WO2004025306A1 - Computer-generated expression in music production - Google Patents

Abstract

A matter performer (MP) system is disclosed for allowing a user to automatically create performances of musical compositions with human expression qualities based on performance-pertinent analysis of musical compositions, uniquely integrating various distinct algorithms for both the analysis and its rendering into a performance. Additionally, the system includes an editor (334) that allows a user to edit the performances generated by it or by an outside source. The system includes a processor (333) for analyzing music data and a processor for setting performance attributes based on the analysis and user preferences, and several interfaces for manipulating and creating music data and performance data. The system also includes a moments database (337) for storing music data, analysis results and performance data.

Description

COMPUTER-GENERATED EXPRESSION IN MUSIC PRODUCTION

FIELD OF THE INVENTION

The present invention relates to methods for computer-generated music. More particularly the present invention relates to methods that enable the introduction of many forms of expression into computer generated music production systems, moreover redefining the relationship between writing, realizing and listening to music.

BACKGROUND OF THE INVENTION

In the past music production was completely dependent on human performers playing acoustic instruments. Today anyone can produce music right on his or her desktop. In fact, the computer has become the dominant component in the commercial music production industry. Music software and hardware enable sophisticated manipulation of both computer generated and human generated music data.

Computers can actually achieve "perfection" that is beyond human physical capabilities. This is especially important in soundtracks for feature films, television programs, video presentations, web sites and commercials, where the integration, in perfect timing, between the pictorial and musical message is absolutely essential.

A sequencer is a device or software that saves the representational information of music, and enables it's editing and performance via sound generators such as samplers and/or synthesizers. Modern sequencers also incorporate digital audio recording and editing capabilities, which turns them into a complete virtual studio. E.g., a performance that is played on a keyboard with a MIDI (Musical Instrument Digital Interface) connection can be recorded by the sequencer and can be transmitted to a "tone generator," i.e. a sound module, which supplies the actual sound(s) for playback.

MIDI is a standard that allows musical instruments to exchange musical data. MIDI, like notation, represents the idea of music. WAVE represents the acoustical phenomenon of music as captured by a microphone. MIDI is used to play multiple instruments from a single keyboard.

However, since information that has been converted into MIDI data can be recorded on a computer and then re-transmitted for automatic playback, virtually all of today's computer- assisted music systems use MIDI.

The MIDI IN connector receives MIDI messages, and the MIDI OUT connector transmits them. When the MIDI OUT of a keyboard instrument etc. is connected to the MIDI IN of a tone generator, playing notes on the keyboard will cause the tone generator to sound. The other connector, MIDI THRU, re-transmits any data that is received at MIDI IN, so that a single keyboard can be used to control two or more tone generators.

MIDI can use a single cable to send multiple streams of musical data - for example, the musical data for a piano performance, a guitar performance, and a bass performance. This is made possible by the concept of "channels." Each channel handles the performance data produced by one performer. Since sixteen channels of MIDI data can exist in a single MIDI cable, it can transmit data that is the equivalent of sixteen musicians performing. Of course it is impossible for one user to play sixteen people's parts simultaneously by oneself, and that's why one needs a sequencer. A sequencer enables the recording of a performance, one channel at a time, and then playing back all the channels at once to produce an ensemble, even though there are no performers other than the one user. This by itself may not seem very different from the capabilities of a tape recorder, but there are some major advantages to using MIDI. I.e., after recording, the user can change the tempo without affecting the pitch, and is free to edit the individual musical parts to personal taste, since each part is usually assigned to its own MIDI channel, thus separating that part's MIDI data from the other parts' MIDI data, based upon the MIDI channel in each MIDI message.

The sequencer records the MIDI data as to which notes were played and how strongly. Even if the tempo is changed, this will only affect the rate at which the notes are sent out-but not the pitch. The pitch, timing, and force of each recorded note can be edited on the sequencer, so that even if the user cannot play a song perfectly in real-time, a great sounding performance can still be produced. MIDI can also handle data to regulate the expression of a performance and to modify the sounds, so that the user can remix the data to create effects such as fade-in and fade-out, or to change a part originally played on a piano so that an organ plays it.

One can compare recording a performance with copying a written text. Using a tape recorder to record a performance is comparable to writing by hand on paper. Although it might be possible to copy the idiosyncrasies of a person's handwriting to some extent, it's difficult to make corrections, especially if using a pen. And if a photocopy is made of the written text, and then a copy of that copy, etc., the copies become progressively less legible. When using a word processor one can easily make changes after writing the text, and even change the font. Similarly, it's easy to make a backup copy of a performance that has been recorded on a MIDI sequencer. However, unlike handwriting directly on paper, a word-processed document needs to be transferred to paper by printing it on a printer. In the case of MIDI, the tone generator corresponds to the printer in the analogy. The process of printing corresponds to the action of transmitting the MIDI data to the tone generator to cause it to produce sound.

Usually a keyboard is used to input the musical performance data. As long as it's MIDI- compatible, any electronic piano, or portable keyboard, can be used for input, or if the user is not a keyboard player, a guitar controller or wind controller can be used to input a performance. One can even use the QWERTY keyboard and mouse of a computer to input a performance instead of a conventional musical instrument. In order for the input musical data (MIDI IN) to be heard as sound, a tone generator needs to be connected to the MIDI OUT.

During the early 1980's, a lot of "pop dance music" was using sequenced phrases played on synthesizers. MIDI was perhaps the first true effort at joint development among a large number of musical products manufacturers. Apple computer made a MIDI interface available for its Macintosh computer, and started promoting the computer in the music market during the mid-to-late 1980's since that was a market completely ignored by their biggest competitor IBM, and one that offered them many potential sales, being that musicians loved the flexibility and ease of use of the new, computer-based "digital sequencers". MIDI took off then.

By 1985, virtually every new musical keyboard on the market had a MIDI interface. All of the musical manufacturers supporting MIDI agreed to start a new organization called the MIDI

Manufacturer's Association (MMA). Everyone started creating MIDI sequencer software that could read/write each other's data files and a standard was created for synchronizing the playback of various sequencers.

Examples of modern sequencer software are CakeWalk's Sonar™, Steinberg's Cubase™ and E Magic's Logic Audio™. Sequencers are perhaps the most versatile tools in music production.

Fig.1 is a prior art example of step entry on Cakewalk's staff view. By moving the pencil 100 over the staff where a G note would occur on the second beat, and clicking once, a G note 110 is seen to be inserted. Because the quarter note length button 120 is depressed, G note 110 of a quarter note is inserted for that beat. A specialized breed of sequencers called "Algorithmic Composition programs" has the capacity to create a complete musical arrangement. Such a program typically allows one to hear, i.e. play, the arrangement that it creates, and even edit the arrangement somewhat. The program composes all of the individual musical parts, for example, drums, bass guitar, piano, etc, parts. One need specify only the chord changes i.e., what chord plays upon which musical bars and beats, what instrumentation and what "style" of music is desired. The program usually offers a variety of rock, jazz, country, waltz, rap, Latin, etc, "styles" of music. Digital audio recorders do not have such a feature.

Most computers have installed software programs, which turn the computer into a sequencer. With a sound card installed inside of the computer, the sequencer can playback musical performances without even needing external MIDI sound modules, since most sound cards now have an internal, multi-timbral General MIDI module, usually a wavetable synth, that can recognize and properly "play" the MIDI messages that the sequencer outputs to the sound card's driver.

An example of a very simple sequencer is Windows Media Player™. This software does not record MIDI messages, nor edit them. It can only playback a MIDI performance (stored in general MIDI File Format). Such simple "playback only" sequencers are often called "MIDI Players".

A MIDI file is a data file. It stores information, just as a text or ASCII file may store the text of a story or newspaper article, but a MIDI file contains musical information. Specifically, a MIDI file stores MIDI data - the data, or commands, that musical instruments transmit between each other to control such things as playing notes and manipulating sounds in various ways.

The only way to record and digitally store an acoustic musical performance is to digitize the analog audio output of all instruments as captured by a the microphone, while that performance is happening, and save it as a WAVE file, for example. The result will be a typically large file that represents the digitized "sound" of all the instruments playing the musical piece in real-time. The act of doing so is referred to as "digital recording", which today is done directly to a large hard drive while the Digital Analog Converter (DAC) digitizes the performance. This is known as "Hard disk recording." Prior to today's hard drive recording there were DAT (Digital Audio Tape) machines that recorded the digitized analog information from the microphones first to TAPE, which than was recorded by programs such as Sound Forge™ to the hard drive for editing, and than re-recorded back to tape. Some Modern software, such as Image-Line's Fruity Loops™ and Sonic Foundry's

Acid™ are loop-based sequencers that have evolved from the drum machines of the early

1980's, that is to say: the basic building block of a sequence is a group of MIDI notes, as with

Fruity Loops™, or a WAVE chunk of an audio recording, as with Acid™. The programs edit these building blocks rather than the individual notes or sounds.

An MP3 file, like a WAVE file, stores digital audio data. So a MIDI file and an MP3 file are different in exactly the same way that a MIDI file and a WAVE file are different. Indeed, a WAVE and MP3 file are two different ways of storing the exact same type of data. The primary difference between a WAVE and MP3 file is that the latter uses compression to squeeze the data down in size, typically resulting in a much smaller file size, particularly as a DOT.WAV, or ".WAV" file for transmission over the Internet .

A computer is capable, for example, of playing quarter notes for hours with each one having exactly 24 clock pulses in it. No human being, no matter how accomplished, can duplicate this feat. At a tempo of 100 beats per minute (BPM), each clock pulse would be .025 seconds long, and you would have to have exactly 24 of them per each quarter note. So, each quarter note would have to be held exactly 0.6 seconds. Humans are not capable of repeating such small timing intervals with perfect accuracy. One quarter note might actually span 26 pulses, the next might work out to be 22 pulses, etc. One can manage to fit 100 quarter notes into a minute, more or less (100 BPM), but will almost certainly not be accurate to a fraction of a second. One will keep getting ahead of, or behind, the ideal clock interval for each note. This imperfection is commonly known as the "human feel". Human brains are even designed to tune into this phenomenon, and reject anything that repeats itself without any variation. This observed human characteristic is a fundamental principle of psychoacoustics. So, the best way to make the human brain become bored by a particular musical passage is to cause the computer to adjust all of its events to perfect clock intervals. Although the brain might not be able to recognize exactly what is going on, it will "know" that the performance is being rendered by a machine devoid of the "human feel".

La grande encyclopedie Larousse, Paris, 1975, defines music as: "language of sounds which permits the musician to express himself". Standard musical Notation is defined as "a visual analogue of musical sound, either as a record of sound heard or imagined, or as a set of visual instructions for performers," which fulfills "the need for a memory aid and the need to communicate." It is the role of the performer to bring to life the "text" of music. This makes the performer an essential link between the composer's work and an audience. Gaining the knowledge of the language, and the ability to carry out the physical movements for actualizing the sounds of the music depends on talent and requires extensive training.

Therefore, so-called "perfect" timing is in direct contradiction to the concept of "meaning" in music. The claim is that humans have a capacity not only to identify "mechanical" timing and reject it as devoid of expression, but rather that humans have actually developed a complex system of perceiving connections between notes based primarily on the linear, or melodic, and vertical, or harmonic, relationship between them as well as even the minutest duration, timing and loudness ("velocity" in MIDI) differences.

Quantization is the technique of "correcting" note timings to specific, i.e., perfect clock intervals. It can affect such things as rhythm, phrasing, and embellishments like trills, glissandos, vibrato, etc. Musicians often use quantization to "correct" rhythmically sloppy playing, not realizing that this also removes the human nuances that our brain favors, when judiciously applied.

Resolution means the maximum number of clock pulses, i.e., the smallest units of time that can occur within a given span of time. Most sequencers specify resolution in terms of Pulses Per Quarter Note (PPQN). This tells how many clock pulses (clocks) there are in every quarter note. For example, a PPQN of 24 means that each quarter-note spans 24 clocks. Roland, one of the largest musical instrument manufacturers, conducted tests that suggest that a minimum of 96 PPQN clock resolution is needed to capture most nuances of a human performance, and a resolution of at least 192 PPQN is necessary for capturing the most subtle human "irregularities" which clue our brain into the fact that a human is performing the music. It has been found that anything less than 192 PPQN is not adequate, whereas 240 are ideal. A few sequencers allow a form of "half-quantization," which only corrects the most "rhythmically off" events, and then only by a random or partial amount. This is perfectly acceptable, since it doesn't "wipe out" the subtle human element, but rather, corrects the most grotesque mistakes, and not in a computer perfect manner. However, without an understanding of the relationship between the notes in the musical structure any "filtered" quantization may be only a pseudo solution.

This explains why computer generated music is often criticized for lack of expression. When compared to a live human performance, computer generated music produces a MIDI event list that sounds "cold", technical and inhuman. Today, there are no "intelligent" solutions for creating expression in computer-generated music, or for enhancing the interpretation in prerecorded music.

Dr. Manfred Clynes has identified what he says are unique patterns in the music of master composers. These patterns, or "pulses," are composer specific according to his theories and show how a composer favors louder and longer notes at consistent points in a phrase, much as poets resort to patterns of phrasing or painters favor certain brush strokes. Dr. Clynes argues that similar patterns exist on higher levels in musical compositions, like the way composers stress certain bars in a movement. He claims, for example, that performances of Beethoven, carry his characteristic strength and intensity when, in four-note sequences, the second note is played shorter and softer than the others, whereas the notes in Mozart's sequences are more consistent in length and amplitude.

Dr. Clynes uses the pulse to avoid the mechanical effect of computer-generated music, but he also relies heavily on another of his theories: predictive amplitude shaping, which calculates the volume of notes in a phrase. For many instruments, like a violin, individual notes can be shaped differently, depending on how the amplitude rises and falls during a note. A note can have a gradual build-up or can diminish quickly. The next note affects the shape of a note's rise and fall. If the next note is a higher pitch, then the amplitude will fall more sharply at the end. If the note that follows is lower, the amplitude will fall more gradually. In addition, the shorter the time between the beginning of the first note and the beginning of the next, the more pronounced the rise or fall.

Johan Sundberg has also developed rule-based performance theories. For example, the contrast between long and short note values, such as half-notes and eighth-notes, is sometimes enhanced by musicians. They play the short notes shorter and the long notes longer than nominally written in the score. Music also has a hierarchical structure, so that small units, such as melodical gestures, join to form sub-phrases, which join to form phrases etc. When musicians play, they mark the endings of these tone groups. This rule marks the phrase and sub-phrase endings by creating minute accelerandos and ritardandos within phrases and sub-phrases according to a parabolic function. Thus it increases the tempo in the beginnings, and decreases it towards the endings. The loudness is changed similarly creating crescendos and diminuendos.

Composers and music production professionals have been using sequencers in an attempt to add a "human touch" to computer-generated music. However, using sequencers for this purpose is difficult technically, very tedious and time-consuming. Moreover, the outcome depends entirely on the musical talent, intuition and analytical skills of the people using these tools. This inefficiency has, therefore, a significant impact on the "bottom line" of music production for CDs, film, television, commercials, etc.

There is clearly a gap in the art, despite some progress that has been made in the last decades.

The expressiveness of electronic music, whether it is performed on a synthesizer or by a computer MIDI sequencer, cannot be compared to that of a live acoustic performance. Using even the best electronic equipment, no one has yet been able to produce convincing violin solos or a virtuoso guitar accompaniment solely with the computer. How does a live musician play, and what underlies performance expressiveness? What does the computer have to learn?

Musicians consider that the essential principle of expression is in the gift of creativity and insight. But there are other important factors expressive playing requires. They are the design of an instrument, the psychophysiology of a performer, and general laws of intoning and performance devices worked out by practice of performance that are appropriate to a particular instrument, genre and manner. There is also a musical language that is learned by a musician and is understandable to a listener. Music is played by rules, sometimes quite plain, sometimes subconscious for a musician, but absolutely unknown to a computer. First of all computers must be taught the basics of musical language and rules of producing music.

Computers have to be taught performance modeling. Conceptually, a score is only a plan for performance. A computer, as well as a human musician, must study a score in order to be able to play it with the necessary nuances, such as: intonation; volume and intensity changes; tempo; timing; articulation; timbre and so on. These elements of performance can be considered as the performance parameters that transform the source data as represented by the score or a quantized MIDI file into what we perceive as expression.

But how can one define these functions for the computer, so that the effects are not arbitrary, but rather intentional and repeatable?

For over 50 years academic research has been attempting to conceptualize the principles of musical expression. Yet, to date the gap between computer and human generated music has never been bridged. Therefore, there is a need for a solution that can integrate human musical expression and interpretation into the computer. The ideal solution will enable not only the transformation of human expression elements into computer-generated music, but will also embody the collective knowledge and intuition of the world's most talented and celebrated musicians. What is needed is a performance-modeling program that is able to observe the rules of musical language, as well as a live performer's interpretation of compositions that are manifestations of the myriad styles of expression that are possible within the boundaries of this language.

A performance-modeling program, such as the one developed by NTONYX, analyzes a musical part in an electronic score, recognizes certain Musical Objects (MO), i.e. notes and their combinations, and then transforms them. The NTONYX Style Enhancer™ copes with the task by recognizing and transforming MO's according to certain rules. The program sets up certain values in the Velocity, Start Time and Duration parameters, and automatically defines Pitch Wheel, Expression, Modulation and other functions. The most interesting results from a musical point of view are in tiny changes of note onsets, fine drawings of volume curves, modulation and tuning inherent to a live performance, which are impossible to reproduce by traditional methods.

Thus, some progress has been made in the art towards computer analysis of the content of musical composition. However, more progress is needed.

Furthermore, despite the large variety of high quality sound production, editing and processing possibilities available today, there are no adequate expression creation capabilities.

Assuming that the ultimate goal of any music production is some kind of expression, the question arises: why do people work with computers in music production at all? The answer is: computers provide people with the kind of control over the material that is vastly superior to what is possible without the computer.

In fact, people in the music business already spend enormous amounts of time, money and energy working with computers on the essential parameters of music expression, such as intensity and timing, in order to obtain results, which cannot be achieved otherwise. However, using existing tools to add "the human touch" to computer-generated music is so difficult and time-consuming that it may even justify, in some cases, reverting back to hiring live musicians, which is costly both in time and money, as it depends on personal abilities, extensive training and often requires lengthy preparation time. Notwithstanding, the benefits of using computers encourages producers to use it whenever possible despite the current technological shortcomings. In short, as far as the ultimate goal of music performance is concerned - creating music expression - there is no technology that can rival the expressions created by live musicians and existing computer tools are primitive and inefficient.

Therefore, there is a need for a method that: analyzes the structural content of any musical composition; correlates this analysis with performance parameters; and qualifies and quantifies these performance parameters in accordance with all the attributes of notes that are relevant to music expression.

SUMMARY OF THE INVENTION

Accordingly, it is a principal object of the present invention to overcome the limitations of prior art computer-generated music systems and provide a method that solves the need for computer-generated analysis of musical compositions as it relates to the performance of music.

It is another principle object of the present invention to provide a method that applies the results of computer-generated analysis to the automatic integration of expression, commonly referred to as the "human-touch," into a computer-generated music system hereinafter called the "Performance Framework."

It is also another object of the present invention to provide a variety of default and optional performance templates.

It is yet another object of the present invention to provide a method that enables a human operator to further manipulate the expression automatically generated by the computer.

It is another object of the present invention to provide a method that enables a human operator to create expression using "smart" tools that refer to general music analysis and the particular analysis of the specific piece. It is also an object of the present invention to provide a method that modifies an existing music file to sound as if a human being is playing it.

It is a further object of the present invention to provide the option, hereinafter referred to as "Saved Preferences" to be "learned" by the program for implementation appropriately in similar instances and induce the "Saved Preferences" to generate performance nuances, thereby enabling the user to automatically apply a "personal" touch to the "Performance

Framework" as well as subsequent editing.

It is still another object of the present invention to provide a method that upgrades and improves digital music production.

It is still a further object of the present invention to provide a method that improves computer music technology serving composers, music producers and performers.

It is yet a further object of the present invention to provide a user-friendly, visually oriented interface for handling expression in computer music technology.

It is yet still a further object of the present invention to provide a method that allows a user to modify the performance qualities of a music file.

Moreover it is an object of the present invention to provide a method that allows a user to choose from a variety of performance templates, to provide master performances of computer generated files - like great pianists, great conductors, musical era oriented performances, specific style performances etc.

A master performer (MP) system is disclosed for allowing a user to analyze a musical composition into functional groups. The system includes a processor for selecting templates and options. The system also includes an algorithms database for storing the analysis and processing algorithms, a functional groups database for storing the functional groups and an editor for editing musical format files.

Definitions and acronyms:

Notation software - a program that allows writing scores into a music-oriented editor; Musical time - an abstract standard of measurement associated with the representation of notes and rests in music notation, expressed in real numbers and signifying the amount of time assigned to each (note or rest). The units used to indicate musical time are whole (corresponding to the number 1 ), halves, quarters, etc; Tempo - a number that specifies the ratio between units of musical time and units of actual time;

Maximal Musical moment - a moment in music where a new event begins - a new note is struck or a note is released. All notes and rests at that moment are considered to be part of that musical moment. The end of the maximal musical moment is determined by the appearance of a new moment. (That is to say, it lasts between the appearances of two events);

Musical moment - any interval of musical time that is included in a maximal musical moment; Beat - The temporal unit of a composition, as is indicated by the (real or imagined) up-and- down movement of a conductors hand. (Harvard Dictionary of Music); Strong and weak Beats - in duple meter the odd beats are defined as strong and the even ones as weak, (starting from 1 ). In triple meter, only the first beat is defined as strong;

Strong and weak measures - Measures are usually grouped by powers of 2. In such groups of measures the odd ones are defined as strong and the even ones as weak. This arrangement may be overridden by the occurrences of cadences and end-of-phrases; Beat hierarchy - the note length division, i.e.: 1/2, 1/4, 1/8 etc; Consonance, Dissonance - "The terms are used to describe the agreeable effect produced by certain intervals (consonant intervals, e.g., octave, third) as against the disagreeable effect produced by others (dissonant intervals, e.g., second, seventh), or similar effects produced by chords. [...] In spite of numerous efforts no wholly satisfactory explanation and definition of consonance and dissonance has yet been found." (Harvard Dictionary of Music).

Performance of music is sensitive to the perceived tension between dissonance and consonance. In addition, the consonance or dissonance of intervals or chords is actually dependant on style.

In each style, each dissonant interval has a finite number of possible solutions (in classical style, for example, these are usually consonant intervals in which one of the tones is a second away from one of the tones of the dissonant interval). Each resolution of a dissonant interval is designated as a resolution of one of the dissonant interval's notes, or of both. The present invention defines a dissonant chord to be a chord that includes a dissonant interval, and a dissonant note to be a note participating in a dissonant interval, which has a resolution, designated as a resolution of this particular note. The present invention accommodates different "dissonance schemes", each of which consists of a classification of intervals into dissonant and consonant (the consonant intervals are those not specified as dissonant), and a list of possible resolutions for each dissonant interval.

The present invention puts forth the following dissonance scheme, which pertains to classical music and contemporary popular music - "dissonance in classical music": An interval is dissonant if and only if it is

• augmented, diminished, a second, a seventh • a fourth against the bass, or

• a fourth with the bass being a major second beneath its lower tone.

Appoggiatura - "An appoggiatura is usually the upper or lower second of a harmonic tone, played instead of this tone on the beat and resolved afterwards into the proper tone" (Harvard Dictionary of Music). "A 'leaning-note'. As a melodic ornament, it usually implies a note one step above or below the 'main' note. It usually creates a dissonance with the prevailing harmony, and resolves by step on the following weak beat" (Grove Dictionary of Music). An embellishing note or tone preceding an essential melodic note or tone (Webster Dictionary). Despite the lack of a single definition of Appoggiatura there is no disagreement among professional musicians as to its meaning in practical applications. In performance practice Appoggiaturas are usually emphasized by increasing the dynamic level on the dissonance and decreasing it on the resolution as well as stretching out the time on both the dissonance and the resolution.

The present invention puts forth the following formal definition of Appoggiatura. Appoggiatura consists of

• A dissonant moment

• A specific note within this moment that is subsequently resolved (the dissonant note)

• The moment of resolution (a moment of resolution is the resolution on the harmonic level) referred to as AppoChord

• A specific note within the moment of resolution or prior to it that is the melodic resolution referred to as AppoNote of the dissonance (this note may not actually appear, in case the note of resolution appears in a different layer - then it is not an apponote, it is merely the note that makes the appochord an appochord) The present invention uses a software/hardware system, hereinafter referred to as the "Master Performer," which analyzes music data and automatically assigns values to all the parameters that are relevant to music expression. The technology improves and upgrades all computer-based music creation systems by enabling commercial media production houses, composers and independent music studios, recording, editing, processing and mastering professionals to:

significantly simplify and improve the process of music production; significantly shorten music production time; and achieve performance quality that is superior to that of most live musicians.

These benefits provide financial savings by substantially cutting production processing time, reducing and even eliminating the need to hire live musicians for recording sessions, which, in the case of a full orchestra, could amount to many hundreds of man-hours. The Master Performer technology is based on an innovative approach to the process of transforming music from the representational format of MIDI or standard musical notation into expressive performances that unfold in real-time.

The program analyzes the specific musical content of any given piece of music and applies flexible quantitative models to the principle parameters' of Duration, Velocity, (or intensity), and Timing, (as well as many other parameters that are instrument specific), of each note and creates a Performance Framework.

The Master Performer does not impose a rigid, pre-defined interpretation upon the music, but allows the user to edit the framework, choose alternative expressions from Template libraries or implement Saved Preferences that the program "learned" from the user.

The technology is an outcome of research into the relationship between linguistic and expressive properties in music, which has yielded a practical software application based on a unifying formulation for both the linear, or melodic and vertical, or harmonic, organizational principles of sounds in western tonal music.

The Master Performer is a leap forward in computer music technology benefiting the professional music production industry as well as the millions of amateur and semi-professional music lovers and enthusiasts. The technology complements and upgrades existing music creation software and hardware and will be available as a stand-alone product, plug-in or rewire to notation, sequencing or loop creating software and as a feature for hardware OEM.

Applicable hardware includes synthesizers, controller keyboards, digital pianos, electronic reproducing pianos, sound modules, samplers, etc. The software includes notation programs, sequencers and loop creators.

The end users of the Master Performer will be professional music creators, such as music production studios, composers, arrangers, music directors for film, television and video, Web and other "new media," dance & theatre companies, as well as music educators, music students and music-hobbyists and enthusiasts.

The Master Performer demonstrates the fallacy of associating "the human feel" in music with either human imperfections or superior qualifications. In fact, human emotional cognition is mostly intuitive and specific technical skills are learned and practiced in order to gain the ability to carry them out. The Master Performer utilizes innovative methods of tracing this cognitive process, thereby offering clear performance options, which in turn it carries out flawlessly.

Identifying a significant shortcoming in the process of computer based music production, the Master Performer addresses the specific need to create and control the expression in computer-generated music. It is the first comprehensive technological solution that transforms music from representational formats such as Notation or MIDI into truly expressive performances based on analysis of music as it relates to expression, which offers flexible quantitative performance models that unfold in real-time.

The Master Performer's algorithms analyze the rhythm, harmony, patterns and structure of the music data and identify units akin to words, sentences and paragraphs as the basic building blocks of music, rather than the individual notes, which like individual letters, do not possess in and of themselves the minimal requirements for composing meaningful statements.

Thus, the computer is enabled to participate in the intelligent application of all the performance parameters that are relevant to the creation of music expression. An innovative set of fundamental principles has been formulated to define the possibilities of expression in music. These principles combine basic music attributes, such as tempo, timing, dynamics and articulation with the make up of music's building blocks and the systemic structural connections between them.

BRIEF DESCRIPTION OF THE DRAWINGS

For a better understanding of the invention in regard to the embodiments thereof, reference is made to the accompanying drawings and description, in which like numerals designate corresponding elements or sections throughout, and in which:

Fig. 1 is a prior art screenshot of music sequencer software;

Fig. 2 is a screenshot of the Editor, where the user can manipulate the default values that have been assigned automatically to a piece of music by the program, in accordance with the principles of the present invention;

Fig. 3a is a flowchart of the inputs to the Editor, where the user can manipulate the default values that have been assigned automatically to a piece of music by the program, in accordance with the principles of the present invention;

Fig. 3b is a flowchart of the progression of the Master Performer system leading to a complete digital performance output, in accordance with the principles of the present invention;

Fig. 4 is musical notation illustrating the concepts of appoggiatura, appochord and apponote, in accordance with the principles of the present invention; and

Fig. 5 is graph illustrating the concept of the To-Beat for relative tempo vs. musical time, in accordance with the principles of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The invention will now be described in connection with certain preferred embodiments with reference to the following illustrative figures so that it may be more fully understood. References to like numbers indicate like components in all of the figures.

A strict rendering of musical scores sounds mechanical and expressionless, like reciting a written text without inflection and punctuation. The algorithms of the Master Performer analyze the rhythm, harmony, patterns and structure of the music data that is presented in the score, and identify units akin to words, sentences and paragraphs. This is done in two steps: dealing exclusively with the mathematically quantifiable properties that effect the inflection and punctuation of music, the program automatically calculates values for all the relevant attributes of every note and creates a performance framework for MIDI playback; and the user is able to create a personal interpretation by modifying the framework, editing it with "smart" tools or choosing an optional expression template from a database.

By separating syntax from expression, the Master Performer technology emulates the complex, yet largely intuitive human process.

Recognizing that seamless fusion of "mind" and "heart", in a flow of inspired music making, unhindered by technical limitations is the exclusive province of human "master" performers, the program still in all offers:

the professional musician freedom from the dependence on superior dexterity; a means for self-expression that bypasses the enslaving demands of acquiring, developing and maintaining the physical and mental nimbleness that is necessary for matching his or her potential for emotional and intellectual expression; and non-professionals, amateurs and even novices an opportunity to experience music making while bypassing the arduous path that is required in order to gain the minimal proficiency of a professional musician. This is accomplished via a library of templates and editing possibilities, which encompass the cumulative theoretical and technical background, practices and experience of several centuries, yet are presented in a clear and simple way.

Fig. 2 is a screenshot of the user interface Editor, where the user can manipulate the default values that have been assigned automatically to a piece of music by the program, in accordance with the principles of the present invention. The user can edit values for velocity 210, duration 220, timing 240 and other relevant parameters for every note. Fig. 2 illustrates a performance framework, which is then further enhanced by actual human expressive input. In the piano piece illustrated in Fig. 2 as seen from top to bottom the: velocity (loudness) view (blue background) 210; special notes view (white background) 220; quantized tempo view (orange background) 230; timing view (green background) 240; and pedal view (beige background) 250.

Measure numbers 260 are also shown at the top of the special notes view 220, which combines features of standard notation with Master Performer exclusive features. Different voices are represented by various respectively colored small circles with connecting lines indicating groups in velocity view 210 and by the same respectively colored small triangles along with their Duration lines in the notes view 220.

For example, in Fig. 2 there are 4 voices represented by the blue circle with lines 212, red circle with lines 214, purple circle with lines 216 and dark yellow circle 218. This yellow circle is connected to a group that started earlier and can be seen in the previous window. The 4 individual red circles 214 in velocity view 210 between measure 34 & 35 (260) are connected, which means that these 4 notes belong to one group 270. All the same notes appearing as circles of a particular color in velocity view 210 also appear in the same color in standard note view 220. Standard note view 220 adds length values (Duration) to each note by extending a line to the right of the small triangles.

For example, the voice (meaning an actual human voice or, more often, the "voice" of an instrument), represented by dark yellow circle 218 in velocity view 210, actually has two counterparts in standard note view 220: an expected dark yellow triangle with a length value of a quarter note 280; and a purple triangle above it (i.e., at a higher pitch) also having a length value of a quarter-note. Thus, dark yellow circle 218 in velocity view 210 is "hiding" a purple circle because they have the same velocity. Also, specific purple circles 216 are shown as a series of 4 eighth notes of differing pitch 285.

Figs. 3a and 3b show how the software algorithms of the present invention are coordinated with existing hardware and software. Fig. 3a is a flowchart of the inputs to the Editor 335, where the user can manipulate the default values that have been assigned automatically to a piece of music by the program, in accordance with the principles of the present invention. Fig. 3b is a flowchart of the progression of the Master Performer system leading to a complete digital performance output 370, in accordance with the principles of the present invention.

All of the primary inputs 310 in Fig. 3a are files that can be saved also as MIDI files. For example, an existing plain MIDI file 311 may be used. Alternatively, a composer may work with prior art forms of "notation" software 313, like Finale or Sibelius, and can prepare a MIDI file of a human-like performance using the Master Performer processing system 330, then run it through a sound module 340, apply processing and mixing 360 to the resulting audio file 350 ("WAV." file) and have it all ready in the form of complete digital performance output 370 for presentation to producers within hours, in contrast to the weeks or even months required for prior art systems involving live musicians and physical recording venues.

Musicians working with sequencers that produce a sequencer specific file 315 can add the missing human touch to an otherwise mechanical rendering of a composition by simply running it through processor 333 and choosing an expression template 334 from the Master Performer's template database. The Master Performer database is independent, and translators enable plug-in into existing programs or rewire, which means one opens whichever program one is used to using, such as Sonar™ or Finale™, and within that program opens the Master Performer.

A MIDI keyboard 317 is a keyboard instrument, which looks like a piano keyboard. The user plays on the keyboard, and a built-in signal processor sends the MIDI information of the playing to the Master Performer processing system 330. A synthesizer keyboard 319 having piano keyboard-type keys functions similarly. Thus, there are means for human live playing input of any MIDI-equipped instrumentation.

Once all the inputs are collected by Master Performer processor 333, and the user makes choices of an expression template 334 and other options 336, Master Performer carries out an analysis in Processor 333. There are three basic modules: translation, analysis and performance. The translator translates input data from software specific representations (for example: Finale: MUS. File, Cubase: .ALL file, standard .mid file, etc.) into Master Performer's own common representation (moment data base). The analysis and performance modules then use the output of this module. The analysis module analyzes the moment database and marks the implied harmony, the groups, patterns etc. Then the performance module uses the output of the analysis module to modify the performance attributes of the musical data. The parameters that are adjusted are the loudness, or intensity, in velocity view 210, the timing in timing view 240 and the duration in the notes view 220.

What a human player would do intuitively, the performance module does using mathematical formulations that have been discovered in developing the present invention. Along with duration in the notes view 220, the note onset can be adjusted to appear slightly before or after the beat. . All the notes in Fig. 2 begin at the beginning of a moment, except the three notes 222, 224 and 226. These have a slight off set and are played slightly ahead of the beat. These type of "inaccuracies" are typical of human playing. Such deviations, which the program applies judiciously, based on the user's preferences, add further subtle human expression. Thus, the invention is a sophisticated tool for a human user for implementing creative and expressive choices and allowing further editing of these choices, while saving considerable time and effort.

According to score indications, default options or preferred user preferences the Master Performer will adjust the Duration of each note. For example, in half-measure 280 in the notes view 220, the purple and dark yellow voices are exactly a full quarter, but the blue voice is somewhat shortened.

It has been discovered that one of the most important means of musical punctuation is the slowing down, with or without intensity diminishment or enhancement, at the end of a group of notes. A piece is seen to comprise a hierarchy of groups of various sizes, analogous to the words, sentences and paragraphs of a textual composition. Smoothing is most critical. Thus, the slowing down process must be equalized. Otherwise, each subsequent group would start slower, and in time the entire section would become unbearably slow. Thus, subsections are judiciously slowed down and speeded up again within the framework of each sub-section at each level, thus maintaining the global tempo in an overall musical "story." The same applies to the velocity. Discovering, and automatically applying, rules for balanced smoothing is a major contribution of the present invention towards adding computer-generated human expression to music. It shows the understanding of the "feeling," as related to time, intensity and articulation in the performance of music.

In Fig.2a the Master Performer distinguishes the melody or voice 1 (in blue), by making it louder than the accompaniment, and therefore it is in the topmost position in the velocity, or loudness, view (blue background) 210. Using Editor 335 and adding saved preferences 337, for example, the user may create a bigger difference between the melody and the accompaniment.

As a further example, the user may decide how fast the music should be. E.g., quantized tempo view (orange background) 230 in Fig. 2 shows a default value of 100%. The user may point and click and change this to 80% or 125%. When the user wants to change the loudness, or "velocity," of a certain note in a group, the program automatically adjusts the velocities of the rest of the notes in the group to realign them with the change. Furthermore, the program may adjust the timing of the group to correspond to the changes in velocity and vice versa. If the user chooses to change a tempo in the timing map, the program will automatically adjust other tempi in the entire group as well as the corresponding velocities, if necessary. The system also learns from stylistic inputs of the user, and can be instructed to repeat them in future usage as new "personal preferences" defaults.

Master Performer processing system 330 generally is rule-based and applies several major rules and many sub-rules. For example, as illustrated by timing view (green background) 240, Master Performer processor 333 may apply a particular rule and automatically slow down at a particular instant.

The user receives an automatic default velocity pattern and has several options: keep the default as is; raise or lower the entire group. The program automatically adjusts the pattern to retain the original relationships between all the velocities in the group; modify the relationship between the velocities. The user indicates a preference by moving one note in the group. The program automatically adjusts the relationships between the notes in the entire group according to the user's wishes and the underlying music principles and also automatically makes the necessary adjustments to the timing; a less professional user would choose an alternative option from an expression table. For example: more aggressive, less sharp. The program would perform the necessary changes automatically; and a novice would probably want to do more global editing and would therefore choose an alternative expression template for the entire section. For example: more delicate. The program would then offer several complete options that would render the passage more delicate.

Method steps of the present invention:

1. Detecting End-of-Group

The following procedure is performed on a single voice (layer). In the context of this rule:

• Note length refers to the aggregate of its actual written length and the rest(s) that immediately follow it.

• The lengths of notes that are melodic resolutions of Appoggiaturas are computed as the sum of their length as defined above, and that of the Appoggiatura's last strong segment

(See: step 3. Detecting Appoggiaturas) as defined above, (i.e.: elongated by the succeeding rests).

1. If a note x is longer than the note that immediately precedes x and the note that immediately succeeds x is not longer than x then x is an end-of-group, with the following exceptions:

• If x ends on a strong beat and is dissonant to the explicit or implicit harmony, or to a note in another voice, and resolves according to the rules of Appoggiatura in a melodic resolution, then the group ends on the melodic resolution of the appoggiatura (see: step 3. Detecting Appoggiaturas: Apponote). • If the note following x is longer than x and is identified as a "new motivic beginning" (see: step 5. Detecting Patterns) than x is the end - of - group.

• If x is inside a slur, or a slur begins on x. This may apply only to certain slurs, depending on factors such as slur length.

• If a pattern such as a sequence or repetition (with or without variations) is detected (see: step 5. Detecting Patterns: Types of patterns) the pattern may dictate an end-of-group at the end of each recurrence of the fragment.

• A recurring small rhythmic pattern (see: step 5. Detecting Patterns: Types of patterns) may dictate extending a group to the point where the patterns stops

2. A cadence (see: step 5. Determining the end - of - phrase) means an end-of-group, unless the note that coincides with the end of the cadence is shorter than the note that precedes it, in which case this point is marked as an overlap: i.e., an end and a beginning of group at the same time. Since under this exception the conclusion of a cadence may be the start of a new group or phrase, the performance implication of the cadence may be modified by the occurrence of such a beginning.

The type of cadence is also detected and stored.

2. Determining the Implicit Harmony

The implicit harmony is determined by comparing the results of two distinct procedures: melodic pattern oriented (M-PROC) and harmonic block oriented (H-PROC).

The algorithm first performs the M-PROC by detecting certain melodic patterns such as Alberti bass, tremolos, arpeggios, etc., and deducing the harmony implicit in them.

Subsequently, the H-PROC looks at different temporal segments of the music and finds the segmentation, which best expresses the harmonic identity of each segment (meaning that each segment should be mostly or to some convincing extent comprised of notes that belong to a single chord). The algorithm compares the results of the two procedures in the context of each segment and determines which procedure is more likely to have detected the correct Implicit Harmony, which then becomes the final output of the entire algorithm.

3. Detecting Appoggiaturas

The following detailed description is an example of a method, by which the algorithm detects relevant attributes of music as defined by the program:

The program detects Appoggiaturas by iterating over all the notes in the music; upon encountering a dissonance on a strong beat the algorithm looks for a resolution on the harmonic level (AppoChord). If such a resolution is found, the algorithm goes on looking for a melodic resolution (AppoNote).

Since sometimes only part of a note (in this context rests that follow a note are considered a part of it) is on a strong beat the algorithm checks for a dissonance not in the first moment of the note but in the first moment of the longest segment of the note such that its end matches the end of the note and this segment correlates exactly with a strong beat. No part of notes that do not have such a segment is perceives as strong.] Example of a Last Strong Segment:

ε- ξ_ ε. ξη = ε. ξ_ ε _ξ ξ η Upon encountering a dissonant note on a strong beat the algorithm suspects an Appoggiatura. A new data structure called "suspected Appoggiatura" is used to store the intermediate data needed to choose between alternative resolutions to the same dissonance. This data structure contains all the data members of the aforementioned Appoggiatura data structure and the note against which the encountered dissonant note is dissonant, called "the co-dissonant note". The first moment of the dissonant note's strong segment is designated as the dissonant moment of the suspected Appoggiatura. The actual note is designated as the dissonant note of the suspected Appoggiatura. The items in the Appoggiatura data structure arrived at by this stage are designated collectively "the dissonant portion of a suspected Appoggiatura". To verify the suspicion of Appoggiatura, the algorithm must find at least a harmonic resolution for the dissonant portion of the suspected Appoggiatura. The harmonic and melodic resolutions of the dissonant portion of the Appoggiatura (which are a moment and a note, respectively) are designated "the resolution portion of an Appoggiatura".

Next, the algorithm looks for the harmonic resolution. A moment is the harmonic resolution of the dissonant portion of the suspected Appoggiatura when this moment does not contain a dissonant interval (except in a half cadence, which may contain a 7^th chord), does not contain a note in the same pitch class as the Appoggiatura's dissonant note, contains an interval defined by the dissonance scheme as the Appoggiatura's dissonant interval's resolution, or contains the dissonant note's resolution note in another octave unless: the moment of resolution contains only one note, a fifth, a sixth or a third and the dissonant interval was a minor or major second or their octave doublings. Having found a dissonant note segment on a strong beat the algorithm looks for a harmonic resolution in the points of time defined by the following series: Yi(i= 1 ,2,...) such that Y_k = the starting point of the segment + 2'^"1. (The length of the segment). At each point Y the following procedure is performed: (i) If the moment at Y qualifies as a harmonic resolution of the dissonant part of the suspected Appoggiatura, then it is designated as the harmonic resolution. Otherwise, (ii) if at Y, there is a rest in the layer of the suspected Appoggiatura's dissonant note then: if there is a moment that ends at Yj and qualifies as a harmonic resolution of the dissonant part of the suspected Appoggiatura, then it is designated as the harmonic resolution. This moment may end a little while beforehand, if for example, the note that is supposed to end at Y| is shortened as an articulation gesture. Else, if the moment at Y_i+1 qualifies as a harmonic resolution of the dissonant part of the suspected Appoggiatura it is designated as the harmonic resolution. Else, if there is such a moment between Yj and Y_i+1, then it is designated as the harmonic resolution. Otherwise, (iii) if the moment at Yj contains a dissonant interval the unresolved suspected Appoggiatura is saved as pending, and the point Y_i+1 is marked as the point in which the algorithm will continue to look for a harmonic resolution for the current suspected Appoggiatura. When a harmonic resolution is found it is designated as the harmonic resolution of all pending suspected Appoggiaturas.

After a harmonic resolution has been found for a suspected Appoggiatura the algorithm looks for a melodic resolution for that Appoggiatura (by examining the notes in the layer of the Appoggiatura's dissonant note between the end of the last strong segment and the beginning of the harmonic resolution).

If, after having found the resolution of a suspected Appoggiatura, the temporal interval between the beginning of the dissonant segment and the beginning of the moment of harmonic resolution does not have a "last strong segment", the suspected Appoggiatura is canceled. In the end, suspected Appoggiaturas are verified in this manner: for every pair of dissonant note and dissonant moment that appear in a suspected Appoggiatura, an actual Appoggiatura is created with the same resolution as was designated for them.

Fig. 4 is musical notation illustrating the concepts of appoggiatura, appochord and apponote, in accordance with the principles of the present invention. The thin, solid arrows 410 indicate Apponotes. The Appochord 420 appears on the 2^nd beat of the 2^nd measure. The thick, solid arrow 430 points to the last strong segment of the note D in the soprano voice (layer) 440. That same note served as an Apponote to the dissonance F (in the bass 450) - E (in the soprano) in the 1^st beat of the 1^st measure, and becomes a new dissonance against the E in the bass on the 1^st beat of the 2^nd measure. That E is itself the Apponote of the Dissonance F (in the bass) - B (in the alto 460) in the 1^st beat of the 1^st measure.

The G in the tenor voice 470 on the 2^nd beat of the 1^st measure is the Apponote of the dissonance A (tenor) - B (alto) on the 1^st beat of the 1^st measure. This G is dissonant to the bass note F, but does not qualify as the dissonant note of an Appoggiatura since it does not have a last strong segment. The Apponote C in the soprano on the 2^nd beat of the 2^nd measure coincides with the Appochord of this complex Appoggiatura.

4. Determining the End - of - Phrase

End-of-Phrase is detected by the presence of: • Cadence

A cadence is defined as " A melodic or harmonic formula, which occurs at the end of a composition, a section, or a phrase, conveying the impression of a momentary or permanent conclusion," (Harvard Dictionary of Music) Examples of Cadences are: Authentic, Plagal, Half- cadence (Authentic or Plagal), deceptive, etc.

A cadence is an end-of-phrase only if it represents a complete harmonic progression or a substitute to it, for example: repetitions of a I - V progression, except for deceptive cadence, which is never an End-Of-Phrase.

If the cadence is immediately followed by the same type of cadence then the end-of-phrase moves to the later cadence.

The type of cadence is also detected and stored.

• A simultaneous end-of-group in all voices.

• Just before a 'New Motivic Beginning' (see: step 5. Detecting Patterns)

Note: If a 'New Motivic Beginning' follows a cadence a new group may not begin in the voice labeled "The melody" (see: step 6. Determining the relationship between the voices) at the end of the cadence.

Note: A cadence may be an end-of-phrase and the beginning of a new group or phrase simultaneously (overlap), in which case the new beginning has an overriding performance implication.

• Fermata means an End-of-Phrase.

Note: if there are several fermatas one after the other, the last is the End-of-Phrase.

Note: End-of-Phrase cannot be the beginning of a sequence or repetition, (see: step 5. Detecting Patterns). Note: Ends of Phrases are used to determine strong/weak measure relationships in the entire piece of music.

5. Detecting Patterns

The term pattern refers to a section of music in which a smaller section of music is repeated as described below in "Types of patterns". The smaller section is called "fragment". A pattern occurs in a single voice unless otherwise stated.

The occurrence of patterns influences the performance in several ways, such as:

i. Direct performance implications; for example: in an exact repetition (a1=a2), a1 can be forte and a2 can be piano, a1 can be legato and a2 can be staccato, etc. ii. If the analysis detects a certain phenomena in some recurrence(s) of a fragment within a pattern the performance implications of that phenomenon may be extended to affect other recurrences; for example: if there is a sequence of recurrences of a fragment and the first one incorporates an appoggiatura the following recurrence(s) will behave like an appoggiatura as well. iii. Indirect performance implications as a result of patterns influencing the analysis procedure. For example, patterns may dictate that an end-of-group occur where it would not be derived from other rules.

Types of patterns:

1. Group Pattern - A pattern in which the fragment is a group. 2. Small Pattern - A recurring pattern within a group. All fragments must maintain the first fragment's strong-weak beat relationship.

3. Exact Repetition - a fragment played twice consecutively.

4. Repetition with Variation - a repetition with small changes.

5. Sequence - a fragment played twice or more consecutively from different starting pitches, maintaining the interval qualitative relationships (i.e. 2nds, 3rds etc. and not major 2nds, minor 3rds etc.) and the rhythm. The interval between starting notes of fragments is fixed. A descending or ascending scale or one-way arpeggios are exceptions and are not to be considered sequences.

6. Sequence with variation - a sequence with small changes. More interval changes are allowed than rhythmic ones. 7. Mirror pattern - a pattern of types1-6 in which a fragment may appear as its mirror image.

8. Rhythmic Dominated Pattern - a sequence with little or no rhythmic variation and any amount of interval variation. 9. Direction Dominated Pattern - a sequence with little or no variation in the direction of the intervals and any amount of rhythmic variation and any amount of variation in the size of the intervals.

10. Canon (with or without variation)- A recurrence of the fragment in different voices. Each recurrence may start on a different pitch, either maintaining the interval qualitative relationships of the first fragment or its mirror image. A canon in which the beginning of a fragment occurs before the end of the previous fragment is called: "Stretto".

11. Motivic Group - recurrence of a group or a number of consecutive groups with or without transposition and/or small changes separated by a non related section of music.

Note: 'A New Motivic Beginning' is a motivic group that has a separating section that is at least twice as long as the group itself, in musical time. 'Reprise' is 'A new beginning' of the first motivic group. 12. Any combination of the above patterns.

6. Determining the relationship between the voices

The algorithm checks the melodic content of the voices according to criteria such as: rhythmic variety, the presence of chorals, canons, accompaniment patterns such as Alberti bass, etc. and rates segments of the individual voices in a number of qualitative criteria such as melodic quality, recurring pattern, etc.

7. Assigning velocity values

Each configuration of ratings as described in section 6 dictates a certain configuration of default velocities for each voice. For example, the algorithm diminishes whatever is labeled accompaniment.

The remainder of this section describes the effect of phenomena detected and determinations made by the analysis on velocities within a single voice (layer). Note: The term velocity is derived from keyboard instruments action (it represents dynamic intensity as a function of the velocity in which the hammerhead travels towards the strings). Other instruments have additional parameters associated with velocity such as: envelope, vibrato, etc. The rules of determining the values of the additional parameters are organic extensions of the rules described in this section.

(For example: a violinist would make a crescendo towards the last strong segment of an appoggiatura's dissonant note).

Each of the following rules (except the acoustic rule) generates a function of relative velocity so that the actual velocity map of the layer is calculated by multiplying all these functions by the default velocity as defined in section 6 and passing it through the acoustic rule. The acoustic rule may also be used before the rule of To-Beats is applied because the rule of To-Beats is sensitive to the results of the previous rules (the rule of default velocities and the rule of Appoggiaturas).

A dynamic intensity graph is associated with each note. This graph describes the dynamic level of the note throughout its entire duration.

Different rules are used for different instruments based on the same principles and dependant upon the nature of an instrument's control over articulation, dynamics, the envelope of its sound, etc. For a piano only constant functions are used representing the force with which the string is struck by the hammerhead (velocity).

Composer markings such as global dynamic indications (piano, forte, etc.) and local indications (such as accent, sforzando, etc.) are used as a default onto which the results of the procedures described here is superimposed.

1. The rule of weak beats relative to preceding strong beats

This rule describes a function d(x), that assigns each note in the layer a "default" velocity value (for x inside a given note d(x) will be its velocity value). This default is based on the assumption that weak beats always relate to the strong beats that precede them. This assumption may be overridden by other rules. Given a note x, which is not on the first beat of an odd cycle of the pulse (in which case the velocity determined in section 6 is retained), we find the imaginary beat y-i of the lowest possible level in the beat hierarchy that is weak and starts at the same point as the given note. For example, if the pulse is in quarter notes in quadruple meter then a cycle of the pulse comprises of four quarter notes, and two cycles of the pulse comprise of eight quarter notes, so that the first beat of each odd cycle of the pulse falls on the beginning of every odd measure (unless the arrangement of cadences and phrases overrides the succession of strong and weak measures: for example after an 7- measure phrase, the pulse-cycle count starts again at 1 in the next phrase). This beat may, in fact, not correspond to the notes "in it", and may even stretch outside the piece. What is required for it to "exist" is that its beginning immediately follow the end of an actual strong beat of the same size.

yi 's corresponding strong beat is called y₀, and y^s hierarchy level L. x is assigned a velocity value which is the velocity of the note that starts on y₀ (in case such a note does not exist, the note that begins the closest to y₀ before y₀ is chosen) times α_L or β , where α and β_L are real numbers < 1. In triple meter the first beat is the strong beat associated with both the second beat and the third beat. If the y₀ is the first beat of an odd cycle of the pulse then the velocity determined in section 6 is used as the velocity at that point.

OIL is used in case the beat that comprises of y₀ and yi is strong, and β is used in case the beat that comprises of y₀ and yi is weak, f the beat that comprises of y₀ and yj is imaginary a default ratio is used. α_L and β_L are defined to meet the following condition: α < α_L+ι < β_L+ι < β for any L. In triple meter "the beat that comprises of y₀ and yi" refers to the beat comprised of all of the three beats (meaning that the beat that "comprises of the first and second beats is the beat that encompasses the first, second and third beats, and this is also the beat that "comprises of the first and third beats).

2. The rule of Appoggiaturas

This rule generates a function a(x) that will be multiplied by the functions generated by the other rules to calculate the velocity of the layer's notes at each point in time. For each Appoggiatura resolution recorded, all the Appoggiatura dissonances belonging to it are examined and divided into groups of the same pitch (such groups will contain more than one note if notes in the dissonance repeat before resolving). For each such group the first dissonance is considered (the dissonance is taken to end at the resolution's beginning); from its beginning to the beginning of its last strong segment each recurrence of the dissonance is assigned a higher velocity factor a(x) according to some rising curve of the user's preference. The velocity at the last strong segment is assigned a user defined velocity factor > 1. The resolution (both apponote and appochord) is assigned a velocity factor < 1. For each dissonance, the note against which it is dissonant is also assigned a user defined velocity factor > 1 , if it occurs at the same time.

3. The rule of To-Beats

This rules deals with To-Beats, which are weak beats that are not the resolution of an appoggiatura, or end-of-group. In triple meter (in moderate to fast tempo), if both the second and third beats are weak then the To-Beat encompasses both, but instead of the linear function described in (ii), a piecewise linear function with two segments (corresponding to the beats) is used so that the angle of the second segment is considerably larger.

The following rules apply to rising To-Beats (To-Beats that lead to a strong downbeat):

(i) To-Beats one level above the pulse level, which contain only one note on that beat are raised in velocity to a level no higher than the following downbeat. (The amount depends on user preference).

(ii) This rule applies to maximal To-Beats of hierarchy levels smaller or equal to the pulse hierarchy level that have notes that start within them. Given such a beat, a function f(x) is created so that for x outside the beat f(x) = 1 (the first point of a beat is considered to be inside the beat), and for x inside the beat it is defined constructively:

• At the beginning of the beat f(x) = either 1 or a higher user defined ratio dependant upon the number of notes in the beat (for one note the ratio will usually be > 1 ) for an extra crescendo effect. From this point, f(x) goes up linearly until the end of the beat so that the left limit of the multiplication of f(x), d(x) and a(x) at the beginning of the next beat will be equal to the value of the multiplication of f(x), d(x) and a(x) at that point. • The number of notes inside the beat is counted, and all the appoggiaturas that start on notes that are among the first two thirds (or a similar section) are marked, except for appoggiaturas whose dissonance's last strong segment is on the second beat in triple meter. For each appoggiatura that was marked, f(x) is multiplied by d(x) and a(x), and if in the resulting function the dissonant note's velocity is lower than the resolution's velocity, the angle of the segment of f(x) that corresponds to the duration of the appoggiatura's dissonant note is replaced so that in the above product function the resolution's velocity will be equal to that of the dissonance. The segment of f(x) that follows the note until the end of the beat is uniformly decremented so that it will begin at the same height as the end of the modified segment.

If among the marked appoggiaturas there are at least two appoggiaturas whose dissonant note's velocity equals that of the resolution, the marked appoggiatura that is second weakest

(meaning that the difference of velocities between its dissonance and resolution is smallest) according to the multiplication of d(x) and a(x) is strengthened by a user defined amount

(meaning that the angle of the segment between the dissonance and the resolution is diminished), and the segments between the dissonance and the resolution of all stronger marked appoggiaturas are changed to the same angle as the segment of that appoggiatura.

Changing the appoggiatura segments of the marked appoggiaturas to the same angle as that of the weakest appoggiatura will achieve a similar result.

• Of course, the remainder of the function following the modified segments is decremented as before.

• Finally the angle of non-marked-appoggiatura segments is changed uniformly so that the left limit of f(x) at the beginning of the next beat will be equal to that of f(x) after step 1. When this rule has been applied in all layers, a corrective step is employed to make sure that an appoggiatura's dissonant note will not be drowned by the effect of this rule in other layers and that the resolution will not stick out.

The following rules apply to falling To-Beats (To-Beats that lead to a weak downbeat):

(i) To-Beats one level above the pulse level, which contain only one note on that beat are lowered in velocity to a level no lower than the following downbeat. (The amount depends on user preference).

(ii) Otherwise, the To-Beat is multiplied by a descending function of the user's preference so that the left limit of the multiplication of this function with all the functions generated by the previous rules at the beginning of the following downbeat will be equal to the value of that multiplication there.

By default, To-Beats are rising. Rule 4 and user preference may override this default.

4. The rule of End-Of-Phrase. This is an exception to the default of rising To-Beats. For each phrase, if it ends on a strong measure, the effect of the last To-Beat before it is augmented to rise more steeply in velocity. Otherwise, the last To-Beat before the End-Of-Phrase falls in velocity. The same is true for End-Of-Group, only the effect is milder.

5. The acoustic rule. In order to compensate for the natural decay of sound in instruments that provide control of the sound envelope only at the beginnings of notes (such as piano, marimba, harpsichord, etc.), this rule increments the velocity of long notes in some cases (governed mostly by user preference) and diminishes all the notes that come after a long note until the nearest opportunity to return to the velocity determined by other rules (this may be the next group, the next downbeat, its To-Beat, an Appoggiatura, etc.). This rule governs directly the actual velocities of the notes, thereby rendering the other rules to the status of an abstract idea, and the functions promulgated by them to the status of recommendation.

8. Assigning tempo values

This section describes the effect on tempo of phenomena detected and determinations made by the analysis. Each tempo rule correlates each occurrence of such phenomena and determinations with a function that represents their effect on the tempo of musical events.

For each such function the x-axis represents musical time and the y-axis represents relative tempo as a percentage of user-determined global tempo. After all these functions have been set, their product is calculated. This product (with several adaptations perhaps), multiplied by the user-determined tempo (to convert percentages to concrete tempo values), is used as the tempo map.

Since it is technically impossible to change the tempo at every point in time, an additional process is required to convert the tempo map to a list of tempo changes to be inserted in the music. These changes are set to occur only at the beginnings of maximal moments, for ease of implementation and simplicity. The conversion is made so that, using the list of tempo changes, each maximal moment would take the same amount of actual time as it would have taken if the original tempo map had been used directly (meaning that the tempo would have been changed at every point in time to the value of the map at that point). The conversion is done in the following manner: The original tempo map is designated as f(x). Given a moment m, we designate its starting point as m_begin and its ending point as m_end (in terms of musical time). The tempo change value assigned to m is the result of the integration of f over x (which is musical time), from m_begin to m_end: ff_x) q(m end) - q(m begin)

Tempo at m = _m_end - m_begin ⁼ m_end - m_begin

g(x) is the primitive function of f(x) (meaning that g'(x)=f(x)).

Composer markings such as global tempo indications (andante, allegro, etc.) and local tempo changes (such as rit., accel., etc.) are used as a default onto which the results of the procedures described here is superimposed.

1. The rule of Appoggiatura

For each appoggiatura a function a(x) is generated so that between the beginning of the last strong segment and the beginning of the appochord resolution a(x) descends linearly (or polynomially) in a user defined angle and afterwards it rises by the inverted angle until the original tempo has been reached.

2. The rule of acceleration

In order to apply this rule, the leading voice in each segment of the musical piece must be selected (automatically). The leading voice is selected for its rhythmic properties that pertain to this rule, and not necessarily for containing the dominant melody. This rule generates a function f(x) so that over every note in the leading voice the relative tempo is set to a ratio > 1 that depends on the musical-time length of the note, increasing as the length decreases (probably linearly). This results in speeding up on shorter notes. The whole function may be later multiplied by an amount set by the user, or determined automatically in order to balance the other rules so that the average tempo over the entire piece will be brought up to match the user defined "global tempo".

3. The rule of To-Beats To-Beats are defined in section 7, rule 3, only that in the context of this rule compound To-Beats are kept in separate parts, and the rule applies only to To-Beats that have notes that start within them. If a To-Beat is contained within another To-Beat (being in a lower hierarchy level), the rule is applied to the smaller beat first. This rule generates a function b(x) for each To-Beat. For each such beat, a target average tempo is assigned depending on its musical-time length (probably linearly), and then between the beginning of the beat to its end, b(x) is set to be a linear (or polynomial) descending function whose angle is determined by the target average tempo (meaning that the average value of all the functions generated by other rules that affect this region and b(x) and the user defined "global tempo" will amount to the target tempo).

Fig. 5 is graph illustrating the concept of the To-Beat for relative tempo 510 vs. musical time 520, in accordance with the principles of the present invention. After the end 540 of the To-Beat 530, the subsequent portion of the function rises back, linearly or polynomially, to the original relative tempo 550, taking a shorter time to do so and starting from a higher tempo than the end of the decline. I.e., (T₄ - T₃) > (T₅ - T₄) and (R₂ > R^. The user will determine the exact ratios.

4. The rule of Ending This rule is an extension to the rules of Appoggiatura and To-Beats. In an End-Of-Group or End- Of-Phrase, the last weak beat before the last note of the group or phrase is found, and the slowing down effected by rule #1 or #3 is increased (more for an End-Of-Phrase than for an End-Of-Group) by amounts determined by the user. Additionally, the return to original tempo is replaced by a constant relative tempo value equal to the average tempo in the weak beat until the beginning of the next group; there occurs the return to original relative tempo.

9. Other performance parameters

Other parameters (for example: Duration) are set according to simple default procedures selected or overridden by the user. These procedures are sensitive to composer markings (such as staccato, legato or tenuto, affecting the duration of notes).

As an example for such a procedure, by default, the sustaining pedal, when called for (in piano music), is applied slightly after a change in harmony.

The ability to create an entire performance on the computer saves the cost of hiring musicians to perform the music. Hiring professional musicians is costly. The cost of hiring a 5 piece band could run anywhere from $1000 to $10000 per session, and a full orchestra, $15000 to $75000, depending on the experience and reputation of the musicians. Musicians usually require a significant amount of time to learn and practice music in preparation for a performance. The Master Performer enables the music director to maintain better control of the production details and achieve optimal results without having to rely on the technical expertise and artistic abilities of others and enables producers to complete the production in a significantly shorter time frame.

The Master Performer frees the user to make intuitive expressive choices and carry them out instantly instead of the tedious trial and error process of working with existing tools when putting together a performance. Editors can make use of the "extra" time saved to better perfect their work and increase their output.

Processing music as MIDI files rather than heavy and complex WAVE files allows for an editing process that is simple and easy and has no impact on the audio quality of the end product and at the same time can be transferred and exchanged between the people involved in the production quickly and easily via e-mail or even mobile phones.

Working with the Master Performer is not only faster and less expensive it yields superior results. The substantially total control that the Master Performer provides over the elements that constitute musical expression means that the quality of music production can be superior to anything that was possible in the past.

The Master Performer replaces the primitive editing capabilities of existing sequencers with a powerful, "smart" and creative expression processors. The graphic interface and the templates are simple and easy to master. Editing procedures do not require a long learning curve. The interface provides maximal functionality with a minimal feature set, so that operation is simple. The musical expressive quality is controlled by indicating a desired change with one click of the cursor in the appropriate place. The program performs all the necessary calculations and automatically changes the numerical value of each parameter. Alternatively, when a template is chosen, an entire section can be modified with one click. Furthermore, the user can save patterns for future use and thereby make future work even simpler and more efficient. The editing process is thus fast and easy with simple curser and keyboard commands. The Master Performer (MP) specifications are as follows:

Scope:

In a preferred embodiment MP processes only note-oriented data, that is, data where pitch, time line placement and length of a tone are clearly specified. In an alternative embodiment, an enhancement to MP may be a component that translates audio files into pitch/length representation; and

MP analyzes the music based upon pre-defined or user-defined assumptions. For classical & western tonal music, MP assumes the basic rules of harmony, voice leading and notation marks as the basis for performance analysis. Optional enhancements to MP include musical analysis outside of western tonal music, to include music forms which are radically different, such as repetitive music, a-tonal, oriental scales etc.

Analysis features:

Identify the Implied Harmony. For example, the implied harmony for the Alberti pattern "do-mi-sol-mi" as a succession of 4 eighth notes will be the chord "do-mi-sol" applied to each one of the 4 different notes;

Identify Appoggiaturas.

Identify musical groups. A human player makes some kind of gesture, which indicates the inflection of the notes as part of groups and groups as part of sub-phrases, phrases, sections, etc. This is one of the main properties of human performance;

Identify cadences.

Identify performance related patterns. Find out which places in the music relate to each other in a way that a player would be inclined or moved to emphasize, or otherwise indicate, this relationship. For example, in a Canon or fugue, the theme alternates between different voices. This basic structural feature calls for special attention and treatment;

Analyze the music according to its acoustic qualities. Usually the lower notes are played softer than higher notes when played at the same time to avoid drowning the higher notes. Certain instruments might have to be intensified relative to others due to their acoustic qualities;

Determine the voice structure hierarchy of the piece. For example, determine which voice plays the melody, which is the harmony, what voices provide counterpoint lines or accompaniment;

Follow the composer's instructions in files of written music. For example, identify articulation, dynamic and expression marks. Follow the performer's original performance in files of recorded music, i.e. sequencer files; and

Analyze the title of the piece and the composer's specific style and historic period. These give further hints about the nature of the desired performance. A fugue will be performed differently if Bach, Beethoven or Bartok wrote it. A fantasy by William Byrd is radically different from a fantasy by Schumann. A mazurka requires special attention to its particular atmosphere and characteristics. A song obviously is performed differently if written by Schubert or Lennon & McCartney.

All performance parameters may be controlled by the user or taken from a predefined template:

For .classical performance in written-music oriented programs, change the tempo and/or offsets of each moment according to its role in the piece. For sequencer style programs, move the moments' offsets according to their role in the piece (refer hereinbelow to Method steps 8: Assigning tempo values);

According to the notes' role in the piece and the style of music, reduce or increase the velocity of each note (refer hereinbelow to Method steps 7: Assigning velocity values);

Adjust the offsets and tempi of notes in different voices, to provide a rhythmically unified performance;

Adjust the duration of notes according to predefined defaults, composer indications, personal preferences, etc;

Smooth the tempo/offset and velocity changes mathematically, to create a unified performance; Optionally, add a random factor to provide a unique performance every time; Add pedals (Controllers, 64 & 67) to piano performances according to harmonic background and changes, music style, score indications, performance practice and personal preferences;

Change other MIDI controllers, such as volume, vibrato, pitch bend, poly press, program change, modulation, breath control, balance, pan, sustain or any other specifically defined controller functions according to the nature of the piece, instrument and style. For example, a Sforzando-piano-forte effect in brass instruments requires manipulating the volume controller during the execution of a single note; Vibrato is a prominent attribute of wind and string instruments; volume is most pertinent to wind instruments;

Control MIDI Sysex (system exclusive) events and any hardware options available on various MIDI devices. This can enhance the' quality of performance for expressive instruments, like brass and strings; and

Modify the envelope attributes of each note, according to the instrument, the notation and the musical context, for example, the attack of a violin down-bow and up-bow, or tonguing in flute playing etc.

User Interface Features:

A performance editor allows the user to edit "moment database" format files. Each note's velocity, duration and tempo/offset is available for the user to modify;

The results of the analysis are displayed in the performance editor. This gives a powerful tool for the professional musician to modify the analysis results, change the performance parameters, tune and refine the results of the automatic MP tool;

A tool for building performance templates, to allow the user to control the way MP creates the performance framework. This, for example, allows a user to create a more

"melodramatic" performance by increasing the tempo slowdown at the end of musical sentences or by exaggerating the dynamic range. Also feasible is a more "dry" performance by making very thin smoothing curves, which makes only minimal rhythmic and velocity changes; Smart editing of the performance data is feasible. The user can create connections between different attributes of the performance, as would a human performer. For example, if the user exaggerated a crescendo, the MP editor will adjust the tempo, e.g. accel., rit., according to defaults, options or preferences; The editor can "learn" the typical changes for a composer, a user or a style, thereby creating automatic changes based on experience. This allows for a "closer to life" experience; and

A tool that shows the results of the analysis: For example: the implied harmony, the appoggiaturas, etc. This tool may serve as a powerful educational feature.

Infrastructure:

To provide the unique ability to analyze the music from a player's point of view, and perform it as a human would, the music needs to be put in a special format. The music is separated into moments. The moments are stored in a "moment database"; and A tool for transforming MIDI and other music file formats to and from the moment database is required.

Analysis algorithms database:

Determine the Implicit harmony Identify Appoggiaturas; Find End Of Groups (EOG); Identify End-Of-Phrase (EOP);

Identify performance related patterns; Acoustic analysis;

Voice structure; Notation analysis; and Contextual analysis.

Processing algorithms database:

Rules of tempo & offset changes; Rules of velocity changes; Tempo/offset adjust; Smoother; Randomizer; Pedal(s) processor; MIDI controllers; Sysex events; and Envelope processor.

The Master Performer provides many benefits to users as follows.

Soundtracks for films, television and commercials can be produced from beginning to end by one or two people. Editing and synchronizing can be performed to perfection without harming the expressive qualities of the performance. In fact, the Master Performer makes it possible to have total control over the expressive message of the music in connection with any other media without any compromises that commonly result from synchronization, overall timing restrictions and picture editing.

Musicians working with sequencers like Cubase, or with loop creators like Fruity Loops can add the missing human touch to an otherwise mechanical rendering of a composition by running it through processor 333 and choosing an expression template 334 from the Master Performer's template database.

Record producers can produce an almost unlimited variety of different performances without having to go through the process of actually recording each one of them, which involves paying for a recording venue, renting or purchasing recording equipment, hiring sound, editing and mastering engineers, rehearsal time and recording session time. With the Master Performer one person in a small studio can carry out the entire process.

Composers working with notation software like Finale™ or Sibelius™ can prepare a MIDI file of a human-like performance using the Master Performer, then run it through a sound module, process the audio and have it all ready for presentation to producers within hours instead of weeks or even months.

Simple and Easy access to active interaction with music using the Master Performer technology could lead to greater involvement with music for a large number of people who are now intimidated and hindered by the technical barrier of mastering a musical instrument on a level that enables true self-expression in music. Dance performances can have the music performed by the Master Performer, controlled by a human "conductor" coordinating the expressive attributes of the music with the movements of the dancers on the stage in real time.

Music educators can utilize the technology to increase the motivation and encourage the engagement of students with music by offering a powerful tool that affords instant feedback and promotes exploration into realms of possibilities that only the most accomplished professionals can reach with conventional methods.

The Master Performer technology is primarily designed for people who work in the digital music production industry, but is not limited to such use. The technology is most effective in any production that need not stay exclusively within the audio domain. In fact, the technology offers the option of dispensing entirely with the dependence on audio source material. As such it can be integrated smoothly into any segment of the audio production market. When the editing is completed on a MIDI file, only then does the final processing turn it into an audio file.

Claims

We claim:

1. A master performer (MP) system for allowing a user to automatically create performances of musical compositions with human expression qualities based on performance-pertinent analysis of musical compositions, integrating various algorithms for both the analysis and its rendering into a performance, the system comprising: a processor for analyzing musical data; a processor for setting performance attributes and creating performances for a specific composition; and a user interface for controlling the analysis and its transformation into performances, and for creating MP templates which provide high-level control over the transformation.

2. The MP system of claim 1 , wherein the music analysis yields groups that function like words in a musical syntax.

3. The MP system of claim 1, wherein the relations between these groups are further analyzed to generate larger groups in higher hierarchy levels.

4. The MP system of claim 1 , further comprising an interface for controlling the analysis and its transformation into a performance.

5. The MP system of claim 1 , further comprising an editor for editing the musical performance, wherein the system allows a user to edit the performances generated externally.

6. The MP system of claim 1 , further comprising a performance template database, to provide the user with high-level control over the way the MP automatically creates the initial performance, and a template builder allowing the user to create his own templates by manipulating the transformation parameters in low-level.

7. The MP system of claim 1 , further comprising a performance editor, to provide the user with control over the low-level performance parameters, such as velocity, tempo, duration, etc. of an existing performance of a specific musical composition. High-level control and low-level control are both possible.

8. The MP system of claim 7, wherein the control is a high-level control.

9. The MP system of claim 7, wherein the control is a low-level control.

10. The MP system of claim 1, further comprising a musical moments database, wherein music data is stored as a list of moments, on which analysis results and performance data are superimposed.

11. The MP system of claim 1 and claim 10, further comprising translators specific to MIDI and other music data formats for loading music into said musical moments database and saving its contents.

12. The MP system of claim 1, further comprising an Implied Harmony detector.

13. The MP system of claim 1, further comprising an Appoggiatura detector.

14. The MP system of claim 1 , further comprising a group detector

15. The MP system of claim 1, further comprising a performance related pattern detector.

16. The MP system of claim 1, further comprising a cadence detector.

17. The MP system of claim 1, further comprising a phrase detector.

18. The MP system of claim 1, further comprising a processor that generates a performance based on the analysis of a specific musical composition and user preferences.

The MP system of claim 1 , for use by media production studios.

20 The MP system of claim 1 , for use by sound engineers.

21 The MP system of claim 1 , for use by record producers.

22 The MP system of claim 1 , for use by composers.

23 The MP system of claim 1 , for use by amateur musicians.

24. The MP system of claim 1 , for use by dance companies.

25. The MP system of claim 1 , for use by music teachers.

26. The MP system of claim 1 , for use by music students.