WO2013087638A1

WO2013087638A1 - Method for digitally processing a set of audio tracks before mixing

Info

Publication number: WO2013087638A1
Application number: PCT/EP2012/075106
Authority: WO
Inventors: Laurent Girin; Nicolas STURMEL; Laurent DAUDET
Original assignee: Institut Polytechnique De Grenoble; Centre National De La Recherche Scientifique (Cnrs)
Priority date: 2011-12-14
Filing date: 2012-12-11
Publication date: 2013-06-20
Also published as: FR2984579B1; FR2984579A1

Abstract

The invention relates to a method for digitally processing a set of tracks (t_{i, j}(n)) forming a mixed input sound signal (m_j (n)). The method includes, for at least one sound reproduction channel (j) of said mixed input signal (m_j(n)): a step (i) of determining a compression gain (g _j(n)) of a dynamic compression processing (n_j(n)) applied to said mixed input signal (m_j(n)); and a step (ii) of multiplying at least one of the tracks from the set of tracks (t_{i, j}(n)) by the determined compression gain (g_j(n)) so as to obtain a track that is dynamically compressed according to said dynamic compression processing.

Description

Digital processing method on a set of audio tracks before mixing

The technical field of the invention is signal processing and more particularly audio signal processing.

It is known to those skilled in the art, the use of audio-digital stations also called DAW according to an acronym for Digital Audio Workstation.

These audio-digital stations make it possible to carry out treatments on a plurality of tracks. These tracks are obtained from different sound synthesis devices or sound capture devices such as microphones. A track may correspond to a source identified by the listener as being a musical instrument, for example a guitar. But this is not necessarily the case. Indeed, the sound produced by the guitar can also be found in other tracks than the one to which the guitar is associated. On the contrary, a sound identified as the sound of a guitar can be associated with a series of tracks.

Treatments made by digital audio stations include linear digital processing such as instantaneous and stationary linear processing or convolution processing. An example of instantaneous and stationary linear processing is amplification and an example of convolution processing is reverberation. The processing performed by the audio-digital stations also include non-linear digital processing such as, for example, dynamic compression. Dynamic compression is a treatment that allows in particular to limit the amplitude and variations of the sound level of the tracks.

The digital audio stations also allow the plurality of digital audio tracks to be mixed to obtain a mixed signal.

One of the technical problems relating to the mixed signals is to be able to separate from the mixed signal one of the sources forming the mixed signal.

For this, it is known, from the state of the art, the use of digital information to facilitate the separation of a source of signal mixed according to a method called IS S according to an acronym known to those skilled in the art Informed Source Separation. This digital information is previously generated using the sources that are going to be mixed and the signal mixed.

However, some of the processing of digital audio stations can disrupt this separation, which becomes complex. There is therefore a need to allow a simpler separation of the sources and / or tracks of a mixed signal.

It is proposed, according to an embodiment of the invention, a method for obtaining a mixed signal for which the separation of a track and / or a source is simpler.

It is also proposed, according to an embodiment of the invention, a dynamic compression identical to that of the mixed signal, performed on at least one of the tracks forming the mixed signal. This dynamic compression is, for example, performed on each of the tracks forming the mixed signal taken separately.

In addition, an object of the invention is also to allow an estimation of the compression gain of a mixed signal stored.

The object of the invention is therefore a digital processing method on a set of tracks forming a mixed input sound signal.

According to a general characteristic, the method comprises, for at least one sound reproduction channel of said mixed input signal:

a step of determining a compression gain of a dynamic compression process applied to said input mixed signal; and

a step of multiplying at least one of the tracks of the set of tracks by the determined compression gain, so as to obtain a dynamically compressed track according to said dynamic compression process.

Thus, a dynamic compression identical to that performed on the mixed input signal can be performed on a single track. This dynamic compression can therefore be performed before the mixing. According to one embodiment, the method further comprises adding the at least one dynamically compressed track to obtain a dynamically compressed output mixed signal.

This is done mixing after dynamic compression of at least one of the tracks. In the case where all the tracks of the mixed input signal are multiplied by the determined compression gain and are added then the output mixed signal obtained is identical to the result of said compression processing performed on the input mixed signal.

According to a characteristic of this embodiment, the mixed input signal has undergone convolution processing and said convolution processing is performed on each of the tracks of the mixed input signal.

The convo lution can thus be distributed over all the tracks of the mixed input signal to be performed before mixing and dynamic compression.

According to an additional embodiment, digital information is available on sources corresponding to the tracks of the mixed output signal and the method further comprises separating said mixed output signal from at least one of the sources using digital information.

In general, in the state of the art, dynamic compression is performed on the entire mixed signal. On the contrary, when the mixing is the last step of the production of the mixed signal, the separation of the track is simpler since it is enough to reverse the addition. This separation did not alter the other mixing tracks nor the dynamics compression processing.

According to one feature, the step of determining the compression gain comprises a dynamic detection step performed on the input mixed signal, said dynamic detection step using the setting parameters and the calculation algorithm of said dynamic compression processing .

Thus, the dynamic detection step applied to the mixed input signal makes it possible to determine the compression gain of dynamics of dynamic compression processing applied to the mixed input signal in a simple and accurate manner.

According to another characteristic, the determined compression gain is stored before the multiplication step.

We can then use a simpler audio-digital station.

In another embodiment, a stored mixed signal calculated from said input mixed signal dynamic compression processing is available and the compression gain determining step includes a step of estimating the compression gain. using said stored mixed signal and the input mixed signal.

It is then proposed to be able to estimate the compression gain of a mixed signal stored. This is particularly advantageous if it is desired to perform a re-mastering of an old record for which the algorithm and / or the parameters of the dynamic compression are lost.

According to one characteristic of this embodiment, said estimation step comprises a step of dividing a function obtained from said mixed signal stored by a function obtained from the mixed input signal to obtain a ratio of the energies of the stored mixed signal and the mixed input signal.

It is thus proposed a mo de of simple calculation to estimate the gain of compression.

Other objects, features and advantages will become apparent upon reading the following description given solely as a non-limitative example and with reference to the accompanying drawings, in which:

- Figures 1 and 2 illustrate audio signal processing method according to the state of the art;

FIGS. 3 and 4 illustrate embodiments of methods for processing a mixed signal according to the invention;

FIGS. 5 and 6 illustrate in greater detail steps of the embodiments of the treatment method of the invention illustrated in FIGS. 3 and 4. FIG. 1 shows schematically the processing carried out by a digital audio station according to the state of the art.

From the sources represented on line A, the station will perform digital processing to obtain sources with effects represented on line B. These digital processes include linear processing such as convolution processing (called convolutional processing in the legend of FIG. 1) or instantaneous and stationary linear processing (called LIS processing in the legend of FIG. 1). Digital processing also includes non-linear processing (referred to as non-linear processing in the legend of Figure 1). A known example of instantaneous and stationary linear processing is amplification. A known example of a convolution processing is the reverberation processing.

Then, from these sources with effects, the station will perform panoramic treatments. This corresponds in particular to the distribution of sources on all sound reproduction channels with particular treatments well known to those skilled in the art. For example, in the illustrated case of a stereophonic recording, it is necessary to broadcast each of the sources with effects on two channels, the right channel and the left channel.

All the results of the panoramic treatments are then added channel by channel to obtain a mixed signal represented on the line D.

The mixed signal is then processed. It undergoes, for example, linear treatments such as convolution treatments (called convolutional processing in the legend of FIG. 1) or instantaneous and stationary linear treatments (called LIS treatments in the legend of FIG. 1). It can also undergo dynamics compression processing which is a nonlinear processing. In general, dynamic compression treatment is the latest treatment performed. This is also generally the only nonlinear processing performed on the mixed signal. We then obtain the processed mixed signal represented on line E. This processed mixed signal is ready to be recorded or broadcast.

We can express all the treatments made from, for example, the tracks of line C of FIG. 1 in the form of mathematical equations, using the notation below:

the set of tracks at line C of FIG. 1 are denoted t 1 (ir 1), with i representing the number of the track and j representing the number of the rendering channel;

a set of tracks t ^ iri) form a source s _kj on the restitution channel j, according to the formula

where I _k is a set of natural numbers represents a group of track numbers;

a convolutional processing of formula c- (") on the rendering channel j is applied to the mixed signal of the line D;

a nonlinear processing of formula "" (") on the rendering channel j is applied to the result of the convolution processing.

With these notations, it is possible to express the processed mixed signal and _j (n) of the line E of FIG. 1 in the form:

m _j (n) = n _j (equation 1),

in which is the formula of a dynamic compression process.

FIG. 2 illustrates in more detail the realization of the dynamics compression treatment of formula ", (").

An input mixed signal is received which has undergone a convolutional processing C _j (n) from line D and possibly a stationary instantaneous linear processing which, if necessary, can be distributed on the tracks. This mixed signal input has formula

However, a dynamic detection step is performed on the mixed input signal. In this step, a computational algorithm and tuning parameters of the audio-digital station are used to determine a compression gain g _j (n) corresponding to the dynamics compression treatment of the formula applied to the mixed input signal. These parameters and the algorithm define the function "■ ("). The operations performed by the calculation algorithm include a calculation of the local energy of the mixed input signal.

Finally, the compression gain g _j (n) is applied to the input mixed signal of formula C _j (n) * t _ij (n) to obtain the processed mixed signal i

m _j (n). The processed mixed signal m _j (n) is therefore dynamically compressed.

FIG. 3 illustrates a mode of implementation of the method according to the invention. It makes it possible to obtain, after step 3, a result similar to that of the line E of FIG. 1 with equation m _j (n) = n _j c _j (n) * ^' t _; ■ (") but in a different way.

FIG. 3 describes a processing carried out on tracks (1) represented at line C of FIG. 1. This method may for example be implemented at least in part by an audio-digital station. The method is described for a rendering channel j.

That said, the process is identical for each of the rendering channels considered.

The treatment of this process comprises 5 steps (step 0, step 1, step 2, step 3, step 4):

step 0 is a step of applying a convolutional processing on the tracks t ^ iri). Indeed, the convolution processing is a linear processing, it can therefore be applied to each of the tracks ti irî) and then perform the mixing instead of applying it only once to the mixed signal of the line D of FIG. of formula ti (n). We obtain the following equation whose result is similar to that of equation 1:

m _j {n) = n _j m _j {n)) (equation 2), wherein m _j (n) = ^' c _j (n) * t _i (n) is a mixed input signal which corresponds to the signal mixed input shown in Figure 1.

step 1 is a step of determining a compression gain of the dynamic compression processing of equation 2. As explained below, the determination of a compression gain of the dynamic compression processing of formula ri _j (n) applied to the mixed input signal w _i (n) will make this dynamic compression process linear.

Furthermore, the compression gain of the formula dynamic compression treatment applied to the input mixed signal w _i (n) is identical to the compression gain g _j (n) (the compression gain of the formula treatment applied to the signal C _j (n) * t _ij (n)) because the mixed input signal m ^ n) has a value identical to the input mixed signal C _j (n) * t _ij (n).

Formula processing "(") is a dynamic compression process. It is a nonlinear process and therefore can not be distributed as the convolutional processing C _j (n) on all the tracks t ^ iri). That being so, the formula dynamics compression treatment applied to the mixed input signal rh _j (n) comprises a multiplication by the compression gain g _j (n).

Thus, if the value of the compression gain is determined, it suffices to multiply the mixed input signal rh _j (n) by this compression gain g ₇ (") to obtain the dynamic compression treatment according to the formula

™ _j (n) = g _j (n) - (equation 3),

in which the multiplication by the gain g ₇ (") is a linear operation which can therefore be distributed on each track t (ir) forming the mixed signal without changing the final result.

In the embodiment of the method illustrated in FIG. 3, a dynamic detection step 11 performed on the mixed input signal rh _j (n) will be used to determine the compression gain.

This dynamic detection step 11 uses a calculation algorithm and setting parameters of the digital audio station identical to those already used for the step illustrated in Figure 2.

It is thus possible to reproduce the function and exactly determine the compression gain g _j (n) of the dynamic compression process applied to the input mixed signal w _i (n).

Step 2 comprises a step of multiplying each of the tracks after convolution by the dynamic compression gain g _j (n). The treatment performed corresponds to the equation:

hi («) = g _j («) · (C _j (n) * t _tj (n)) (equation 4)

Advantageously, the gain determined in step 1 is stored before the multiplication operation of step 2. Indeed, in the case where the determined gain is not stored then, the digital audio station must comprise two buses. . Namely, a bus with the mixed signal connected to a dynamic detection block and a bus with the at least one track to multiply. On the contrary, in the case where the gain is stored alone a bus is sufficient since the stored gain is directly multiplied with the bus having the at least one track to be multiplied. We can then use a simpler audio-digital station.

Steps 11 and 2 are illustrated in greater detail in FIG. 5. In this figure, the mixed input signal of formula m _j (ri) = ^' c _j (ri) * t _ij (ri) is received on which we apply the detection i

dynamic 11 to determine the compression gain g _j {n). Then we multiply at least one track by the compression gain g _j (n) determined according to the formula: t _i . {n) = g. {n) · (c ■ {n) * t _i . {ri)).

The method of Figure 3 further comprises:

a step 3 of summing at least a portion of the tracks dynamically compressed according to the formula:

m _j {ri) = ^' t _ij (ri) (equation 5), i

where t _] {n) = g _j (n) · (c (n) * t _] (ri)).

We can therefore introduce a source ¾ ■ image of the source s _kj :

S _k ( ⁿ )

(equation 6), where I _k is a set of natural numbers representing a group of track numbers.

a separation step 4, during which digital information "info" is available on sources corresponding to the tracks of the mixed output signal according to equation 6 and said mixed signal is separated from at least one of the sources using digital information "info". Since the summation (3) is carried out in the last step, the separation will not be disturbed by the other treatments, only a subtraction is sufficient.

FIG. 6 illustrates in greater detail the principle of separation step 4 in the simplified case of a single rendering channel. That being so, the person skilled in the art can easily extend this case to any number of rendering channels by reproducing this principle for each of the rendition channels.

The set of sources S i, S ₂ , S ₃ , S ₄ is mixed according to a step called Mix from which a mixed signal is obtained. These sources may comprise a plurality of tracks according to the formula:

where I _k is a set of natural numbers representing a group of track numbers.

A step of analysis and extraction AE is then performed on each of the sources and on the mixed signal. From this step AE, digital information "info" is obtained.

Then, according to the step IS corresponding to step 4, the information "info" and the signal mixed with the different sources are received.

With the information "info", it is then possible to separate at least one source of the mixed signal by using, for example, the subtraction.

The separate sources can then be remixed according to the Re-Mix step.

FIG. 4 illustrates another mode of implementation of the method according to the invention. It is distinguished from that of FIG. 3 by step 1 which comprises a step of estimating compression gain 12 instead of step 1 1 of dynamic detection. According to step 12, the algorithm and adjustment parameters of dynamic detection of the dynamic compression treatment of formula ", (") are no longer available. An estimate of the compression gain g _j {n) is therefore provided.

For this we consider the following calculation which makes it possible to obtain an estimator of the compression gain g _j (n) from the ratio of the energies of the stored mixed signal and the mixed input signal, for example according to the equation below. :

~

S _j n) = (equation 7),

in which the function Η is a Hilbert transform and in which m = n, {Σc (()) * t _i , (())! and m = Vc - (n) * t _; . ( ').

Claims

1. Digital processing method on a set of tracks (t _; ■ (")) forming an input mixed sound signal (rh _j (n)), characterized in that it comprises, for at least one sound reproduction channel ( j) of said mixed input signal (rh _j (n)):

a step of determining (1) a compression gain (g _j (n)) of a dynamic compression process (". (")) applied to said input mixed signal (rh _j (n)) ); and

a step of multiplying (2) at least one of the tracks of the set of tracks (^ _; ()) by the determined compression gain (g ₇ (")), so as to obtain a dynamically compressed track (t _i (n)) according to said dynamic compression process.

2. Method according to claim 1, further comprising an addition (3) of the at least one dynamically compressed track

(t _i (n)) to obtain a mixed output signal dynamically compressed.

The method of claim 2, wherein the input mixed signal has undergone convolution processing (c (w)) and (0) said convolutional processing is performed on each of the tracks (t _i (n)). of the mixed input signal.

4. Method according to one of claims 2 or 3, wherein there is digital information (info) on sources corresponding to the tracks of the mixed output signal and the method further comprises a separation (4) of said mixed signal of output of at least one of the sources using the digital information (info).

5. Method according to one of claims 1 to 4, wherein the step of determining the compression gain (1) comprises a dynamic detection step (11) performed on the input mixed signal, said dynamic detection step using the setting parameters and the calculation algorithm of said dynamic compression process ("(")).

6. Method according to one of claims 1 to 5, wherein the determined compression gain (g _j (n)) is stored before the multiplication step.

The method according to one of claims 1 to 4, wherein a stored mixed signal (m (n)) calculated from said dynamic compression processing (". (")) Of the input mixed signal is available and the step of determining (1) the compression gain comprises a step of estimating (12) the compression gain using said stored mixed signal (m (n)) and the mixed input signal (m (n )).

The method of claim 7, wherein said estimating step (12) comprises a step of dividing a function obtained from said mixed signal stored by a function obtained from the input mixed signal to obtain a report energies of the stored mixed signal and the input mixed signal.