CN100511425C - Method and system for improving ADPCM voice coding step length regulation - Google Patents
Method and system for improving ADPCM voice coding step length regulation Download PDFInfo
- Publication number
- CN100511425C CN100511425C CNB2003101165291A CN200310116529A CN100511425C CN 100511425 C CN100511425 C CN 100511425C CN B2003101165291 A CNB2003101165291 A CN B2003101165291A CN 200310116529 A CN200310116529 A CN 200310116529A CN 100511425 C CN100511425 C CN 100511425C
- Authority
- CN
- China
- Prior art keywords
- length
- frame
- signal
- change function
- higher limit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Abstract
The method and system for improving ADPCM voice coding step length regulation includes partitioning one voice signal to form several voice frames and pre-coding the voice frames to obtain the optimal voice frame step length varying functions and upper step length limits; and then performing formal coding with the optimal voice frame step length varying functions and upper step length limits of the voice frames to improve voice quality and avoid great error.
Description
Technical field
The present invention relates to a kind of adaptive differential pulse code modulated (Adaptive DifferentialPulse-Code Modulation; ADPCM) Methods for Coding, particularly a kind of is the method and system of improving the step-length adjusting of ADPCM voice coding.
Technical background
Fig. 1 is the simplified system calcspar of known adpcm encoder 10, and it comprises a quantizer 12, prediction device 14 and step size automatic regulator 16.12 pairs one difference signal Δs of quantizer X quantizes and produces a numerical code C[n] with quantize after difference signal Δ X ' [n], this difference signal Δ X is voice signal X[n] with estimate the difference of signal X ' [n], difference signal Δ X ' [n] after this quantification produces a signal S and inputs to prediction device 14 to produce the new signal X ' [n] that estimates with estimating signal X ' [n] after totalizer combines, step size automatic regulator 16 is according to this numerical code C[n] and output one step change function M (C[n]) to quantizer 12.
Fig. 2 is the system block diagrams of ADPCM decoder 20, and it comprises a de-quantizer 22, a prediction device 24 and step size automatic regulator 26.Step size automatic regulator 26 receives this numerical code C[n], and export step change function M (C[n]) according to this, de-quantizer 22 is according to step change function M (C[n]) this numerical code of de-quantization C[n] produce difference signal Δ X, this difference signal Δ X with estimate signal X ' [n] and after totalizer combines, revert to voice signal X[n], prediction device 24 is according to this voice signal X[n] produce this and estimate signal X ' [n].
The quantizer 12 of known adpcm encoder 10 has regulatory function, it adjusts according to the step-length step_size (n) that step change function M (C[n]) will import quantizer 12, to adapt to the variation of present input difference signal AX, yet, in the step-length renewal process of quantizer 12, be to decide next step value, normally present step-length step_size (n) be multiplied by a step change function M (C[n]) according to present coded data, shown in the following formula (1):
Step_size (n+1)=step_size (n) * M (C[n]) ... formula (1)
Wherein, step_size (n+1) is for giving the used step value of next sampling spot.
Step change function M (C[n]) only with present numerical code C[n] relevant, in general, one step change function M (C[n]) and numerical code C[n are all arranged in step size automatic regulator 16 and 26] the table of comparisons, shown in following table one, and be a default value, it is not regulated at different characteristics of signals, therefore, when the amplitude variations of voice signal is excessive, and pairing step change function M (C[n]) can't do best processing to voice signal, thereby cause distortion.
Table one:
| Numerical code C[n] | Step change function M (C[n]) |
| 0，1，2，3，8，9，10，11 | 0.9 |
| 4，12 | 1.2 |
| 5，13 | 1.6 |
| 6，14 | 2.0 |
| 7，15 | 2.4 |
With reference to table, numerical code C[n wherein] be one 4 bit data, as numerical code C[n] be 0,1,2,3,8,9,10 or 11 o'clock, step change function M (C[n]) be 0.9, as numerical code C[n] when being 4 or 12, step change function M (C[n]) be 1.2, as numerical code C[n] when being 5 or 13, step change function M (C[n]) be 1.6, as numerical code C[n] when being 6 or 14, step change function M (C[n]) be 2.0, as numerical code C[n] when being 7 or 15, step change function M (C[n]) be 2.4, as shown in Table 1, different numerical code C[n] and corresponding separately one fixing step change function M (C[n]), have nothing to do with signal itself.
In addition, known adpcm encoder 10 is all set a higher limit to step size, in case halt long excessive and cause distortion, and this higher limit is unique, yet voice signal is all inequality in the amplitude and the dynamic change scope of each time point, needs bigger step-length when scope is big, and scope hour only needs smaller step size, and single fixing step-length higher limit can not satisfy all scopes.
Therefore, a kind of can finding out with the difference of scope can reach signal to noise ratio (S/N ratio) (Signal-to-NoiseRatio; SNR) the step-length higher limit and the ADPCM coding method and the system of step change function are to be the institute Ji.
Summary of the invention
Purpose of the present invention is to propose the method and system that a kind of ADPCM of improvement voice coding step-length is regulated.
Purpose of the present invention, other is to provide a kind of ADPCM voice coding step-length control method and system that finds out the most suitable step change function and step-length higher limit with precoding.
Purpose of the present invention is to provide a kind of again and improves voice coding tonequality and prevent excessive ADPCM coding method and the system that causes error of step-length.
According to the present invention, the method that a kind of ADPCM of improvement voice coding step-length is regulated comprises that cutting apart a voice signal forms a plurality of frames, and each frame of precoding to be obtaining having the step change function and the step-length higher limit of maximum signal to noise ratio, and formally encodes with step change function and step-length higher limit that each frame has a maximum signal to noise ratio.
Because voice signal has the characteristic of slow variation, so the utmost point in the short time signal variation characteristic all very close, therefore the present invention is cut into a plurality of frames that continue with voice signal, and be that unit encodes and regulates with the frame, moreover, the present invention utilizes the mode of precoding according to the maximum upper limit that calculates optimal step change function of each frame and step-length, cause tangible error to improve tonequality and to prevent that step-length is excessive, after the maximum upper limit of optimal step change function that obtains each frame and step-length, again each frame is formally encoded, to obtain best tonequality and to prevent that error from occurring.
After precoding finishes, the optimal step change function of each frame and the maximum upper limit of step-length will be stored in the comparison list, according to this table of comparisons, the step change function and the step size higher limit of ADPCM coded system will be adjusted with frame, therefore, coding of the present invention and decoding system can be made best adjusting at different sound properties, to prevent distortion and to improve tonequality.
Description of drawings
For the personage who has the knack of this skill, cooperate the accompanying drawing follow from following being described in detail, the present invention can more clearly be understood, and its above-mentioned and other purpose and advantage will become more obvious, wherein:
Fig. 1 is known adpcm encoder;
Fig. 2 is known ADPCM decoder;
Fig. 3 is the waveform synoptic diagram of voice signal;
Fig. 4 is the process flow diagram that the present invention improves the ADPCM voice coding;
Fig. 5 is the system block diagrams of adpcm encoder of the present invention; And
Fig. 6 is the system block diagrams of ADPCM decoder of the present invention.
The reference numeral explanation
10 adpcm encoders
12 quantizers
14 prediction devices
20 ADPCM decoders
22 de-quantizer
24 prediction devices
100 voice signals
200 read the into speech data of a frame
202 precodings
20202 make I=1
20204 make J=1
20206 is the step-length upper limit with MaxStepsize (J), is that the step change function is done precoding to entire frame with M (I)
20208 calculate the signal to noise ratio (S/N ratio) after the precoding and write down I and the J value
20210 whether J is more than or equal to k
20212 make J=J+1
20214 whether I is more than or equal to n
20216 make I=I+1
20218 finish precoding, and have found out pairing I of maximum S R value and J value
204 is this step change function with M (I), formally encodes with MaxStepsize (J) step-length higher limit
Whether 206 be last frame
300 adpcm encoders
302 dispensers
304 quantizers
306 prediction devices
308 step size automatic regulators
310 SNR counters
400 adpcm encoders
402 de-quantizer
404 step size automatic regulators
406 prediction devices
Embodiment
Fig. 3 is the waveform synoptic diagram of a voice signal 100, because voice signal 100 has the characteristic of slow variation, therefore the utmost point in the short time signal 100 variation characteristics very close, the present invention utilizes this characteristic that voice signal 100 is divided into a plurality of frames, characteristics of signals in each frame is very close, thereby the signal in same frame can use identical step change function to encode.In this embodiment, the length of each frame is L, then is that unit carries out precoding and formal coding again with the frame, and its flow process as shown in Figure 4.Present embodiment is set the higher limit of k kind step-length, ascending be respectively MaxStepsize (1), MaxStepsize (2) ... to MaxStepsize (k), and n kind step change function, be in regular turn M (1), M (2) ... to M (n), choose optimal step-length higher limit and step change function for each frame.After beginning coding, at first step 200 is read the into speech data of a frame, then in step 202, the frame of being read is into carried out precoding to obtain the optimal step change function of this frame M (I) and step-length upper limit value M axStepsize (J), and then step 204 is formally encoded with this step change function M (I) and step-length upper limit value M axStepsize (J), after finishing formal coding, step 206 judges whether this frame is last frame, then finish coding if yes, otherwise then get back to step 200, continue again next frame is carried out precoding and formal coding.
Refer again to Fig. 4, in precoding step 202, in order to determine optimum stepsize higher limit and the step change function of each frame in k kind step-length upper limit value M axStepsize (J) and n step change function M (I), in step 20202 and 20204, make I=1 and J=1 respectively, then in step 20206, with MaxStepsize (J) is the step-length higher limit, with M (I) is that the step change function is done precoding to entire frame, follow the signal to noise ratio (S/N ratio) after step 20208 is calculated precoding, and record I and J value, come step 20210 to judge that whether J is more than or equal to k again, if otherwise carry out step 20212, make J=J+1, and repeating step 20206 to 20210, otherwise then carry out step 20214, step 20214 is to judge that whether I is more than or equal to n, if otherwise carry out step 20216, make I=I+1, and repeating step 20204 to 20214 again, otherwise, carry out step 20218 and finish precoding, and found out pairing I of maximum S R value and J value, M (I) and MaxStepsize (J) with maximum S R value, the optimum stepsize that is the frame that reads changes function and step-length higher limit.Method not only can be regulated the step change function according to difference by this, more can regulate step change function and step-length higher limit according to different frames, thereby obtains the ADPCM coding of the most suitable voice signal characteristic.
Fig. 5 is the system block diagrams of adpcm encoder 300 of the present invention, and it comprises a dispenser 302, quantizer 304, prediction device 306, step size automatic regulator 308 and SNR counter 310.Dispenser 302 is cut apart voice signal X[n] form a plurality of frames, can utilize counter to write down the length of frame, 304 couples of difference DELTA X of quantizer quantize and produce a numerical code C[n] with quantize after difference signal Δ X ' [n], this difference signal Δ X is voice signal x[n] with estimate the difference of signal X ' [n], difference signal Δ X ' [n] after the quantification with estimate signal X ' [n] and after totalizer combines, produce a signal S and input to prediction device 306 to produce the new signal X ' [n] that estimates, step size automatic regulator 308 is according to numerical code C[n] and output one step change function M (I, C[n]) to quantizer 304.To voice signal X[n] in each frame when doing precoding, step size automatic regulator 308 provides various step change function and step-length higher limit, and utilize SNR counter 310 to calculate the SNR value of corresponding various step change functions and step-length higher limit, and then obtain the optimal step change function of each frame M (I) and step-length upper limit value M axStepsize (J), therefore, the step change function M (I that final step size automatic regulator 308 is determined, C[n]) and numerical code C[n] the table of comparisons also be the function of frame, because each frame all has its best step change function M (I, C[n]) and step-length upper limit value M axStepsize (J), tonequality improved so when coding, can reduce distortion.The step-length renewal process of system 300 is to decide next step value according to present coded data and frame, shown in the following formula (2):
Step_size (n+1)=step_size (n) * M (I, C[n]) ... formula (2)
Wherein, step_size (n) is present step value, and step_size (n+1) is next step value.
The system 300 of Fig. 5 can utilize the process control of software to realize on existing hardware structure.And, thereby the length L of frame and step change function M (I, C[n]) can revise with the step-length higher limit, to cooperate suitable voice signal X[n].
Fig. 6 is the system block diagrams of ADPCM decoder 400 of the present invention, and it comprises a de-quantizer 402, a prediction device 404 and step size automatic regulator 406.Step size automatic regulator 406 receives this numerical code C[n], and export step change function M (I according to this, C[n]), it is the function of speech data and frame, de-quantizer 402 is according to step change function M (I, C[n]) this numerical code of de-quantization C[n] and produce difference signal Δ X, this difference signal Δ X with estimate signal X ' [n] and after totalizer combines, revert to voice signal X[n], prediction device 404 is according to this voice signal X[n] produce this and estimate signal X ' [n].Similarly, step size automatic regulator 406 employed step change function M (I, C[n]) and numerical code C[n] the table of comparisons be along with the voice signal X[n that is imported] and the difference of frame and changing.
In different embodiment, the length L of frame can adopt revocable size, for example, cuts apart according to the scope and the variation of voice signal 100.
More than the narration done for preferred embodiment of the present invention be the purpose of illustrating, accurately be disclosed form and be not intended to limit the present invention, based on above instruction or to make an amendment or change from embodiments of the invention study be possible, embodiment has the knack of this operator and utilizes the present invention to select in practical application with various embodiment and narrate for explaining orally principle of the present invention and allowing, and technological thought attempt of the present invention is decided by following claim and equalization thereof.
Claims (9)
1. one kind is improved the method that ADPCM voice coding step-length is regulated, and comprises the following steps:
Cut apart a voice signal and form a plurality of frames;
Each this frame of precoding in regular turn is to obtain the step change function with maximum signal to noise ratio and a step-length higher limit of each this frame correspondence; Formal this speech signal of coding and
Wherein, the step of this each this frame of precoding comprises the following steps:
Set a plurality of step-length higher limits;
Set a plurality of step change functions;
Calculate the signal to noise ratio (S/N ratio) of each this frame when different step change functions and step-length higher limit; And
Find out this step change function with maximum signal to noise ratio and this step-length higher limit of each this frame correspondence.
2. the method for claim 1, wherein this step of cutting apart a voice signal comprises with a regular length and cuts apart this voice signal.
3. the method for claim 1, wherein this step of cutting apart a voice signal comprises with an on-fixed length and cuts apart this voice signal.
4. ADPCM coded system comprises:
One dispenser is cut apart a voice signal to form a plurality of frames with a length;
One quantizer is estimated the difference of signal and is produced a numerical code to quantize this voice signal and; And
One step size automatic regulator, providing a step change function and step-length higher limit to this quantizer,
Wherein,
When each frame in the voice signal is done precoding, the step size automatic regulator provides various step change function and step-length higher limit, and utilize the SNR counter to calculate the SNR value of corresponding various step change function and step-length higher limit, and then obtain optimal step change function of each frame and step-length higher limit.
5. coded system as claimed in claim 4, wherein, this length is a fixed value.
6. coded system as claimed in claim 4, wherein, this length is an on-fixed value.
7. coded system as claimed in claim 4 more comprises a SNR counter, calculating the SNR value that a plurality of default step change functions and step-length higher limit are caused, and then finds out for each this frame and to have maximum S R value step change function and step-length higher limit.
8. coded system as claimed in claim 4, wherein, each this frame has the step change function and the step-length higher limit an of the best.
9. an ADPCM decoding system, it receives a numerical code, and to produce a voice signal, this decoding system comprises:
One de-quantizer produces a difference signal with this numerical code of de-quantization;
One totalizer is to estimate signal and to form this voice signal in conjunction with this difference signal and; And
One step size automatic regulator is for each frame provides a step change function and step-length higher limit to this de-quantizer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2003101165291A CN100511425C (en) | 2003-11-14 | 2003-11-14 | Method and system for improving ADPCM voice coding step length regulation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2003101165291A CN100511425C (en) | 2003-11-14 | 2003-11-14 | Method and system for improving ADPCM voice coding step length regulation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1617227A CN1617227A (en) | 2005-05-18 |
| CN100511425C true CN100511425C (en) | 2009-07-08 |
Family
ID=34760669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2003101165291A Expired - Fee Related CN100511425C (en) | 2003-11-14 | 2003-11-14 | Method and system for improving ADPCM voice coding step length regulation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100511425C (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103851767A (en) * | 2014-03-12 | 2014-06-11 | 上海夏普电器有限公司 | Design method for condensation prevention of air outlet of air conditioning indoor unit |
-
2003
- 2003-11-14 CN CNB2003101165291A patent/CN100511425C/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN1617227A (en) | 2005-05-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| RU2608878C1 (en) | Level adjustment in time domain for decoding or encoding audio signals | |
| AU2008346515B2 (en) | Audio encoder and decoder | |
| RU2335809C2 (en) | Audio coding | |
| CN1735925B (en) | Reducing scale factor transmission cost for MPEG-2 AAC using a lattice | |
| US7512535B2 (en) | Adaptive postfiltering methods and systems for decoding speech | |
| US7693710B2 (en) | Method and device for efficient frame erasure concealment in linear predictive based speech codecs | |
| KR101147878B1 (en) | Coding and decoding methods and devices | |
| US20120253797A1 (en) | Multi-mode audio codec and celp coding adapted therefore | |
| US7324937B2 (en) | Method for packet loss and/or frame erasure concealment in a voice communication system | |
| US9252728B2 (en) | Non-speech content for low rate CELP decoder | |
| RU2719008C1 (en) | Audio encoder for encoding an audio signal, a method for encoding an audio signal and a computer program which take into account a detectable spectral region of peaks in the upper frequency range | |
| JP2005513539A (en) | Signal modification method for efficient coding of speech signals | |
| US9940942B2 (en) | Advanced quantizer | |
| RU2648953C2 (en) | Noise filling without side information for celp-like coders | |
| AU768744B2 (en) | Method for quantizing speech coder parameters | |
| US20080275709A1 (en) | Audio Encoding and Decoding | |
| CN104517612B (en) | Variable bitrate coding device and decoder and its coding and decoding methods based on AMR-NB voice signals | |
| US11636863B2 (en) | Stereo signal encoding method and encoding apparatus | |
| US6012026A (en) | Variable bitrate speech transmission system | |
| CN100511425C (en) | Method and system for improving ADPCM voice coding step length regulation | |
| RU2707144C2 (en) | Audio encoder and audio signal encoding method | |
| KR102428419B1 (en) | time noise shaping | |
| US20200194014A1 (en) | Method And Apparatus For Reconstructing Signal During Stereo Signal Encoding | |
| AU2012201692A1 (en) | Audio Encoder and Decoder | |
| CN1327410C (en) | Method and apparatus for transcoding between different speech encoding/decoding systems and recording medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090708 Termination date: 20181114 |