Classifications

• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
  • G10L19/0212—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders using orthogonal transformation
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
  • G10L19/022—Blocking, i.e. grouping of samples in time; Choice of analysis windows; Overlap factoring
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
  • G10L19/16—Vocoder architecture
  • G10L19/173—Transcoding, i.e. converting between two coded representations avoiding cascaded coding-decoding
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
  • G10L19/16—Vocoder architecture
  • G10L19/18—Vocoders using multiple modes
  • G10L19/20—Vocoders using multiple modes using sound class specific coding, hybrid encoders or object based coding
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
  • G10L19/26—Pre-filtering or post-filtering
• • H—ELECTRICITY
  • H03—ELECTRONIC CIRCUITRY
  • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
  • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
  • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction

Definitions

• the present invention relates to the field of coding digital signals.
• the coding according to the invention is particularly suitable for the transmission and / or storage of digital audio signals such as audio-frequency signals (speech, music or other).
• the invention advantageously applies to the unified coding of speech, music and mixed content signals, by means of multi-mode techniques that alternate at least two coding modes and whose algorithmic delay is adapted to conversational applications (typically ⁇ 40 ms).
• CELP Code Excited Linear Prediction
• ACELP ACELP variant
• the linear prediction coders are predictive coders. They aim to model the production of speech from at least some of the following: a short-term linear prediction to model the vocal tract, a long-term prediction to model the vibration of the vocal cords into voiced period, and an excitation derived from a vector quantization dictionary usually called fixed dictionary (white noise, algebraic excitation) to represent ⁇ "innovation" which could not be modeled by prediction.
• a short-term linear prediction to model the vocal tract
• a long-term prediction to model the vibration of the vocal cords into voiced period
• fixed dictionary white noise, algebraic excitation
• transform coders MPEG AAC encoder or ITU-T G.722.1 Annex C, for example
• MDCT type critical transforms for "Modified Discrete Transform"
• a “critical-sampling transform” is a transform for which the number of coefficients in the transformed domain is equal to the number of time samples analyzed.
• LPD Linear Predictive Domain
• FD mode (For "Frequency Domain") using an MDCT technique (For "Modified Discrete Transform")
• MDCT Modified Discrete Transform
• the CELP coding - including its ACELP variant - is a predictive coding based on the source-filter model.
• the filter generally corresponds to an all-pole transfer function filter 1 / A (z) obtained by linear prediction (LPC for Linear Predictive
• the synthesis uses the quantized version, 1 / ⁇ ( ⁇ ), of the filter 1 / A (z).
• the source - that is, the excitation of the predictive linear filter 1 / A (z) - is in general the combination of long-term prediction excitation modeling the vibration of the vocal cords, and a stochastic excitation (or innovation) described in the form of algebraic codes (ACELP), noise dictionaries, etc.
• the search for "optimal" excitation is carried out by minimizing a quadratic error criterion in the domain of the signal weighted by a transfer function filter (A) (z) generally derived from the linear prediction filter A (z).
• CELP Code Division Multiple Access
• BV16 BV32
• iLBC iLBC
• SILK coders which are still based on linear prediction.
• predictive coding including CELP coding
• the MDCT transform coding is divided into three steps for the encoder:
• Transformation DCT-IV Discrete Cosine Transform
• TDAC transformation type calculation variants that can use for example a Fourier transform (FFT) instead of a DCT transform.
• FFT Fourier transform
• the MDCT window is generally divided into 4 adjacent portions of equal lengths called "quarters".
• the signal is multiplied by the analysis window and the folds are made: the first quarter (windowed) is folded (that is to say inverted in time and overlapped) on the second and the fourth quarter is folded on the third.
• folding from one quarter to another is done in the following way: The first sample of the first quarter is added (or subtracted) to the last sample of the second quarter, the second sample of the first quarter is added (or subtracted) the second-last sample of the second quarter, and so on until the last sample of the first quarter that is added (or subtracted) to the first sample of the second quarter.
• temporal folding corresponds to mixing two temporal segments and the relative level of two temporal segments in each "folded quarter" is a function of the analysis / synthesis windows.
• the resolution of the systems of equations mentioned is generally made by unfolding, multiplication by a wisely selected synthesis window then addition-recovery of the common parts.
• This overlap addition ensures at the same time the smooth transition (without discontinuity due to quantization errors) between two consecutive decoded frames, in fact this operation behaves like a crossfade.
• the window for the first or fourth quarter is zero for each sample, we are talking about an MDCT transformation without time folding in this part of the window.
• the smooth transition is not ensured by the MDCT transformation, it must be done by other means such as for example an external crossfade.
• Transform coding (including MDCT encoding) can theoretically easily accommodate different input and output sampling rates, as illustrated by the combined implementation in Appendix C of G.722.1 including G.722.1 coding; However, it is also possible to use transform coding with pre / post-processing operations with resampling (by FIR filter, filter banks or IIR filter), possibly with separate coding of the high band which may be an extension of parametric band - these high-band resampling and coding operations are not discussed here, but the 3GPP e-AAC + coder provides an example of such a combination (resampling, low-band transform coding, and band extension).
• the acoustic band coded by the different modes can vary according to the selected mode and the bit rate.
• the mode decision can be made in open loop (or "open-loop" in English) for each frame, that is to say that the decision is made a priori based on data and observations available , or in closed loop as in the AMR-WB + coding.
• the transitions between LPD and FD modes are important to ensure sufficient quality without switching faults, since the FD and LPD modes are different in nature - one is based on a coding by transformed in the frequency domain of the signal, while the other uses predictive linear (time) coding with filter memories that are updated at each frame.
• An example of inter-mode switching management corresponding to the USAC RMO coding is detailed in the article by J. Lecomte et al., "Efficient cross-fade Windows for transitions between LPC-based and non-LPC based audio coding", 7-10 May 2009, 126th AES Convention. As explained in this article, the main difficulty lies in the transitions between LPD to FD modes and vice versa.
• This technique thus makes it possible, during the selection at 110 of the coding mode and the switching (at 150) of the FD-type coding towards LPD, of making it without any transition fault (artifacts) since at the time of coding the frame with the technique LPD, the memories (or states) of the CELP coder (LPD) have already been updated by the generator 160 from the reconstructed signal S a (n) of the frame m.
• the technique described in the application WO2013 / 016262 proposes a step of resampling the memories of the FD type encoder.
• This technique has the disadvantage on the one hand that it requires access to the decoded signal encoder and thus force a local synthesis in the encoder.
• it requires to update the memories of the filters (which may include a resampling step) at the time of encoding and decoding FD type, and a set of operations returning to achieve a CELP type analysis / coding in the previous FD type frame.
• These operations can be complex and are superimposed on conventional coding / decoding operations in the LPD type transition frame, which causes a peak of complexity in "multi-mode" coding.
• the present invention improves the situation.
• decoding according to an inverse transform decoding of a previous frame of samples of the digital signal received and coded according to a transformed pa r coding; decoding according to a predictive decoding of a current frame of samples of the digital signal received and coded according to a predictive coding.
• the method is such that the predictive decoding of the current frame is a transition predictive decoding that does not use an adaptive dictionary from the previous frame and that further comprises:
• an addition-overlay step which combines a synthesized signal segment by predictive decoding of the current frame and a signal segment synthesized by inverse transform decoding, corresponding to a stored segment of the decoding of the previous frame.
• the states reset is performed without the need for the decoded signal of the previous frame, it is performed very simply by zero or predetermined constant values.
• the complexity of the decoder is thus reduced compared to state memory updating techniques requiring analysis or other calculations.
• the transition artifacts are then avoided by the implementation of the addition-overlap step which makes it possible to make the link with the previous frame.
• transition predictive decoding it is not necessary to reset the adaptive dictionary memories for this current frame since it is not used. This further simplifies the implementation of the transition.
• the inverse transform decoding has a lower processing delay than the predictive decoding and in that the first decoded current frame segment by predictive decoding is replaced by a segment resulting from the decoding of the corresponding previous frame. delay offset and stored when decoding the previous frame.
• the signal segment synthesized by inverse transform decoding is corrected before the addition-overlap step by applying an inverse window compensating the windowing previously applied to the segment.
• the decoded current frame has an energy that is close to that of the original signal.
• the signal segment synthesized by inverse transform decoding is pre-sampled at the sampling frequency corresponding to the decoded signal segment of the current frame. This makes it possible to perform a flawless transition in the case where the sampling rate of the transform decoding is different from that of the predictive decoding.
• a state of the predictive decoding is in the list of the following states:
• the calculation of the coefficients of the linear prediction filter of the current frame is performed by decoding the coefficients of a single filter and by assigning identical coefficients to the end linear prediction filter. , middle and start of frame.
• the frame start coefficients are not known.
• the decoded values are then used to obtain the coefficients of the linear prediction filter of the complete frame. This is therefore done in a simple manner without providing significant degradation to the decoded audio signal.
• the calculation of the coefficients of the linear prediction filter of the current frame comprises the following steps:
• the coefficients corresponding to the frame medium filter are decoded with a smaller error.
• the coefficients of the frame start linear prediction filter are reset to a predetermined value corresponding to an average value of the long-term prediction filter coefficients and in that the linear prediction coefficients of the frame current are determined using the thus predetermined values and decoded values of the end-of-frame filter coefficients.
• the frame start coefficients are considered to be known with the predetermined value. This makes it possible to find the coefficients of the complete frame more accurately and to stabilize the predictive decoding more quickly.
• a predetermined default value depends on the type of frame to be decoded.
• the decoding is well adapted to the signal to be decoded.
• the invention also relates to a method for encoding a digital audio signal, comprising the steps of:
• the method is such that the predictive coding of the current frame is a transition predictive coding that does not use an adaptive dictionary from the previous frame and that further comprises:
• a step of resetting at least one state of the predictive coding to a predetermined default value a step of resetting at least one state of the predictive coding to a predetermined default value.
• the reset of the states is carried out without the need for reconstruction of the signal of the previous frame and thus of local decoding. It is performed very simply by zero or predetermined constant values. The complexity of the coding is thus reduced compared to the techniques for updating state memories that require analysis or other calculations.
• the coefficients of the linear prediction filter are part of at least one state of the predictive coding and in that the calculation of the coefficients of the linear prediction filter of the current frame is performed by determining the values encoded coefficients of a single prediction filter, either middle or end of frame and assignment of identical coded values for the coefficients of the frame start and end prediction filter or frame medium.
• At least one state of the predictive coding is coded directly.
• the bits normally reserved for encoding the set of coefficients of the frame medium or frame start filter are used, for example, to directly code at least one state of the predictive coding, for example the memory of the deemphasis filter.
• the coefficients of the linear prediction filter are part of at least one state of the predictive coding and in that the calculation of the coefficients of the linear prediction filter of the current frame comprises the following steps:
• the coefficients corresponding to the frame medium filter are coded with a lower error percentage.
• the coefficients of the linear prediction filter are part of at least one state of the predictive coding
• the coefficients of the linear start-of-frame prediction filter are reset to a predetermined value corresponding to an average value of the long-term prediction filter coefficients and that the linear prediction coefficients of the frame current are determined using the thus predetermined values and the coded values of the coefficients of the end of frame filter.
• the frame start coefficients are considered as known with the predetermined value. This makes it possible to obtain a good estimate of the prediction coefficients of the preceding frame, without additional analysis, in order to calculate the prediction coefficients of the complete frame.
• a predetermined default value depends on the type of frame to be coded.
• the invention also relates to a digital audio signal decoder, comprising:
• an inverse transform decoding entity adapted to decode a previous frame of samples of the digital signal received and coded according to transform coding
• a predictive decoding entity capable of decoding a current frame of samples of the digital signal received and encoded according to a predictive coding.
• the decoder is such that the predictive decoding of the current frame is a transition predictive decoding that does not use an adaptive dictionary derived from the previous frame and that it further comprises:
• a reset module capable of resetting at least one predictive decoding state by a predetermined default value
• a processing module capable of effecting an overlay which combines a synthesized signal segment by predictive decoding of the current frame and a signal segment synthesized by inverse transform decoding, corresponding to a stored segment of the decoding of the previous frame.
• the invention relates to a digital audio signal encoder, comprising:
• transform coding entity capable of encoding a previous frame of samples of the digital signal
• the encoder is such that the predictive coding of the current frame is a transition predictive coding that does not use an adaptive dictionary from the previous frame and that it further comprises:
• a reset module capable of resetting at least one state of the predictive coding by a predetermined default value.
• the decoder and the encoder provide the same advantages as the decoding and coding methods that they implement respectively.
• the invention relates to a computer program comprising code instructions for implementing the steps of the decoding method as described above and / or coding as described above, when these instructions are executed by a processor.
• the invention also relates to a storage means, readable by a processor, whether or not integrated into the decoder or the encoder, possibly removable, storing a computer program implementing a decoding method and / or a coding method as described. previously.
• FIG. 1 illustrates a transition method, between a transform coding and a predictive coding, of the state of the art and described previously;
• FIG. 2 illustrates the transition to the coder between a coded frame according to a transform coding and a coded frame according to a predictive coding, according to an implementation of the invention
• FIG. 3 illustrates an embodiment of the coding method and the coder according to the invention
• FIG. 4 illustrates in flowchart form the steps implemented in a particular embodiment, for determining the coefficients of the linear prediction filter during the predictive coding of the current frame, the previous frame having been coded according to a coding by transformed;
• FIG. 5 illustrates the transition to the decoder between a decoded frame according to an inverse transform decoding and a decoded frame according to a predictive decoding, according to an implementation of the invention
• FIG. 6 illustrates an embodiment of the decoding method and the decoder according to the invention
• FIG. 7 illustrates in flowchart form the steps implemented in one embodiment of the invention, for determining the coefficients of the linear prediction filter during the predictive decoding of the current frame, the previous frame having been decoded according to FIG. reverse transform decoding;
• FIG. 8 illustrates the addition-recovery step implemented during decoding according to one embodiment of the invention
• FIG. 9 illustrates a particular mode of implementation of the transition between decoding by transform and predictive decoding when they have different delays
• FIG. 10 illustrates a hardware embodiment of the encoder or decoder according to the invention.
• FIG. 2 schematically illustrates the coding principle during a transition between a transform coding and a predictive coding according to the invention.
• FD transform coder
• LPD predictive encoder
• the windows of the encoder FD are synchronized so that the last non-zero portion of the window (right) corresponds to the end of a new frame of the input signal.
• the division into frames illustrated in Figure 2 includes the "lookahead" (or future signal) and the actually encoded frame is typically shifted in time (delayed) as explained further with respect to Figure 5.
• the coder performs the procedure of folding and transformation by DCT as described in the state of the art (MDCT).
• MDCT state of the art
• the LPD encoder is derived from the ITU-T G.718 coder whose CELP coding operates at an internal frequency of 12.8 kHz.
• the LPD encoder according to the invention can operate at two internal frequencies 12.8 kHz or 16 kHz depending on the flow.
• FIG. 3 illustrates an embodiment of an encoder and a coding method according to the invention.
• the particular embodiment is in the transition frame between a FD transform coded using an MDCT and an ACELP type predictive codec.
• a decision module Determines whether the frame to be processed is to be encoded in predictive coding ACELP or in transform coding FD.
• a complete transform step MDCT is performed (E302) by the transform coding entity 302.
• This step comprises among other things a windowing with an aligned low delay window as illustrated in FIG. folding step and a transformation step in the DCT domain.
• the frame FD is then quantized in a step (E303) by a quantization module 303 and the data thus encoded are written in the bit stream (bitstream) at E305 by the bit stream construction module 305.
• ACELP are reset in a step (E306) to default values predetermined in advance (not necessarily zero). This reset step is implemented by the reset module 306, for at least one state of the predictive coding.
• a predictive coding step for the current frame is then implemented in E308 by a predictive coding entity 308.
• the coded and quantized information is written to the bit stream at step E305.
• This predictive coding E308 may, in a particular embodiment, be a transition coding as defined under the name of C mode 'in the ITU-T G.718 standard, in which the coding of the excitation is direct and n does not use an adaptive dictionary from the previous frame. An independent excitation of the previous frame is then coded.
• This realization allows the LPD-type predictive coders to stabilize much more quickly (compared to a conventional CELP coding that would use an adaptive dictionary that would be zeroed out.) This further simplifies the implementation of the transition according to the invention.
• the coding of the excitation may not be in a transition mode but will use a CELP coding similar to G.718 and may use an adaptive dictionary (without forcing or limiting the classification) or classic CELP coding with adaptive and fixed dictionaries.
• This variant is however less advantageous because the adaptive dictionary has not been recalculated and has been set to zero, the coding will be suboptimal.
• the CELP coding in the TC mode transition frame may be replaced by any other type of coding independent of the previous frame, for example by using the iLBC type coding scheme.
• a step E307 for calculating the coefficients of the linear prediction filter for the current frame is performed by the calculation module 307.
• the predictive coding (block 304) performs two linear prediction analyzes per frame as in the G.718 standard, with a coding of the LPC coefficients in the form of ISF (or LSF equivalently) obtained at the end of frame (NEW) and a very low rate coding of the LPC coefficients obtained in the middle of the frame (MID), with a subframe interpolation between the LPC coefficients of the previous frame end (OLD), and those of the frame current (MID and NEW).
• ISF or LSF equivalently
• the prediction coefficients in the FD-type previous frame (OLD) are not known because no LPC coefficient is encoded in the FD encoder.
• the interpolation of the LPC coefficients in the The Impedance Spectral Pairs (ISP) or Line Spectral Pairs (LSP) domain may be modified in the second LPD frame that follows the transition LPD frame.
• ISP The Impedance Spectral Pairs
• LSP Line Spectral Pairs
• the bits that could be reserved for the coding of the set of middle frame (MID) or frame start LPC coefficients are used for example to directly code at least one state of the predictive coding, for example the memory of the filter de-emphasis.
• a first step E401 is the initialization of the coefficients of the prediction filter and the equivalent representations ISF or LSF according to the implementation of step E306 of FIG. 3, that is to say to predetermined values, for example according to the long-term average value on a prior learning basis of the LSP coefficients.
• Step E402 encodes the end of frame filter coefficients (LSP NEW) and the obtained code values (LEP NEW Q) as well as the predetermined reset values of the start frame filter coefficients (LSP OLD) are used in E403 for encoding the coefficients of the frame middle prediction filter (MID LSP).
• An E404 replacement step of the frame start coefficient values (LSP OLD) by the coded values of the frame medium coefficients (LSP MID Q) is performed.
• Step E405 makes it possible to determine the coefficients of the linear prediction filter of the current frame from these values thus coded (OLD LSP, LSP MID Q, LSP NEW Q).
• the coefficients of the linear prediction filter of the previous frame are initialized to a value which is already available "for free” in a variant of encoder FD using an LPC-type spectral envelope.
• a "normal" coding as used in G.718 can be used, the linear prediction coefficients per sub-frame being calculated as an interpolation between the values of the prediction filters OLD, MID and NEW, this operation thus allows the LPD coder to obtain, without further analysis, a good estimate of the LPC coefficients in the previous frame.
• the LPD coding may by default only encode a set of LPC coefficients (NEW), the previous variants of embodiment are simply adapted to take into account that no set of coefficients is available in the middle of frame (MID).
• the initialization of the states of the predictive coding may be performed with default values predetermined in advance which may, for example, correspond to different type of frame to be encoded (for example the values of initialization can be different if the frame comprises a signal of the voiced or unvoiced type).
• FIG. 5 schematically illustrates the decoding principle during a transition between a transform decoding and a predictive decoding according to the invention.
• FD transform decoder
• LPD predictive decoder
• the transform decoder (FD) uses "Tukey" type low delay synthesis windows (the invention is independent of the type of window used) and the total length of which is equal to two frames (including zero values). ) as shown in the figure.
• an inverse DCT transformation is applied to the decoded frame.
• the latter is unfolded and the synthesis window is applied to the unfolded signal.
• the synthesis windows of the encoder FD are synchronized so that the non-zero part of the window (left) corresponds to a new frame.
• the frame can be decoded to the point A since the signal does not have aliasing before this point.
• the states or memories of the predictive decoding are reset to predetermined values.
• the state memory of the CELP decoding internal frequency resampling filter (12.8 or 16 kHz) at the output frequency fs. It is considered here that resampling can be performed according to the input and internal frequencies by FIR filter, filterbank or IIR filter, knowing that an implementation of FIR type simplifies the use of the state memory which corresponds to the input signal passed.
• BPF low frequency post-filter
• FIG. 6 illustrates an embodiment of a decoder and a decoding method according to the invention.
• the particular embodiment is in the transition frame between a FD transform coded using an MDCT and an ACELP type predictive codec.
• a decision module Determines whether the frame to be processed must be decoded in predictive decoding ACELP or in FD decoding.
• a decoding step E602 by the transform decoding entity 602 makes it possible to obtain the frame in the transformed domain.
• the step may also contain a resampling step at the sampling frequency of the ACELP decoder.
• This step is followed by an inverse MDCT transformation E603 comprising, an inverse DCT transformation, a temporal unfolding, and the application of a synthesis window and an addition-overlay step with the previous frame as described later in reference to Figure 8.
• the part for which the temporal folding has been canceled is framed in a step E605 by the framing module 605.
• the part which includes time folding is kept in memory (MDCT mem) to make a step of adding -recoverment in E609 by the processing module 609 with the possible next frame decoded by the core FD.
• the memorized portion of the MDCT decoding that is used for the addition-overlap step does not include time-folding, for example in the case where there is a sufficiently large time shift between the MDCT decoding and the decoding. CELP.
• Step E609 uses the transform coder memory (MDCT mem) as described above, i.e. say the decoded signal after the point A but which has folding (in the illustrated case).
• MDCT mem transform coder memory
• the signal is used up to the point B which is the folding point of the transform.
• this signal is first compensated by the inverse of the window previously applied to the segment AB.
• the segment AB is corrected by the application of an inverse window compensating the windowing previously applied to the segment. The segment is no longer "windowed” and its energy is close to that of the original signal.
• the two AB segments are then weighted and summed in order to obtain the final AB signal.
• the weighting functions preferably have a sum equal to 1 (of the linear or quadratic sinusoidal type, for example).
• the addition-overlay step combines a synthesized signal segment by predictive decoding of the current frame and a signal segment synthesized by inverse transform decoding, corresponding to a stored segment of the decoding of the previous frame.
• the signal segment synthesized by FD type inverse transform decoding is previously resampled to the sampling frequency. corresponding to the decoded signal segment of the current LPD frame.
• This re-sampling of the MDCT memory can be done with or without delay with conventional techniques by FIR type filter, filter bank, IIR filter or using "splines".
• the coding modes FD and LPD operate at different internal sampling frequencies, it will be possible, in an alternative, to resample the synthesis of the CELP coding (possibly post-processed with in particular the addition of a high band estimated or coded) and apply the invention.
• This resampling of the synthesis of the LPD coder can be done with or without delay with conventional techniques by FIR type filter, filter bank, IIR filter or using "splines".
• an intermediate delay step (E604) in order to temporally align the two decoders if the decoder FD has less delay than the decoder CELP (LPD).
• a signal part whose size corresponds to the delay between the two decoders is then stored in memory (Delay Mem).
• Figure 9 illustrates this case.
• the embodiment here proposes advantageously taking advantage of this difference in delay D to replace the first segment D resulting from the predictive decoding LPD by that resulting from the FD transform decoding and then to proceed to the addition-recovery step (E609 ) as previously described, on segment AB.
• the inverse transform decoding has a lower processing delay than that of the predictive decoding
• the first current frame segment decoded by predictive decoding is replaced by a segment derived from the decoding of the previous frame corresponding to the delay offset and set to memory when decoding the previous frame.
• a reset step (E606) of the states of the predictive decoding ACELP is applied. This reset step is implemented by the reset module 606, for at least one state of the predictive decoding. Reset values are pre-determined default values (not necessarily zero).
• the initialization of the LPD decoding states can be done with pre-determined default values that can for example correspond to different type of frame to be decoded according to what was done during the encoding.
• a predictive decoding step for the current frame is then implemented in E608 by a predictive decoding entity 608, before the addition-overlap step (E609) described above.
• the step may also contain a resampling step at the sampling rate of the MDCT decoder.
• this predictive coding E608 may be a predictive transition decoding, if this solution has been chosen at the encoder, in which the decoding of the excitation is direct and does not use an adaptive dictionary. In this case, the adaptive dictionary memory does not need to be reset. Non-predictive decoding of the excitation is then performed.
• This embodiment allows the predictive decoders LPD type to stabilize much faster because in this case, it does not use the adaptive dictionary memory that had been previously reset. This further simplifies the implementation of the transition according to the invention.
• the predictive decoding of the long-term excitation is replaced by a non-predictive decoding of the excitation.
• a step E607 for calculating the coefficients of the linear prediction filter for the current frame is performed by the calculation module 607.
• the prediction coefficients in the FD-type previous frame are not known because no LPC coefficient is encoded in the FD encoder and the values have been reset to zero.
• a first step E701 is the initialization of the prediction filter coefficients (LSP OLD) according to the implementation of the step E606 of FIG. 6.
• the step E702 decodes the coefficients of the end of frame filter (LSP NEW) and the decoded values obtained (LSP NEW) as well as the predetermined reset values of the start frame filter coefficients (LSP OLD) are used together in E703 to decode the coefficients of the frame medium prediction filter (MID LSP).
• An E704 replacement step of the frame start coefficient values (LSP OLD) by the decoded values of the frame medium coefficients (MID LSP) is performed.
• Step E705 makes it possible to determine the coefficients of the linear prediction filter of the current frame from these values thus decoded (LSP OLD, MID LSP, LSP NEW).
• the coefficients of the linear prediction filter of the previous frame are initialized to a predetermined value, for example according to the long-term average value of the LSP coefficients.
• a "normal" decoding as used in G.718 can be used, the subframe linear prediction coefficients being calculated as an interpolation between the values of the prediction filters OLD, MID and NEW. This operation allows the LPD encoder to stabilize more quickly.
• FIG. 10 there is described a hardware device adapted to realize an encoder or decoder according to an embodiment of the present invention.
• This encoder or decoder can be integrated into a communication terminal, a communication gateway or any type of equipment such as a set top box, or audio stream reader.
• This device DISP comprises an input for receiving a digital signal which in the case of the encoder is an input signal x (n) and in the case of the decoder, the bit stream bst.
• the device also comprises a processor PROC of digital signals adapted to perform coding / decoding operations in particular on a signal from the input E.
• This processor is connected to one or more MEM memory units adapted to store information necessary for controlling the device for coding / decoding.
• these memory units include instructions for implementing the decoding method described above and in particular for implementing the decoding steps according to an inverse transform decoding of a previous frame of samples of the digital signal.
• these memory units include instructions for implementing the coding method described above and in particular for implementing the steps of coding a previous frame of samples of the digital signal according to a transform encoding, receiving a current frame of samples of the digital signal to be coded according to a predictive coding, a step of resetting at least one state of the predictive coding to a predetermined default value.
• These memory units may also include calculation parameters or other information.
• a storage means readable by a processor, integrated or not integrated with the encoder or the decoder, possibly removable, stores a computer program implementing a decoding method and / or a method of coding according to the invention.
• Figures 3 and 6 may for example illustrate the algorithm of such a computer program.
• the processor is also adapted to store results in these memory units.
• the device comprises an output S connected to the processor for providing an output signal which in the case of the encoder is a signal in the form of bst bit stream and in the case of the decoder, an output signal x (n).

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Computational Linguistics (AREA)
• Signal Processing (AREA)
• Health & Medical Sciences (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Human Computer Interaction (AREA)
• Acoustics & Sound (AREA)
• Multimedia (AREA)
• Spectroscopy & Molecular Physics (AREA)
• Theoretical Computer Science (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)

Priority Applications (10)

Applications Claiming Priority (2)

Publications (2)

Family

ID=50179701

Family Applications (1)

Country Status (11)

Families Citing this family (6)

Citations (1)

Family Cites Families (37)

• 2013
  • 2013-11-15 FR FR1361243A patent/FR3013496A1/fr not_active Withdrawn
• 2014
  • 2014-11-14 BR BR112016010522-2A patent/BR112016010522B1/pt active IP Right Grant
  • 2014-11-14 US US15/036,984 patent/US9984696B2/en active Active
  • 2014-11-14 EP EP14821711.0A patent/EP3069340B1/fr active Active
  • 2014-11-14 CN CN201480062220.5A patent/CN105723457B/zh active Active
  • 2014-11-14 MX MX2016006253A patent/MX353104B/es active IP Right Grant
  • 2014-11-14 ES ES14821711.0T patent/ES2651988T3/es active Active
  • 2014-11-14 KR KR1020217018976A patent/KR102388687B1/ko active Active
  • 2014-11-14 KR KR1020167014550A patent/KR102289004B1/ko active Active
  • 2014-11-14 RU RU2016123462A patent/RU2675216C1/ru active
  • 2014-11-14 WO PCT/FR2014/052923 patent/WO2015071613A2/fr not_active Ceased
  • 2014-11-14 JP JP2016529922A patent/JP6568850B2/ja active Active

Patent Citations (1)

Non-Patent Citations (2)

Also Published As

Similar Documents

Legal Events