EP3175444B1 - Frame loss management in an fd/lpd transition context - Google Patents

Description

The present invention relates to the field of coding / decoding of digital signals in particular for the loss of frame correction.

The invention is advantageously applied to the coding / decoding of sounds that can contain speech and music mixed or alternately.

To effectively code low-speed speech sounds, CELP ( Code Excited Linear Prediction ) techniques are recommended. To effectively encode musical sounds, transform coding techniques are preferred.

CELP coders are predictive coders. They aim to model the production of speech from various elements: a short-term linear prediction to model the vocal tract, a long-term prediction to model the vibration of vocal cords in voiced period, and an excitation derived from a fixed dictionary (white noise, algebraic excitation) to represent the "innovation" which could not be modeled.

Transform coders such as MPEG AAC, AAC-LD, AAC-ELD or ITU-T G.722.1 Annex C use critical-sampling transforms to compact the signal in the transformed domain. 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 in each frame analyzed.

One solution for efficiently coding a mixed speech / music content signal consists in selecting, over time, the best technique between at least two coding modes, one of the CELP type, the other of the transformed type.

This is the case, for example, of the 3GPP AMR-WB + and MPEG USAC codecs (for "Unified Speech Audio Coding" ). The applications targeted by AMR-WB + and USAC are not conversational, but correspond to broadcasting and storage services, without strong constraints on the algorithmic delay.

• Pour les signaux de type parole: modes LPD (pour "Linear Prédictive Domain" en anglais) comprenant deux modes différents dérivés du codage AMR-WB+:
  • Un mode ACELP
  • Un mode TCX (pour « Transform Coded eXcitation » en anglais) appelé wLPT (pour "weighted Linear Prédictive Transform" en anglais) utilisant une transformée de type MDCT (contrairement au codec AMR-WB+) qui utilise une transformée FFT.
• Pour les signaux de type musique: mode FD (pour "Frequency Domain" en anglais) utilisant un codage par transformée MDCT (pour "Modified Discrete Cosine Transform" en anglais) de type MPEG AAC (pour "Advanced Audio Coding" en anglais) sur 1024 échantillons.

The initial version of the USAC codec, called Reference Model 0 (RMO), is described in the M. Neuendorf et al., A Novel Scheme for Low Bitrate Unified Speech and Audio Coding - MPEG RM0, 7-10 May 2009, 126th AES Conventi we. This RMO codec alternates between several encoding modes:

• For speech type signals: LPD (for Linear Predictive Domain ) modes comprising two different modes derived from the AMR-WB + coding:
  • ACELP mode
  • A mode TCX (for " Transform Coded eXcitation " in English) called wLPT (for "weighted Linear Predictive Transform" in English) using an MDCT type transform (unlike the AMR-WB + codec) that uses an FFT transform.
• For music type signals: FD mode (for "Frequency Domain" in English) using a transform coding MDCT (for "Modified Discrete Cosine Transform" in English) MPEG AAC type (for "Advanced Audio Coding" in English) on 1024 samples.

In the USAC codec, the transitions between LPD and FD modes are crucial to ensure sufficient quality without switching faults, knowing that each mode (ACELP, TCX, FD) has a specific "signature" (in terms of artifacts) and that the FD and LPD modes are different in nature - the FD mode is based on transform coding in the signal domain, while the LPD modes use predictive linear coding in the perceptually weighted domain with filter memories to be handled correctly. The management of intermode switches in the USAC RMO codec is detailed in the article of 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 Conventi we. As explained in this article, the main difficulty lies in the transitions between LPD to FD modes and vice versa. We retain here only the case of transitions from ACELP to FD.

In order to understand how it works, the principle of transform coding MDCT is described here through a typical embodiment.

• Pondération du signal par une fenêtre appelé ici "fenêtre MDCT" de longueur 2M ;
• Repliement temporel (ou "time-domain aliasing" en anglais) pour former un bloc de longueur M ;
• Transformation DCT (pour "Discrete Cosine Transform" en anglais) de longueur M.

At the coder, the MDCT transformation is typically divided into three steps, the signal being cut into frames of M samples before MDCT coding:

• Weighting of the signal by a window called here "MDCT window" of length 2M;
• Time folding (or "time-domain aliasing" in English) to form a block of length M;
• Transformation DCT (for "Discrete Cosine Transform" in English) of length M.

The MDCT window is divided into 4 adjacent portions of equal lengths M / 2, here called "quarters".

The signal is multiplied by the analysis window and folds are made: the first quarter (windowed) is folded (ie inverted in time and overlapped) on the second quarter and the fourth quarter is folded on the third.

More precisely, the temporal 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 ) to the second-last sample in the second quarter, and so on until the last sample in the first quarter that is added (or subtracted) to the first sample in the second quarter.

We thus obtain, from 4 quarters, 2 folded quarters where each sample is the result of a linear combination of 2 samples of the signal to be coded. This linear combination induces a temporal folding.

The folded two quarters are then coded together after DCT (Type IV) transformation. For the next frame we shift one half of window (or 50% overlap), the third and fourth quarters of the previous frame then become the first and second quarters of the current frame. After folding, we send a second linear combination of the same pairs of samples as in the previous frame, but with different weights.

At the decoder, after inverse DCT transformation, the decoded version of these folded signals is thus obtained. Two consecutive frames contain the result of 2 different folds of the same quarters, that is to say for each pair of samples one has the result of 2 linear combinations with different but known weights: an equation system is thus solved to obtain the decoded version of the input signal, the time folding can be thus removed using 2 consecutive decoded frames.

The resolution of the systems of equations mentioned can in general be made implicitly by unfolding, multiplication by a wisely chosen 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. When the window for the first quarter or fourth quarter is zero for each sample, we are talking about an MDCT transformation without time folding in this part of the window. In this case the smooth transition is not ensured by the MDCT transformation, it must be done by other means such as for example an external crossfade.

It should be noted that implementation variants of the MDCT transformation exist, in particular on the definition of the DCT transform, on how to fold the block to be transformed temporarily (for example, the signs applied to the folded quarters can be reversed). left and right, or fold the second and third quarter over the first and fourth quarters respectively), etc. These variants do not change the principle of the MDCT analysis-synthesis with the reduction of the sample block by windowing, temporal folding then transformation and finally windowing, folding and addition-recovery.

In order to avoid artifacts at the time of transitions between CELP type coding and MDCT coding, the international patent application WO2012 / 085451 , the contents of which are incorporated by reference in the present application, proposes a method for encoding a transition frame. The transition frame is defined as the transform-encoded current frame that succeeds a previous frame encoded by predictive coding. According to the new method mentioned above, a part of the transition frame, for example a subframe of 5 ms, in the case of a CELP core coding at 12.8 kHz, and two additional CELP frames of 4 ms each, in the case of a CELP core coding at 16 kHz, are coded by a predictive coding restricted with respect to the predictive coding of the previous frame.

Restricted predictive coding consists in using the stable parameters of the preceding frame coded by a predictive coding, such as the coefficients of the linear prediction filter and coding only a few minimum parameters for the additional subframe in the transition frame.

Since the previous frame has not been coded with transform coding, the cancellation of the temporal folding in the first part of the frame is impossible. The patent application WO2012 / 085451 Moreover, it is proposed to modify the first half of the MDCT window so as not to have time folding in the first quarter normally folded. It is also proposed to integrate a part of the addition-overlap (also called "fade-in" or " overlap-add " in English) between the decoded CELP frame and the decoded MDCT frame by modifying the coefficients of the window. analysis / synthesis. With reference to the figure 4e of the aforementioned application, the mixed lines (lines alternating points and lines) correspond to the MDCT coding folding lines (top figure) and the MDCT decoding unfolding lines (bottom figure). In the figure above, the bold lines separate the frames of new samples at the input of the encoder. The coding of a new MDCT frame can be started when a so-defined frame of new input samples is fully available. It is important to note that these lines in bold at the coder do not correspond to the current frame but to the block of new samples arriving for each frame: the current frame is in fact delayed by 5 ms which correspond to an anticipation, called " lookahead " in English. In the bottom figure, the bold lines separate the decoded frames at the output of the decoder.

At the encoder, the transition window is zero to the point of folding. Thus the coefficients of the left part of the folded window will be identical to those of the unfolded window. The portion between the folding point and the end of the CELP transition subframe (TR) corresponds to a (half) sinusoidal window. At the decoder, after unfolding, the same window is applied to the signal. On the segment between the folding point and the beginning of the MDCT frame, the coefficients of the window correspond to a window of form sin ² . To ensure the overlap-addition between the decoded subframe CELP and the signal from the MDCT, it suffices to apply a window of cos ² type to the part of the subframe CELP in overlap and to summon the latter with the MDCT frame. The method is perfect reconstruction.

However, coded audio signal frames may be lost on the channel between the encoder and the decoder.

The techniques of correction of loss of existing frames are most often very dependent on the type of coding used. An example is proposed in US2014 / 0019142 A1 . In the case of speech coding based on predictive technologies, such as CELP, the frame loss correction is often related to the speech model. For example, the ITU-T Standard G.722.2, as of July 2003, proposes to replace a lost packet by extending the long-term prediction gain by attenuating it, and by extending the spectral line frequencies (ISF in English). for " Immitance Spectral Frequencies "), representing the coefficients A (z) of the LPC filter, making them tend towards their respective averages. The fundamental period (or "pitch") is also repeated. The contribution of the fixed dictionary is filled with random values. Applying such methods for transform or PCM decoders would require CELP analysis at the decoder, which would introduce significant additional complexity. It should also be noted that more advanced methods of frame loss correction during CELP decoding are described in ITU-T G.718 for both 8 and 12 kbit / s rates as well as interoperable decoding rates with AMR-WB.

Another solution is presented in the ITU-T G.711 standard, which describes a transform coder for which the frame loss correction algorithm, treated in the "Appendix I" part, consists in finding a tonal delay. (a fundamental period) in the already decoded signal and to repeat it by applying an overlap between the already decoded signal and the repeated signal. This overlapping addition erases the audio artifacts but requires additional delay to the decoder (corresponding to the duration of the overlap-addition) to be implemented.

In the case of transform coding, a common technique for correcting a frame loss is to repeat the last frame received. Such a technique is implemented in several standardized coders / decoders (G.719, G.722.1 and G.722.1C in particular). For example, in the case of the G.722.1 decoder, an MLT transform (for " Modulated Lapped Transform "), equivalent to an MDCT transform, with a 50% overlap and a sinusoidal window, makes it possible to ensure slow enough transition between the last lost frame and the repeated frame to erase the artifacts related to the simple repetition of the frame.

Such a technique is inexpensive but its main defect is the inconsistency between the signal just before the loss of frame and the repeated signal. This results in a phase discontinuity that can introduce important audio artifacts if the overlap time between the two frames is low, as is the case when the windows used for the MLT are so-called low delay windows.

At the decoder level, according to the existing techniques, when a frame is missing, a replacement frame is generated by using a suitable packet loss masking algorithm (for " Packet Loss Concealment "). Note that in general a packet may contain several frames, so the term PLC may be ambiguous, and it is here taken to indicate the correction of the current frame lost. For example, following a correctly received and decoded CELP frame, if the next frame is lost, a PLC-based replacement frame adapted to CELP encoding is used, exploiting the CELP encoder memories. Following a MDCT frame correctly received and decoded, if the next frame is lost, a replacement frame based on a PLC adapted to the MDCT encoding is generated.

In the context of the transition between CELP and MDCT frames, and considering that the transition frame is composed of a CELP subframe (which is at the same sampling rate as the previous CELP frame) and a MDCT frame with a modified MDCT window canceling the folding "left", there are situations for which existing techniques provide no solution.

In a first situation, a previous CELP frame has been correctly received and decoded, a current transition frame is lost, and the next frame is an MDCT frame. In this case, the PLC algorithm, after receiving the CELP frame, does not know that the lost frame is a transition frame and therefore generates a replacement CELP frame. Thus, as previously explained, the first folded portion of the next MDCT frame can not be compensated and the delay between the two types of encoder can not be bridged with the CELP subframe contained in the transition frame (which is lost with the transition frame). There is no known solution to deal with this situation.

In a second situation, a previous CELP frame at 12.8 kHz is correctly received and decoded, a current CELP frame at 16 kHz is lost, and the next frame is a transition frame. The PLC algorithm then generates a CELP frame at the frequency of the last correctly received frame, ie 12.8 kHz, and the CELP transition subframe (partially coded from CELP parameters of the 16 kHz CELP frame lost ) can not be decoded.

The present invention improves this situation.

• décodage prédictif d'une trame précédente du signal numérique, codée par un ensemble de paramètres de codage prédictif ;
• détection de la perte d'une trame courante du signal numérique codé;
• génération par prédiction, à partir d'au moins un paramètre de codage prédictif codant la trame précédente, d'une trame de remplacement de la trame courante ;
• génération par prédiction, à partir d'au moins un paramètre de codage prédictif codant la trame précédente, d'un segment supplémentaire de signal numérique ;
• stockage temporaire de ce segment supplémentaire de signal numérique.

For this purpose, a first aspect of the invention relates to a method for decoding a coded digital signal according to a predictive coding and according to a transform coding, comprising the following steps:

• predictive decoding of a previous frame of the digital signal, encoded by a set of predictive coding parameters;
• detecting the loss of a current frame of the encoded digital signal;
• generating by prediction, from at least one predictive coding parameter encoding the previous frame, a replacement frame of the current frame;
• generating by prediction, from at least one predictive coding parameter coding the preceding frame, an additional segment of digital signal;
• temporary storage of this additional segment of digital signal.

Thus, an additional segment of digital signal is available each time a replacement CELP frame is generated. The predictive decoding of the previous frame includes the predictive decoding of a correctly received CELP frame or the generation of a replacement CELP frame by a CELP-adapted PLC algorithm.

On the one hand, this additional segment makes possible a transition between a CELP coding and a transform coding, even in the case of a frame loss.

Indeed, in the first situation described above, the transition with the next MDCT frame can be provided by the additional segment. As described below, the additional segment may be added to the next MDCT frame to compensate for the first folded portion of this MDCT frame by cross-fading the area containing the non-canceled time folding.

In the second situation described above, the decoding of the transition frame is made possible by the use of the additional segment. Indeed, if it is not possible to decode the transition CELP subframe (unavailability of CELP parameters of the previous frame coded at 16 kHz), it is possible to replace it with the additional segment as described above. after.

On the other hand, the calculations relating to the management of the frame loss and the transition are distributed over time. Indeed, the additional segment is generated and stored for each generated replacement CELP frame. The transition segment is therefore generated as soon as a frame loss is detected, without waiting for a transition to be subsequently detected. The transition is therefore anticipated at each frame loss, which avoids having to manage a "complexity peak" at the moment when a new correct frame is received and decoded.

• réception d'une trame suivante de signal numérique codé comprenant au moins un segment codé par transformée ; et
• décodage de la trame suivante comprenant une sous-étape d'addition avec recouvrement entre le segment supplémentaire de signal numérique et le segment codé par transformée. La sous-étape d'addition avec recouvrement rend possible un fondu-enchaîné du signal de sortie. Un tel fondu-enchaîné limite l'apparition d'artefacts sonores (par exemple de type « bruit métallique ») et assure une cohérence énergétique du signal.

In one embodiment, the method further comprises the following steps:

• receiving a next coded digital signal frame comprising at least one transform coded segment; and
• decoding the next frame comprising an overlap adding sub-step between the additional digital signal segment and the transform coded segment. The overlapping addition sub-step makes it possible to cross-fade the output signal. Such a fade-out limits the appearance of sound artifacts (for example of the "metallic noise" type) and ensures an energy coherence of the signal.

In another embodiment, the next frame is fully coded according to transform coding and the lost current frame is a transition frame between the predecessor coded predictive coding frame and the transform coded next frame.

In a variant, the preceding frame is coded according to a predictive coding by a predictive coder core operating at a first frequency. In this variant, the following frame is a transition frame comprising at least one sub-frame coded according to a predictive coding by a predictive coding core operating at a second frequency distinct from the first frequency. For this purpose, the following transition frame may comprise a bit indicating the frequency of the predictive coding core used.

Thus, the CELP encoding type (12.8 or 16 kHz) used in the transition CELP subframe can be indicated in the bitstream of the transition frame. The invention thus provides to add a systematic indication (one bit) in a transition frame, in order to allow the detection of a CELP coding / decoding frequency difference between the transition CELP subframe and the previous CELP frame.

avec :

• r est un coefficient représentatif de la longueur du segment supplémentaire généré ;
• i un instant d'un échantillon de la trame suivante, compris entre 0 et L/r ;
• L la longueur de la trame suivante ;
• S(i) l'amplitude de la trame suivante après addition, pour l'échantillon i ;
• B(i) l'amplitude du segment décodé par transformée, pour l'échantillon i ;
• T(i) l'amplitude du segment supplémentaire de signal numérique, pour l'échantillon i. L'addition avec recouvrement peut donc être effectuée à partir de combinaisons linéaires et d'opérations simples à mettre en oeuvre. Le temps requis pour le décodage est ainsi réduit tout en sollicitant moins le ou les processeurs utilisés pour ces calculs. Dans des variantes, d'autres formes de fondu enchaîné pourront être mises en oeuvre sans changer le principe de l'invention.

In another embodiment, the overlap addition is given by applying the following formula implementing a linear weighting: $$S\left(i\right)=B\left(i\right).\frac{i}{\left(\mathrm{The}/r\right)}+\left(1-\frac{i}{\left(\mathrm{The}/r\right)}\right).T\left(i\right)$$

with:

• r is a coefficient representative of the length of the additional segment generated;
• i a moment of a sample of the next frame, between 0 and L / r;
• L the length of the next frame;
• S (i) the amplitude of the next frame after addition, for sample i;
• B (i) the amplitude of the transform decoded segment, for sample i;
• T (i) the amplitude of the additional digital signal segment, for sample i. The overlap addition can therefore be performed from linear combinations and simple operations to implement. The time required for the decoding is thus reduced while requiring less the processor (s) used for these calculations. In variants, other forms of cross-fade can be implemented without changing the principle of the invention.

• copie dans une mémoire temporaire, des mémoires du décodeur mises à jour lors de l'étape de génération par prédiction de la trame de remplacement ;
• génération du segment supplémentaire de signal numérique au moyen de la mémoire temporaire.

In one embodiment, the step of generating by prediction of the replacement frame further comprising an update of internal memories of the decoder, the step of generating by prediction of an additional segment of digital signal may comprise the sub-frames. -following steps :

• copying into a temporary memory, decoder memories updated during the step of generating by prediction of the replacement frame;
• generating the additional digital signal segment using the temporary memory.

Thus, the internal memories of the decoder are not updated for the generation of the additional segment. Therefore, the generation of the additional signal segment does not impact the decoding of the next frame, in the event that the next frame is a CELP frame.

Indeed, if the next frame is a CELP frame, the internal memories of the decoder must correspond to the states of the decoder at the end of the replacement frame.

• génération par prédiction d'une trame supplémentaire, à partir d'au moins un paramètre de codage prédictif codant la trame précédente ;
• extraction d'un segment de la trame supplémentaire.

In one embodiment, the step of generating by prediction of an additional segment of digital signal comprises the following substeps:

• generation by prediction of an additional frame, from at least one predictive coding parameter encoding the previous frame;
• extraction of a segment of the additional frame.

In this embodiment, the additional digital signal segment corresponds to the first half of the additional frame. Thus, the efficiency of the method is further improved because the temporary calculation data used for the generation of the replacement CELP frame is directly available for the generation of the additional CELP frame. Typically, the registers and caches, on which the temporary calculation data are stored, may not be updated in order to reuse these data directly for the generation of the additional CELP frame.

A second aspect of the invention is directed to a computer program comprising instructions for implementing the method according to the first aspect of the invention, when these instructions are executed by a processor.

• une unité de détection de la perte d'une trame courante du signal numérique ;
• un décodeur prédictif comportant un processeur agencé pour effectuer les opérations suivantes :
  • * décodage prédictif d'une trame précédente du signal numérique, codée par un ensemble de paramètres de codage prédictif ;
  • * génération par prédiction, à partir d'au moins un paramètre de codage prédictif codant la trame précédente, d'une trame de remplacement de la trame courante ;
  • * génération par prédiction, à partir d'au moins un paramètre de codage prédictif codant la trame précédente, d'un segment supplémentaire de signal numérique ;
  • * stockage temporaire de ce segment supplémentaire de signal numérique dans une mémoire temporaire.

A third aspect of the invention provides a decoder of a coded digital signal according to a predictive coding and a transform coding, comprising:

• a unit for detecting the loss of a current frame of the digital signal;
• a predictive decoder comprising a processor arranged to perform the following operations:
  • predictive decoding of a previous frame of the digital signal coded by a set of predictive coding parameters;
  • generation by prediction, from at least one predictive coding parameter coding the preceding frame, of a replacement frame of the current frame;
  • generation by prediction, from at least one predictive coding parameter encoding the previous frame, of an additional segment of digital signal;
  • * Temporary storage of this additional segment of digital signal in a temporary memory.

• * réception d'une trame suivante de signal numérique codé comprenant au moins un segment codé par transformée ; et
• * décodage de la trame suivante comprenant une sous-étape d'addition avec recouvrement entre le segment supplémentaire de signal numérique et le segment codé par transformée.

In one embodiment, the decoder according to the third aspect of the invention further comprises a transform decoder comprising a processor arranged to perform the following operations:

• receiving a next coded digital signal frame comprising at least one transform coded segment; and
• decoding the next frame comprising an overlap adding sub-step between the additional digital signal segment and the transform coded segment.

At the encoder level, the invention may include inserting in the transition frame an information bit on the CELP core used for encoding the transition subframe.

• la figure 1 illustre un décodeur audio selon un mode de réalisation de l'invention ;
• la figure 2 illustre un décodeur CELP d'un décodeur audio, tel que le décodeur audio de la figure 1, selon un mode de réalisation de l'invention.

• la figure 3 est un diagramme illustrant les étapes d'un procédé de décodage, mis en oeuvre par le décodeur audio de la figure 1, selon un mode de réalisation de l'invention ;
• la figure 4 illustre un dispositif de calcul selon un mode de réalisation de l'invention.

Other features and advantages of the invention will appear on examining the detailed description below, and the attached drawings in which:

• the figure 1 illustrates an audio decoder according to an embodiment of the invention;
• the figure 2 illustrates a CELP decoder of an audio decoder, such as the decoder audio of the figure 1 according to one embodiment of the invention.

• the figure 3 is a diagram illustrating the steps of a decoding method, implemented by the audio decoder of the figure 1 according to one embodiment of the invention;
• the figure 4 illustrates a computing device according to one embodiment of the invention.

The figure 1 illustrates an audio decoder 100 according to one embodiment of the invention.

No audio coder structure is presented. However, the coded digital audio signal received by the decoder according to the invention may be derived from an encoder able to encode an audio signal in the form of CELP frames, MDCT frames and CELP / MDCT transition frames, such as the encoder described in the application WO2012 / 085451 . For this purpose, a transition-coded transition frame may furthermore comprise a segment (a sub-frame for example) coded by a predictive coding. The encoder may further add a bit in the transition frame to identify the frequency of the CELP core used. The CELP coding example is given for illustrative purposes to describe any type of predictive coding. Similarly, the MDCT coding example is given for illustrative purposes to describe any type of transform coding.

The decoder 100 comprises a reception unit 101 of a coded digital audio signal. The digital signal is encoded as CELP frames, MDCT frames and CELP / MDCT transition frames. In variants of the invention, other modes than CELP and MDCT modes are possible, and other combinations of modes are possible, without changing the principle of the invention. In addition, the CELP coding may be replaced by another type of predictive coding, and the MDCT coding may be replaced by another type of transform coding.

The decoder 100 further comprises a classification unit 102 capable of determining - generally by simple reading of the bitstream and interpretation of the indications received from the coder - whether a current frame is a CELP frame, an MDCT frame, or a transition frame. Depending on the classification of the current frame, the latter may be transmitted to a CELP decoder 103 or to an MDCT decoder 104 (or both, in the case of a transition frame, the transition CELP subframe being transmitted to a decoding unit 105 described below). Moreover, in the case where the current frame is a correct (or received) transition frame and where the CELP coding can operate at at least two frequencies (12.8 and 16 kHz), the classification unit 102 may determine the CELP encoding type used in the additional CELP subframe - this type of encoding is indicated as an output bit rate of the encoder.

An example of a decoder structure CELP 103 is shown with reference to FIG. figure 2 .

A reception unit 201, which may include a demultiplexing function, is able to receive CELP coding parameters of the current frame. These parameters can include excitation parameters (gain vectors, fixed dictionary vector, adaptive dictionary vector for example) transmitted to a decoding unit 202 capable of generating an excitation. In addition, the CELP coding parameters may comprise LPC coefficients represented in the form of LSF or ISF, for example. The LPC coefficients are decoded by a decoding unit 203 capable of supplying the LPC coefficients to a synthetic LPC filter 205.

The synthesis filter 205, excited by the excitation generated by the unit 202, synthesizes a frame (or generally a subframe) of digital signal transmitted to a de-emphasis or deemphasis filter 206 (function of the form 1 / (1-αz ^-1 ) with for example a = 0.68). At the output of the de-emphasis filter, the CELP decoder 103 may comprise a low-frequency post-processing 207 (or " bass-post filter " in English) similar to that described in the ITU-T G.718 standard. The CELP decoder 103 further comprises a resampling 208 of the signal synthesized at the output frequency (the output frequency of the MDCT decoder 104), and an output interface 209. In variants of the invention, post-processing CELP synthesis can be implemented before or after resampling.

• des états servant au décodage de l'excitation ;
• la mémoire du filtre de synthèse 205 ;
• la mémoire du filtre de dé-emphase 206 ;
• des mémoires du post-traitement 207 ;
• des mémoires de l'unité de ré-échantillonnage 208.

Furthermore, in the case where the digital signal is decomposed into high and low frequency bands before coding, the CELP decoder 103 may comprise a high frequency decoding unit 204, the low frequency signal being decoded by the units 202 to 208 described. above. CELP synthesis may involve updating internal CELP coder states (or updating internal memories), such as:

• states for decoding the excitation;
• the memory of the synthesis filter 205;
• the memory of the de-emphasis filter 206;
• post-processing memories 207;
• memories of the resampling unit 208.

With reference to the figure 1 , the decoder further comprises a loss of frame management unit 108 and a temporary memory 107.

In order to decode a transition frame, the decoder 100 further comprises a decoding unit 105 able to receive the transition CELP subframe and the decoded transition frame by transforming the output of the MDCT decoder 104 in order to decode the frame of the decoder. addition-based transition with recovery of the received signals. The decoder 100 may further comprise an output interface 106.

The operation of the decoder 100 according to the invention will be better understood with reference to the figure 3 which is a diagram showing the steps of a method according to an embodiment of the invention.

At a step 301, a coded digital audio signal current frame may or may not be received by the reception unit 101 from an encoder. It is considered that the previous audio signal frame is a correctly received and decoded frame or a frame of replacement.

It is detected at a step 302 if the coded current frame is missing or if it has been received by the reception unit 101.

In the case where the coded current frame has been received, it is determined in a step 303, by the classification unit 102, whether the coded current frame is a CELP frame.

In the case where the coded current frame is a CELP frame, the method comprises a step 304 of decoding and resampling the coded CELP frame, by the decoder CELP 103. The aforementioned internal memories of the decoder CELP 103 can then be set in step 305. At a step 306, the decoded and resampled signal is outputted from the decoder 100. The excitation parameters of the current frame, as well as the LPC coefficients, can be stored in the memory 107. .

In the case where the coded current frame is not a CELP frame, the current frame comprises at least one segment coded according to a transform coding (MDCT frame or transition frame). It is then checked at a step 307 if the coded current frame is an MDCT frame. If this is the case, the current frame is decoded at a step 308 by the MDCT decoder 104 and the decoded signal is transmitted at the output of the decoder 100 in step 306.

If, on the other hand, the current frame is not an MDCT frame, then it is a transition frame which is decoded at a step 309 by decoding both the CELP transition subframe and the MDCT transformed current frame and by performing the overlap addition of the signals from the CELP decoder and the MDCT decoder to obtain a digital signal transmitted at the output of the decoder 100 in step 306.

In the case where the current subframe has been lost, it is determined in a step 310 whether the previous received and decoded frame was a CELP frame. If this is not the case, a PLC algorithm adapted to the MDCT, implemented in the lossy frame management unit 108, generates a replacement frame MDCT decoded by the decoder MDCT 104 in order to obtain a signal digital output, at a step 311.

If the last frame correctly received was a CELP frame, a PLC algorithm adapted to the CELP is implemented by the frame loss management unit 108 and the CELP decoder 103 to generate a replacement CELP frame at a step 312.

• estimation par interpolation des paramètres LSF et du filtre LPC en fonction des paramètres LSF de la trame précédente, en mettant à jour, à une étape 313, les mémoires des quantificateurs prédictifs LSF (qui peuvent être par exemple de type AR or MA) - un exemple de mise en oeuvre de l'estimation des paramètres LPC en cas de perte de trame pour le cas des paramètres ISF est donné dans les clauses 7.11.1.2 « ISF estimation and interpolation » et 7.11.1.7 « Spectral envelope concealment, synthesis, and updates » de la norme UIT-T G.718. Alternativement l'estimation décrite dans la clause I.5.2.3.3 de la norme UIT-T G.722.2 Appendix I pourra également être utilisée dans le cas d'une quantification de type MA ;
• estimation de l'excitation à partir du gain adaptatif et du gain fixe de la trame précédente, en mettant ces valeurs à jour, à l'étape 313, pour la trame suivant. Un exemple d'estimation de l'excitation est décrit dans les clauses 7.11.1.3 « Extrapodation of future pitch », 7.11.1.4 « Construction of the periodic part of the excitation », 7.11.1.15 « Glottal pulse resynchronization in low-delay », 7.11.1.6 «Construction of the random part of the excitation ». Le vecteur du dictionnaire fixe est typiquement remplacé dans chaque sous-trame par un signal aléatoire, le dictionnaire adaptatif utilise un pitch extrapolé et les gains de dictionnaires issus de la trame précédente ont typiquement été atténués selon la classe du signal dans la dernière trame reçue. Alternativement l'estimation de l'excitation décrite dans la norme UIT-T G.722.2 Appendix I pourra également être utilisée ;
• synthétiser le signal à partir de l'excitation et du filtre de synthèse 205 mis à jour et en utilisant la mémoire de synthèse de la trame précédente, en mettant à jour la mémoire de synthèse de la trame précédente à l'étape 313 ;
• dé-emphase du signal synthétisé en utilisant l'unité de dé-emphase 206, et en mettant à jour, à l'étape 313, la mémoire de l'unité de dé-emphase 206 ;
• de façon optionnelle, post-traitement 207 du signal synthétisé en mettant à jour, à l'étape 313, la mémoire du post-traitement - on peut noter que le post-traitement peut être désactivé pendant la correction de perte de trame car les informations qu'ils utilisent ne sont pas fiables car simplement extrapolées, dans ce cas les mémoires du post-traitement doivent quand-même être mise à jour pour permettre un fonctionnement normal à la prochaine trame reçue ;
• ré-échantillonnage du signal synthétisé à la fréquence de sortie par le ré-échantillonnage 208, en mettant à jour la mémoire du filtre 208 à l'étape 313.

The PLC algorithm can include the following steps:

• estimation by interpolation of the LSF parameters and of the LPC filter according to the LSF parameters of the preceding frame, by updating, in a step 313, the memories of the predictive quantifiers LSF (which may be for example of the AR or MA type) - a example of implementation of the estimation of the LPC parameters in case of loss of frame for the case of the ISF parameters is given in clauses 7.11.1.2 " ISF estimation and interpolation " and 7.11.1.7 " Spectral envelope concealment, synthesis, and updates "of ITU-T G.718. Alternatively the estimation described in clause I.5.2.3.3 of the ITU-T G.722.2 Appendix I may also be used in the case of AM type quantization;
• estimating the excitation from the adaptive gain and the fixed gain of the previous frame, updating these values, in step 313, for the next frame. An example of estimation of excitation is described in clauses 7.11.1.3 " Extrapodation of future pitch", 7.11.1.4 " Construction of the periodic part of the excitation ", 7.11.1.15 " Glottal pulse resynchronization in low-delay " , 7.11.1.6 "Construction of the random part of the excitation". The vector of the fixed dictionary is typically replaced in each subframe by a random signal, the adaptive dictionary uses an extrapolated pitch and the dictionary gains from the previous frame have typically been attenuated according to the class of the signal in the last received frame. Alternatively the excitation estimate described in ITU-T G.722.2 Appendix I may also be used;
• synthesizing the signal from the exciter and the updated synthesis filter 205 and using the synthesis memory of the previous frame, updating the synthesis memory of the previous frame in step 313;
• de-emphasis of the synthesized signal using the de-emphasis unit 206, and updating, in step 313, the memory of the de-emphasis unit 206;
• optionally, post-processing 207 of the synthesized signal by updating, in step 313, the memory of the post-processing - it may be noted that the post-processing can be deactivated during the frame loss correction because the information they use are unreliable because simply extrapolated, in this case the memories of the post-processing must still be updated to allow normal operation to the next received frame;
• resampling the synthesized signal at the output frequency by resampling 208, updating the memory of the filter 208 in step 313.

The updating of the internal memories allows the decoding of a possible future frame coded by CELP prediction without discontinuity. Note that in ITU-T G.718, recovery and synthesis energy control techniques are also employed (for example in clauses 7.11.1.8 and 7.11.1.8.1) when decoding a received frame after a loss of frame correction. This aspect is not considered here because it is outside the scope of the invention.

At a step 314, the memories thus updated can be copied into the temporary memory 107. The decoded replacement CELP frame is transmitted at the output of the decoder at a step 315.

• estimation par interpolation des paramètres LSF et du filtre LPC en fonction des paramètres LSF de la trame CELP précédente, sans mettre à jour les mémoires des quantificateurs LSF. L'estimation par interpolation peut être mise en oeuvre selon la même méthode que celle employée pour l'estimation par interpolation pour la trame de remplacement décrite ci-avant (sans mise à jour des mémoires des quantificateurs LSF) ;
• estimation de l'excitation à l'aide du gain adaptatif et du gain fixe de la trame CELP précédente, sans mettre à jour ces valeurs pour la trame suivante. L'excitation peut être déterminée selon la même méthode que celle employée pour la détermination de l'excitation pour la trame de remplacement (sans la mise à jour des valeurs de gain adaptatif et de gain fixe) ;
• synthétiser un segment de signal (une demi-trame ou une sous-trame par exemple) à partir de l'excitation et du filtre de synthèse 205 recalculé et en utilisant la mémoire de synthèse de la trame précédente ;
• dé-emphase du signal synthétisé en utilisant l'unité de dé-emphase 206;
• de façon optionnelle, post-traitement du signal synthétisé en utilisant la mémoire du post-traitement 207 ;
• ré-échantillonnage du signal synthétisé à la fréquence de sortie par le ré-échantillonnage 208, en utilisant les mémoires de ré-échantillonnage 208.

In a step 316, the method according to the invention provides the prediction generation of an additional segment of digital signal, by implementing a PLC algorithm adapted to the CELP. Step 316 may include the following substeps:

• Interpolation estimation of LSF parameters and LPC filter according to LSF parameters of the previous CELP frame without updating the LSF quantizer memories. The interpolation estimation can be implemented according to the same method as that used for the interpolation estimation for the replacement frame described above (without updating the memories of the LSF quantizers);
• estimation of the excitation using the adaptive gain and the fixed gain of the previous CELP frame, without updating these values for the next frame. The excitation can be determined by the same method as that used to determine the excitation for the replacement frame (without the update of the adaptive gain and fixed gain values);
• synthesizing a signal segment (for example a half-frame or a sub-frame) from the excitation and the recalculated synthesis filter 205 and using the synthesis memory of the previous frame;
• de-emphasis of the synthesized signal using the de-emphasis unit 206;
• optionally, post-processing the synthesized signal using the post-processing memory 207;
• resampling of the synthesized signal at the output frequency by resampling 208, using the resampling memories 208.

It is important to note that for each of the steps, the invention provides for storing in temporary variables the states of the CELP decoding that are modified at each step, before performing these steps, so that the predetermined states can be restored. their stored values after generating the temporary segment.

The additional signal segment generated is stored in the memory 107 at a step 317.

At a step 318, a next digital signal frame is received by the reception unit 101. It is verified in a step 319 that the next frame is an MDCT frame or a transition frame.

If this is not the case, then the next frame is a CELP frame and is decoded by the CELP decoder 103 at a step 320. The additional segment synthesized in step 316 is not used and can be deleted. the memory 107.

In the case where the next frame is an MDCT frame or a transition frame, it is decoded by the MDCT decoder 104 in a step 322. In parallel, the additional digital signal segment stored in the memory 107 is retrieved at a step 323 by the management unit 108 and transmitted to the decoding unit 105.

In the case where the following frame is an MDCT frame, the additional signal segment obtained makes it possible to carry out an overlay by the unit 103 in order to correctly decode the first part of the next MDCT frame at a step 324. for example, when the additional segment is a sub-frame half, a linear gain between 0 and 1 may be applied during the overlap over the first half of the MDCT frame and a gain linear between 1 and 0 is applied to the additional signal segment. Without this additional signal segment, MDCT decoding can give rise to discontinuities due to quantization errors.

• si la trame CELP précédente a été codée par un coeur à une première fréquence (par exemple 12,8 kHz) et que la sous-trame CELP de transition a été codée par un coeur à une deuxième fréquence (par exemple 16 kHz), alors la sous-trame de transition ne peut être décodée, et le segment de signal supplémentaire permet alors à l'unité de décodage 105 d'assurer l'addition recouvrement avec le signal issu du décodage MDCT de l'étape 322. Par exemple, lorsque le segment supplémentaire est une moitié de sous-trame, un gain linéaire entre 0 et 1 peut être appliqué lors de l'addition recouvrement sur la première moitié de la trame MDCT et un gain linéaire entre 1 et 0 est appliqué sur le segment de signal supplémentaire. ;
• si la trame CELP précédente et la sous-trame CELP de transition ont été codées par un coeur à la même fréquence, alors la sous-trame CELP de transition peut être décodée et utilisée par l'unité de décodage 105 pour l'addition-recouvrement avec le signal numérique issu du décodeur MDCT 104 ayant décodé la trame de transition.

In the case where the following frame is a transition frame, two cases are to be distinguished as considered below. It is recalled that the decoding of the transition frame is based not only on the classification of the current frame as a "transition frame" but also an indication of the CELP coding type (12.8 or 16 kHz) when several coding frequencies CELP are possible. So :

• if the preceding CELP frame has been coded by a core at a first frequency (for example 12.8 kHz) and the transition CELP subframe has been coded by a core at a second frequency (for example 16 kHz), then the transition subframe can not be decoded, and the additional signal segment then allows the decoding unit 105 to provide overlap with the signal from the MDCT decoding of step 322. For example, when the additional segment is a sub-frame half, a linear gain between 0 and 1 can be applied during overlap over the first half of the MDCT frame and a linear gain between 1 and 0 is applied on the signal segment additional. ;
• if the previous CELP frame and the transition CELP subframe have been encoded by a core at the same frequency, then the transition CELP subframe can be decoded and used by the decoding unit 105 for overlay with the digital signal from the decoder MDCT 104 having decoded the transition frame.

avec :

• r un coefficient représentatif de la longueur du segment supplémentaire généré, la longueur étant égale à L/r. Aucune restriction n'est attachée à la valeur r, qui sera choisie de manière à permettre un recouvrement suffisant entre le segment de signal supplémentaire et la trame MDCT de transition décodée. Par exemple, r peut être égal à 2 ;
• i un instant correspondant à un échantillon de la trame suivante, compris entre 0 et L/r;
• L la longueur de la trame suivante (par exemple 20 ms) ;
• S(i) l'amplitude de la trame suivante après addition, pour l'échantillon i ;
• B(i) l'amplitude du segment décodé par transformée, pour l'échantillon i ;
• T(i) l'amplitude du segment supplémentaire de signal numérique, pour l'échantillon i.

The overlap addition between the additional signal segment and the decoded MDCT frame can be given by the following formula: $$S\left(i\right)=B\left(i\right).\frac{i}{\left(\mathrm{The}/r\right)}+\left(1-\frac{i}{\left(\mathrm{The}/r\right)}\right).T\left(i\right)$$

with:

• r a coefficient representing the length of the additional segment generated, the length being equal to L / r. No restriction is attached to the value r, which will be chosen to allow sufficient overlap between the additional signal segment and the decoded transition MDCT frame. For example, r may be 2;
• i an instant corresponding to a sample of the next frame, between 0 and L / r;
• L the length of the next frame (for example 20 ms);
• S (i) the amplitude of the next frame after addition, for sample i;
• B (i) the amplitude of the transform decoded segment, for sample i;
• T (i) the amplitude of the additional digital signal segment, for sample i.

The digital signal obtained after overlap is transmitted at the output of the decoder at a step 325.

Thus, the invention provides, on loss of a current frame following a previous CELP frame, the generation of an additional segment in addition to a replacement frame. In some cases, and especially if the next frame is a CELP frame, such an additional segment is not used. However, its calculation does not induce any additional complexity insofar as the coding parameters of the previous frame are reused. On the other hand, when the next frame is an MDCT frame or a transition frame with a CELP subframe at a core frequency different from the core frequency used for encoding the previous CELP frame, the additional signal segment generated and stored allows the decoding of the next frame, which was not allowed by the solutions of the prior art.

The figure 4 represents an example of a computing device 400 that can be integrated in the CELP encoder 103 and in the MDCT encoder 104.

The device 400 comprises a random access memory 404 and a processor 403 for storing instructions allowing the implementation of steps of the method described above (implemented by the CELP encoder 103 or by the MDCT encoder 104). The device also comprises a mass memory 405 for storing data intended to be stored after the application of the method. The device 400 further comprises an input interface 401 and an output interface 406 respectively for receiving the frames of the digital signal and transmitting the decoded signal frames.

The device 400 may further comprise a digital signal processor (DSP) 402. This DSP 402 receives the digital signal frames for shaping, demodulating and amplifying, in a manner known per se, these frames.

The present invention is not limited to the embodiments described above as examples; it extends to other variants.

Thus, an embodiment has been described above in which the decoder is a separate entity. Of course, a set-top box can be embedded in any type of larger device such as a mobile phone, a computer, etc.

In addition, an embodiment has been described proposing a particular architecture of the decoder. These architectures are given for illustrative purposes only. Thus, an arrangement of the components and a different distribution of the spots assigned to each of these components is also conceivable.

Claims (11)

1. Method for decoding a digital signal coded using predictive coding and using transform coding, comprising the following steps:

   - predictive decoding (304) of a preceding frame of the digital signal, which preceding frame is coded by a set of predictive coding parameters;

   - detection (302) of the loss of a current frame of the coded digital signal;

   - predictive generation (312), on the basis of at least one predictive coding parameter coding the preceding frame, of a replacement frame for the current frame;

   - predictive generation (316), on the basis of at least one predictive coding parameter coding the preceding frame, of an additional digital signal segment;

   - temporary storage (317) of said additional digital signal segment.
2. Method according to Claim 1, furthermore comprising the following steps:

   - reception (318) of a following coded digital signal frame comprising at least one transform-coded segment; and

   - decoding (322; 323; 324) of the following frame, comprising a sub-step of addition with overlap between the additional digital signal segment and said transform-coded segment.
3. Method according to Claim 2, wherein the following frame is coded entirely using transform coding, and wherein the lost current frame is a transition frame between the preceding frame coded using predictive coding and the following frame coded using transform coding.
4. Method according to Claim 2, wherein the preceding frame is coded using predictive coding by a predictive coder core operating at a first frequency, and
   wherein the following frame is a transition frame comprising at least one subframe coded using predictive coding by a predictive coder core operating at a second frequency different from the first frequency.
5. Method according to Claim 4, wherein the following frame comprises a bit indicating the frequency of the predictive coder core that is used.
6. Method according to one of Claims 2 to 5, wherein the addition with overlap is given by applying the following formula: $$S\left(i\right)=B\left(i\right).\frac{i}{\left(L/r\right)}+\left(1-\frac{i}{\left(L/r\right)}\right).T\left(i\right)$$

   

   where:

   - r is a coefficient representative of the length of the generated additional segment;

   - i is an instant corresponding to a sample of the following frame, of between 0 and L/r;

   - L is the length of the following frame;

   - S(i) is the amplitude of the following frame after addition, for the sample i;

   - B(i) is the amplitude of the transform-decoded segment, for the sample i;

   - T(i) is the amplitude of the additional digital signal segment, for the sample i.
7. Method according to one of the preceding claims, wherein the step of predictive generation of the replacement frame furthermore includes updating (313) of internal memories of the decoder, and
   wherein the step of predictive generation of an additional digital signal segment includes the following sub-steps:

   - copying (314), into a temporary memory (107), the memories of the decoder that are updated during the step of predictive generation of the replacement frame;

   - generation (316) of the additional digital signal segment by way of the temporary memory.
8. Method according to one of the preceding claims, wherein the step of predictive generation of an additional digital signal segment includes the following sub-steps:

   - predictive generation of an additional frame, on the basis of at least one predictive coding parameter coding the preceding frame;

   - extraction of a segment of the additional frame; and

   wherein the additional digital signal segment corresponds to the first half of the additional frame.
9. Computer program including instructions for implementing the method according to any one of the preceding claims when these instructions are executed by a processor.
10. Decoder for decoding a digital signal coded using predictive coding and using transform coding, comprising:

    - a unit (108) for detecting the loss of a current frame of the digital signal;

    - a predictive decoder (103) including a processor designed to perform the following operations:

    * predictive decoding of a preceding frame of the digital signal, which preceding frame is coded by a set of predictive coding parameters;

    * predictive generation, on the basis of at least one predictive coding parameter coding the preceding frame, of a replacement frame for the current frame;

    * predictive generation, on the basis of at least one predictive coding parameter coding the preceding frame, of an additional digital signal segment;

    * temporary storage of said additional digital signal segment in a temporary memory (107).
11. Decoder according to Claim 10, furthermore including a transform decoder (104) including a processor designed to perform the following operations:

    * reception of a following coded digital signal frame comprising at least one transform-coded segment; and

    * transform-decoding of the following frame;

    said decoder furthermore comprising a decoding unit (105) comprising a processor designed to perform an addition with overlap between the additional digital signal segment and said transform-coded segment.

Images