EP2489039B1 - Optimized low-throughput parametric coding/decoding - Google Patents

Description

The present invention relates to the field of coding / decoding of digital signals.

The coding and decoding according to the invention is particularly suitable for the transmission and / or storage of digital signals such as audio-frequency signals (speech, music or other).

More particularly, the present invention relates to the parametric encoding / decoding of multichannel audio signals.

This type of coding / decoding is based on the extraction of spatial information parameters so that at decoding, these spatial characteristics can be reconstructed for the listener.

This type of parametric encoding applies in particular for a stereo signal. Such a coding / decoding technique is for example described in the document Breebaart, J. and van de Par, S and Kohlrausch, A. and Schuijers, titled "Parametric Coding of Stereo Audio" in EURASIP Journal on Applied Signal Processing 2005: 9, 1305-1322 . This example is repeated with reference to figures 1 and 2 describing respectively an encoder and a parametric stereo decoder.

So, the figure 1 describes an encoder receiving two audio channels, a left channel (denoted L for Left in English) and a right channel (denoted R for Right in English).

The channels L (n) and R (n) are processed by the blocks 101, 102 and 103, 104 respectively which perform a short-term Fourier analysis. The transformed signals L [j] and R [j] are thus obtained.

The block 105 performs a channel reduction matrix or "Downmix" in English to obtain from the left and right signals, a sum signal, a mono signal in this case, in the frequency domain.

An extraction of spatial information parameters is also performed in block 105.

The ICLD ( InterChannel Level Difference ) type parameters, also called interchannel intensity differences, characterize the energy ratios per frequency subband between the left and right channels.

où L[j] et R[j] correspondent aux coefficients spectraux (complexes) des canaux L et R, les valeurs B[k] et B[k+1], pour chaque bande de fréquence k, définissent la découpe en sous-bande du spectre et le symbole * indique le conjugué complexe.They are defined in dB by the following formula: $$\mathrm{ICLD}\left[k\right]=10.{\log }_{10}\left(\frac{\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}\mathrm{The}\left[j\right]\cdot {\mathrm{The}}^{*}\left[j\right]}{\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}R\left[j\right]\cdot R^{*}\left[j\right]}\right)\mathrm{dB}$$

where L [j] and R [j] correspond to the spectral (complex) coefficients of the L and R channels, the values B [k] and B [k + 1] , for each frequency band k, define the sub-division. spectrum band and the symbol * indicates the complex conjugate.

où ∠ indique l'argument (la phase) de l'opérande complexe.A parameter of ICPD type (for " InterChannel Phase Difference " in English) also called phase difference by frequency subband, is defined according to the following relation: $$\mathrm{ICLD}\left[k\right]=\angle \left(\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}\mathrm{The}\left[j\right]\cdot R^{*}\left[j\right]\right)$$

where ∠ indicates the argument (phase) of the complex operand.

An interchannel time lag called ICTD (for "interchannel time difference") can also be defined equivalent to ICPD.

An interchannel coherence parameter ICC (for " InterChannel Coherence " in English) represents inter-channel correlation.

These ICLD, ICPD and ICC parameters are extracted from the stereo signals, by block 105.

The mono signal is passed in the time domain (blocks 106 to 108) after short-term Fourier synthesis (inverse FFT, windowing and OverLap-Add or OLA) and a mono coding (block 109) is realized. . In parallel, the stereo parameters are quantized and coded in block 110.

In general, the spectrum of the signals ( L [ j ], R [ j ]) is divided according to a nonlinear frequency scale of ERB ( equivalent Rectangular Bandwidth ) or Bark type, with a number of subbands typically ranging from 20 to 34. This scale defines the values of B (k) and B (k + 1) for each subband k. The settings (ICLD, ICPD, ICC) are encoded by scalar quantization possibly followed by entropy coding or differential coding. For example, in the aforementioned article, the ICLD is encoded by a non-uniform quantizer (ranging from -50 to +50 dB) with differential coding; the non-uniform quantization step exploits the fact that the higher the value of the ICLD, the lower the auditory sensitivity to variations of this parameter.

At the decoder 200, the mono signal is decoded (block 201), a de-correlator is used (block 202) to produce two versions M (n) and M '(n) of the decoded mono signal. These two signals passed in the frequency domain (blocks 203 to 206) and the decoded stereo parameters (block 207) are used by the stereo synthesis (block 208) to reconstruct the left and right channels in the frequency domain. These channels are finally reconstructed in the time domain (blocks 209 to 214).

As an alternative example, the document WO 2006/108464 A1 describes a technique for transmitting spatial information parameters of similar type based on a prior grouping of said parameters for two consecutive frames in time and in frequency. The group of parameters requiring the lowest coding rate is chosen to be transmitted to the decoder.

In stereo signal coding techniques, a stereo intensity coding technique ( Intensity Stereo Coding ) consists of coding the sum (M) channel as well as the ICLD energy ratios as defined above.

Stereo intensity coding exploits the fact that the perception of high frequency components is mainly related to the temporal (energy) envelopes of the signal.

For mono signals, there are also quantization techniques with or without memory such as "Coded Pulse Modulation" (MIC) coding or its adaptive version called "Adaptive Differential Coded Pulse Modulation" (ADPCM).

Of particular interest here is ITU-T Recommendation G.722, which uses ADPCM for Adaptive Differential Pulse Code Modulation (ADPCM).

The input signal of a G.722-type encoder is in an expanded band with a minimum bandwidth of [50-7000 Hz] with a sampling frequency of 16 kHz. This signal is decomposed into two sub-bands [0-4000 Hz] and [4000-8000 Hz] obtained by decomposition of the signal by so-called quadrature mirror filters. Quadrature Mirror Filters (QMF) in English, then each of the subbands is separately encoded by an ADPCM encoder.

The low band is coded by a 6, 5 and 4 bit nested code ADPCM coding while the high band is coded by a 2 bit ADPCM coder per sample. The total bit rate is 64, 56 or 48 bit / s depending on the number of bits used for decoding the low band.

Recommendation G.722 was first used in ISDN (Integrated Services Digital Network) and then in enhanced IP voice telephony applications in HD (High Definition) or HD voice in English.

A quantized signal frame according to the G.722 standard consists of 6, 5 or 4 bit low band (0-4000 Hz) and 2 high band (4000-8000 Hz) coded quantization indices. Since the transmission frequency of the scalar indices is 8 kHz in each subband, the bit rate is 64, 56 or 48 kbit / s. In the G.722 standard, the 8 bits are distributed as follows: 2 bits for the high band, 6 bits for the low band. The last or last two bits of the low band can be "stolen" or replaced by data.

• Une extension de la bande acoustique de 50-7000 Hz (bande élargie) à 50-14000 Hz (bande super-élargie). En anglais la bande super-élargie est appelée Superwideband (SWB).
• Une extension de mono à stéréo. Cette extension stéréo peut étendre un codage mono en bande élargie ou un codage mono en bande super-élargie.

ITU-T has recently launched a standardization activity called G.722-SWB (as part of the Q.10 / 16 question described for example in the document: ITU-T: Annex Q10.J ITU-T G.722 and ITU-T G.711 WB, January 2009, WD04_G722G711SWBToRr3.doc) and extending the G.722 Recommendation in two ways:

• An extension of the acoustic band from 50-7000 Hz (wide band) to 50-14000 Hz (super-wide band). In English the super-enlarged band is called Superwideband (SWB).
• An extension of mono to stereo. This stereo extension can extend wide-band mono coding or super-wide band mono coding.

In the context of G.722-SWB, G.722 coding operates with short 5 ms frames.

We are particularly interested here in the stereo extension of the G.722 broadband coding

• Une extension stéréo de G.722 à 56 kbit/s avec un débit additionnel de 8 kbit/s, soit 64 kbit/s au total
• Une extension de G.722 à 64 kbit/s avec un débit additionnel de 16 kbit/s, soit 80 kbit/s au total

Two G.722 stereo extension modes are to be tested in the G.722-SWB standardization:

• A G.722 stereo extension at 56 kbit / s with an additional bit rate of 8 kbit / s, or 64 kbit / s in total
• A 64 kbit / s G.722 extension with an additional bit rate of 16 kbit / s, or 80 kbit / s in total

The spatial information represented by the ICLD or other parameters requires a bit rate (additional stereo extension) all the more important as the coding frames are short.

By way of example, in the context of G.722-SWB normalization, assuming that a stereo extension of G.722 (wide band) is performed by the intensity coding technique, we obtain the next stereo expansion rate.

For a G.722 (mono) sum signal with a frame of 5 ms and a division of the broadband spectrum (0-8000 Hz) in 20 sub-bands, 20 ICLD parameters to be transmitted every 5 ms are obtained. It can be assumed that these ICLD parameters are coded with a bit rate (average) of the order of 4 bits per subband. The stereo extension rate of G.722 thus becomes 20 x 4 bits / 5 ms = 16 kbit / s. Thus the stereo extension of G.722 by ICLD with 20 sub-bands leads to an additional bit rate of the order of 16 kbit / s. However, according to the state of the art, the coding of the ICLD alone is not generally sufficient to achieve good stereo quality.

This example thus illustrates the difficulty of performing a stereo extension of an encoder such as G.722 with short frames (of 5 ms).

Direct encoding of the ICLD (without other parameters) gives an additional bit rate (stereo extension) around 16 kbit / s which is already the maximum possible bit rate for the G.722 extension.

There is therefore a need to represent a stereo signal or more generally multichannel, effectively, at a rate as low as possible, with acceptable quality, when coding frames are short.

The present invention improves the situation.

For this purpose, it proposes in one embodiment, a parametric encoding method of a multichannel digital audio signal as in claim 1.

The invention also applies to a parametric decoding method of a multichannel digital audio signal as in claim 5.

The invention also relates to an encoder as in claim 8.

The invention also relates to a decoder as in claim 9.

It also relates to a computer program comprising code instructions for implementing the steps of the encoding method as described and to a computer program comprising code instructions for implementing the steps of a decoding method. as described, when these are executed by a processor.

• la figure 1 illustre un codeur mettant en oeuvre un codage paramétrique connu de l'état de l'art et précédemment décrit;
• la figure 2 illustre un décodeur mettant en oeuvre un décodage paramétrique connu de l'état de l'art et précédemment décrit;
• la figure 3 illustre un codeur selon un mode de réalisation de l'invention, mettant en oeuvre un procédé de codage selon un mode de réalisation de l'invention;
• la figure 4 illustre un décodeur selon un mode de réalisation de l'invention, mettant en oeuvre un procédé de décodage selon un mode de réalisation de l'invention;
• la figure 5 illustre la division d'un signal audio numérique en trames dans un codeur mettant en oeuvre un procédé de codage selon un mode de réalisation de l'invention;
• la figure 6 illustre un procédé de codage et un codeur selon un autre mode de réalisation de l'invention; et
• les figures 7a et 7b illustrent respectivement un dispositif apte à mettre en oeuvre le procédé de codage et le procédé de décodage selon un mode de réalisation de l'invention.

Other features and advantages of the invention will appear more clearly on reading the following description, given solely by way of nonlimiting example, and with reference to the appended drawings, in which:

• the figure 1 illustrates an encoder implementing a parametric coding known from the state of the art and previously described;
• the figure 2 illustrates a decoder implementing a parametric decoding known from the state of the art and previously described;
• the figure 3 illustrates an encoder according to one embodiment of the invention, implementing a coding method according to one embodiment of the invention;
• the figure 4 illustrates a decoder according to one embodiment of the invention, implementing a decoding method according to one embodiment of the invention;
• the figure 5 illustrates the division of a digital audio signal into frames in an encoder implementing a coding method according to an embodiment of the invention;
• the figure 6 illustrates a coding method and an encoder according to another embodiment of the invention; and
• the Figures 7a and 7b respectively illustrate a device adapted to implement the coding method and the decoding method according to one embodiment of the invention.

With reference to the figure 3 , a first embodiment of a stereo signal encoder implementing a coding method according to a first embodiment is now described.

This parametric stereo encoder operates in wideband with stereo signals sampled at 16 kHz with 5 ms frames. Each channel (L and R) is first pre-filtered by a high pass filter (HPF for High Pass Filter English) removing components below 50 Hz (blocks 301 and 302). Then a mono signal (M) is calculated by the block 303, an exemplary embodiment of which is given in the form: $$M\left(\mathrm{not}\right)=\frac{1}{2}\left(\mathrm{The}\text{'}\left(\mathrm{not}\right)+R\text{'}\left(\mathrm{not}\right)\right)$$

This signal is encoded (block 304) by a G.722 type encoder, as described, for example, in ITU-T Recommendation G.722, 7 kHz audio-coding within 64 kbit / s , Nov. 1988.

The delay introduced in the G.722 type coding is 22 samples at 16 kHz. The L and R channels are aligned in time (blocks 305 and 308) with a delay of T = 22 samples and analyzed in frequency per transform, for example by discrete Fourier transform with sinusoidal overlapping windowing. the example here is 50% (blocks 306, 307 and 309, 310). Each window thus covers 2 frames of 5 ms or 10 ms (160 samples).

The division of the signal into frames is defined with reference to the figure 5 . This figure illustrates the fact that the analysis window (solid line) of 10 ms covers the current frame of index t and the future frame of index t + 1 and the fact that a recovery of 50% is used between the window of the current frame and the window (dotted line) of the previous frame.

Taking into account the future frame induces an additional algorithmic delay of 5 ms to the encoder.

For the frame t, the spectra obtained, L [ t, j ] and R [ t, j ] ( j = 0 ... 79), at the output of the blocks 307 and 310 of the figure 3 , include 80 complex samples, with a resolution of 100 Hz per frequency band.

The block 311 for extracting spatial information parameters is now detailed.

• {B(k)} _k=0,..,20 = [0, 1, 2, 3, 4, 5, 6, 7, 9, 11, 13, 16, 19, 23, 27, 31, 37, 44, 52, 61, 80]

This comprises, in the case of the processing in the frequency domain, a first module 313 for splitting the spectra L [ t, j ] and R [ t, j ] into a predetermined number of frequency subbands, for example here in 20 subbands according to the scale defined below:

• { B (k) } _{k = 0, .., 20} = [0, 1, 2, 3, 4, 5, 6, 7, 9, 11, 13, 16, 19, 23, 27, 31, 37, 44, 52, 61, 80]

This scale delimits (in number of Fourier coefficients) the frequency subbands of index k = 0 to 19. For example the first sub-band ( k = 0) goes from the coefficient B (k) = 0 to B (k + 1) -1 = 0; it is therefore reduced to a single coefficient (100 Hz).

Similarly, the last subband ( k = 19) goes from the coefficient B (k) = 61 to B (k + 1) - 1 = 79, it comprises 19 coefficients (1900 Hz).

The module 314 comprises means for obtaining the spatial information parameters of the stereo signal.

For example, the parameters obtained are the interchannel intensity difference parameters, ICLD.

où ${\sigma }_L^2\left[t,k\right]$

et ${\sigma }_R^2\left[t,k\right]$

représentent respectivement l'énergie du canal gauche (L) et du canal droit (R).For each frame of index t, the ICLD of the sub-band k = 0 , ..., 19 is calculated according to the equation: $$\mathrm{ICLD}\left[k\right]=10.{\log }_{10}\left(\frac{{\sigma }_{\mathrm{The}}^2\left[t,k\right]}{{\sigma }_R^2\left[t,k\right]}\right)\mathrm{dB}$$

or ${\sigma }_{\mathrm{The}}^2\left[t,k\right]$

and ${\sigma }_R^2\left[t,k\right]$

represent respectively the energy of the left channel (L) and the right channel (R).

In a particular embodiment, these energies are calculated as follows: $$\{\begin{array}{c}{\sigma }_{\mathrm{The}}^2\left[t,k\right]=\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}\mathrm{The}\left[t,j\right]\cdot {\mathrm{The}}^{*}\left[t,j\right]+\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}\mathrm{The}\left[t-1,j\right]\cdot {\mathrm{The}}^{*}\left[t-1,j\right]\\ {\sigma }_R^2\left[t,k\right]=\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}R\left[t,j\right]\cdot R^{*}\left[t,j\right]+\underset{j=B\left[k\right]}{\overset{B\left[k+1\right]-1}{\Sigma }}R\left[t-1,j\right]\cdot R^{*}\left[t-1,j\right]\end{array}$$

This formula amounts to combining the energy of two successive frames, which corresponds to a temporal support of 10 ms (15 ms if we count the effective temporal support of two successive windows).

The module 314 therefore produces a series of ICLD parameters defined previously.

These ICLD parameters are divided into the division module 315, into several blocks. In the embodiment illustrated here, the parameters are divided into two blocks according to the following two parts: {ICLD [ t, k ]} _{k = 0, ..., 9} and {ICLD [ t , k ]} _{k = 10 , ..., 19} .

The division of ICLD parameters into contiguous blocks makes it possible to perform a differential coding of the scalar quantization indices.

The module 316 then makes a selection (St.) of a block to be encoded according to the index of the current frame to be coded.

In the example described herein, for the frames t even index, the block ICLO {[t, k]} _{k = 0, ..., 9} is encoded at 312 and transmitted to the frame t odd index the block {ICLD [ t, k ]} _{k = 10, ..., 19} is coded at 312 and transmitted.

The coding of these blocks at 312 is carried out for example by non-uniform scalar quantization.

• 5 bits pour le premier paramètre ICLD,
• 4 bits pour les 8 paramètres ICLD suivants,
• 3 bits pour le dernier (dixième) paramètre ICLD.

Thus, the coding of an ICLD block is achieved with:

• 5 bits for the first ICLD parameter,
• 4 bits for the following 8 ICLD parameters,
• 3 bits for the last (tenth) ICLD parameter.

la quantification à 5 bits de ICLD[t,k] consiste à trouver l'indice de quantification i tel que $$\mathrm{i}=\mathrm{arg\; min\_j}=\mathrm{0}\dots \mathrm{30}\left|\mathrm{ICLD}\left[\mathrm{t},\mathrm{k}\right]-\mathrm{tab\_ild\_q}\mathrm{5}\left[\mathrm{j}\right]\right|{}^{\wedge }\mathrm{2}$$

A more detailed example of realization is for example as below: For the quantization table: $$\mathrm{tab\_ild\_q}5\left[31\right]=\left\{\begin{array}{cccccccccccccccccccccccccccccccc}-50,& -45,& -40,& -35,& -30,& -25,& -22,& -19,& -16,& -13,& -10,& -8,& -6,& -4,& -2,& 0,& 2,& 4,& 6,& 8,& 10,& 13,& 16,& 19,& 22,& 25,& 30,& 35,& 40,& 45,& 50& \end{array}\right\}$$

the 5-bit quantization of ICLD [t, k] consists in finding the quantization index i such that $$\mathrm{i}=\mathrm{arg\; min\_j}=\mathrm{0}...\mathrm{30}\left|\mathrm{ICLD}\left[\mathrm{t},\mathrm{k}\right]-\mathrm{tab\_ild\_q}\mathrm{5}\left[\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="imgf0004.png"\; state="\{\{state\}\}">j</figure-callout>}\right]\right|{}^{\wedge }\mathrm{2}$$

la quantification à 4 bits de ICLD[t,k] consiste à trouver l'indice de quantification i tel que $$\mathrm{i}=\mathrm{arg\; min\_j}=\mathrm{0}\dots \mathrm{15}\left|\mathrm{ICLD}\left[\mathrm{t},\mathrm{k}\right]-\mathrm{tab\_ild\_q}\mathrm{4}\left[\mathrm{j}\right]\right|{}^{\wedge }\mathrm{2}$$

Likewise for the quantization table: $$\mathrm{tab\_ild\_q}4\left[15\right]=\left\{\begin{array}{ccccccccccccccc}-16,& -13,& -10,& -8,& -6,& -4,& -2,& 0,& 2,& 4,& 6,& 8,& 10,& 13,& 16\end{array}\right\}$$

the 4-bit quantization of ICLD [t, k] consists of finding the quantization index i such that $$\mathrm{i}=\mathrm{arg\; min\_j}=\mathrm{0}...\mathrm{15}\left|\mathrm{ICLD}\left[\mathrm{t},\mathrm{k}\right]-\mathrm{tab\_ild\_q}\mathrm{4}\left[\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="imgf0004.png"\; state="\{\{state\}\}">j</figure-callout>}\right]\right|{}^{\wedge }\mathrm{2}$$

Finally for the quantization table tab_ild_q3 [7] = {-16, -8, -4, 0,4, 8, 16} the 3-bit quantization of ICLD [t, k] consists of finding the quantization index i such as $$\mathrm{i}=\mathrm{arg\; min\_j}=\mathrm{0}...\mathrm{15}\left|\mathrm{ICLD}\left[\mathrm{t},\mathrm{k}\right]-\mathrm{tab\_ild\_q}\mathrm{3}\left[\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="imgf0004.png"\; state="\{\{state\}\}">j</figure-callout>}\right]\right|{}^{\wedge }\mathrm{2}$$

In total 5 + 8x4 +3 = 40 bits are therefore necessary for the coding of an ICLD block. The frame being 5 ms, we thus obtain 40 bits / 5 ms = 8 kbit / s as additional bit rate for the stereo coding extension.

This bit rate is therefore not too great and is sufficient to efficiently transmit the stereo parameters.

Two successive frames suffice in this embodiment to obtain the spatial information parameters of the multichannel signal, the length of two frames being most often the length of an analysis window for a 50% overlap frequency transformation. .

Alternatively, a shorter recovery window could be used to reduce the delay introduced.

• obtention (Obt.), par trame de longueur prédéterminée, de paramètres d'information spatiale du signal multicanal;
• division (Div.) des paramètres d'information spatiale en une pluralité de blocs de paramètres;
• sélection (St.) d'un bloc de paramètres en fonction de l'indice de la trame courante;
• codage (Q) du bloc de paramètres sélectionné pour la trame courante.

So, the coder described with reference to the figure 3 implements a parametric encoding method of a multichannel digital audio signal comprising a coding step (G.722 Cod) of a signal resulting from a channel reduction matrix for the multichannel signal. The method further comprises the following steps:

• obtaining (Obt.), per frame of predetermined length, spatial information parameters of the multichannel signal;
• dividing (Div) spatial information parameters into a plurality of parameter blocks;
• selecting (St.) a parameter block according to the index of the current frame;
• encoding (Q) of the selected parameter block for the current frame.

In the embodiment described above, it was in the context of an expanded band encoder operating with a sampling frequency of 16 kHz and a particular subband cut.

In another possible embodiment, the encoder may operate at other frequencies (such as 32 kHz) and with different subband cutting.

• pour les trames d'indice t pair: codage d'un bloc de 9 paramètres {ICLD [t,k]} _k=1,...,9 par quantification scalaire non-uniforme avec:
  • 5 bits pour le premier paramètre ICLD[t,k] avec k=1
  • 4 bits pour les 8 paramètres ICLD suivants
• pour les trames d'indice t impair: codage d'un bloc de 10 paramètres {ICLD[t,k]} _k=10,...,19 comme présenté précédemment
  • 5 bits pour le premier paramètre ICLD,
  • 4 bits pour les 8 paramètres ICLD suivants,
  • 3 bits pour le dernier (dixième) paramètre ICLD.

One can also exploit the fact that the parameter ICLD [ t , k ] for k = 0 can be neglected. Its calculation and therefore its coding can be avoided. In this case the coding of the ICLD parameters becomes:

• for frames of even t- index: coding of a block of 9 parameters {ICLD [ t , k ]} _{k = 1, ..., 9} by non-uniform scalar quantization with:
  • 5 bits for the first parameter ICLD [ t , k ] with k = 1
  • 4 bits for the following 8 ICLD parameters
• for frames of odd t- index: coding of a block of 10 parameters {ICLD [ t , k ]} _{k = 10, ..., 19} as previously presented
  • 5 bits for the first ICLD parameter,
  • 4 bits for the following 8 ICLD parameters,
  • 3 bits for the last (tenth) ICLD parameter.

Thus, in this embodiment, 37 bits are used for frames of even t- index and 40 bits for frames of odd t- indexes.

Similarly, in an alternative embodiment, instead of dividing the ICLD parameters into contiguous blocks, these parameters can be divided differently, for example into interleaving to obtain 2 parts: {ICLD [ t , 2 k ]} _{k = 0 ,. .., 9} and ICLD [ t , 2 k +1]} _{k = 0, ..., 9} .

It should be noted that the coding method thus described is easily generalized in the case where the parameters are divided into more than 2 blocks. In an alternative embodiment, the ICLD parameters are divided into 4 blocks: $$\begin{array}{ccccc}{\left\{\mathrm{ICLD}\left[t,k\right]\right\}}_{k=0,...,4},& {\left\{\mathrm{ICLD}\left[t,k\right]\right\}}_{k=5,...,9},& {\left\{\mathrm{ICLD}\left[t,k\right]\right\}}_{k=10,...,14}& \mathrm{and}& {\left\{\mathrm{ICLD}\left[t,k\right]\right\}}_{k=15,...,19}\mathrm{.}\end{array}$$

et ${\sigma }_R^2\left[t,k\right]\mathrm{.}$

The coding of the ICLD parameters is then distributed over 4 successive frames with storage of the parameters decoded in the previous frames during the decoding. The calculation of the ICLD must then be modified to include more than 2 frames in the calculation of the energies ${\sigma }_{\mathrm{The}}^2\left[t,k\right]$

and ${\sigma }_R^2\left[t,k\right]\mathrm{.}$

• 5 bits pour le premier paramètre ICLD
• 4 bits pour les 4 paramètres ICLD suivants

avec un total de 21 bits par trame. Le débit est donc plus faible encore que dans le mode de réalisation précédent, la contrepartie étant que les paramètres ICLD sont remis à jour dans au moins un bloc toutes les 20 ms au lieu de toutes les 10 ms. Pour certains paramètres stéréo et suivant le type de signal, cette variante peut cependant introduire des défauts de spatialisation audible.In this variant embodiment, the coding of the ICLD parameters can then use the following allocation:

• 5 bits for the first ICLD parameter
• 4 bits for the following 4 ICLD parameters

with a total of 21 bits per frame. The rate is therefore even lower than in the previous embodiment, the counterpart being that the ICLD parameters are updated in at least one block every 20 ms instead of every 10 ms. For some stereo parameters and depending on the type of signal, this variant may however introduce audible spatialization defects.

However, the interest of transmitting the stereo or spatial parameters at a slower rate than that of the frames is always great. The imperfect auditory perception of interchanal energy variations is exploited.

Finally, the encoding method thus described applies to the encoding of other parameters than the ICLD parameter. For example, the coherence parameter (ICC) can be calculated and transmitted selectively in a manner similar to the ICLD.

The two parameters can also be calculated and coded according to the coding method described above.

The figure 4 illustrates a decoder in one embodiment of the invention as well as the decoding method that it implements.

The portion of bit stream scalable and received from the G.722 encoder is demultiplexed and decoded by a G.722 type decoder (block 401) in 56 or 64 kbit / s mode. The synthesized signal obtained corresponds to the mono signal M ( n ) in the absence of transmission errors.

A short-term discrete Fourier transform analysis with the same windowing as the encoder is performed on M ( n ) (blocks 402 and 403) to obtain the spectrum M [ j ].

The part of the bit stream associated with the stereo extension is also demultiplexed at block 404.

The operation of the synthesis block 405 is now detailed.

For even-numbered frames t , a first parameter block {ICLD ^q [ t , k ]} _{k = 0,..., 9} is decoded in the module 404 and these decoded parameters are stored in the module 412. For the frames t of odd index is decoded in the module 404 a second block of parameters {ICLD ^q [ t, k ]} _{k = 10, ..., 19} and stored in the module 412 these decoded parameters.

A more detailed example of realization is for example as below:

For the quantization table: $$\mathrm{tab\_ild\_q}5\left[31\right]=\left\{\begin{array}{cccccccccccccccccccccccccccccccc}-50,& -45,& -40,& -35,& -30,& -25,& -22,& -19,& -16,& -13,& -10,& -8,& -6,& -4,& -2,& 0,& 2,& 4,& 6,& 8,& 10,& 13,& 16,& 19,& 22,& 25,& 30,& 35,& 40,& 45,& 50& \end{array}\right\}$$

the decoding of a 5-bit index i consists in synthesizing the ICLD parameter ^q [ t , k ] as $${\mathrm{ICLD}}^{\mathrm{q}}\left[\mathrm{t},\mathrm{k}\right]=\mathrm{tab\_ild\_q}\mathrm{5}\left(\mathrm{i}\right)$$

le décodage d'un indice i à 4 bits consiste à synthétiser le paramètre ICLD^q[t,k] comme $${\mathrm{ICLD}}^{\mathrm{q}}\left[\mathrm{t},\mathrm{k}\right]=\mathrm{tab\_ild\_q}\mathrm{4}\left(\mathrm{i}\right)$$

Likewise for the quantization table: $$\mathrm{tab\_ild\_q}4\left[15\right]=\left\{\begin{array}{ccccccccccccccc}-16,& -13,& -10,& -8,& -6,& -4,& -2,& 0,& 2,& 4,& 6,& 8,& 10,& 13,& 16\end{array}\right\}$$

the decoding of a 4 bit i index consists in synthesizing the ICLD parameter ^q [t, k] as $${\mathrm{ICLD}}^{\mathrm{q}}\left[\mathrm{t},\mathrm{k}\right]=\mathrm{tab\_ild\_q}\mathrm{4}\left(\mathrm{i}\right)$$

Finally for the quantization table tab_ild_q3 [7] = {-16, -8, -4, 0, 4, 8, 16} the decoding of a 3-bit index i is to synthesize the ICLD parameter ^q [t, k ] as $${\mathrm{ICLD}}^{\mathrm{q}}\left[\mathrm{t},\mathrm{k}\right]=\mathrm{tab\_ild\_q}\mathrm{3}\left(\mathrm{i}\right)$$

In even-numbered frames, then, in the module 413, for the missing part of the parameters, the stored values {ICLD ^q [ t -1, k]} _{k = 10, .... 19} in the previous frame are used. either: ICLD ^q [ t, k ] = ICLD ^q [ t - l, k ] for k = 10 ... 19. Similarly, in odd-numbered frames, for the missing part {ICLD ^q [ t -1, k ]} _{k = 0,..., 9} the values stored in the previous frame are used.

avec $$\{\begin{array}{c}{c}_1\left[t,k\right]=\sqrt{\frac{2c^2\left[t,k\right]}{1+c^2\left[t,k\right]}}\\ c_2\left[t,k\right]=\sqrt{\frac{2}{1+c^2\left[t,k\right]}}\end{array}$$

où $$c\left[t,k\right]={10}^{\mathit{ICLD}\left[t,k\right]/20}$$

The parameters for each of the frequency bands are thus obtained. The spectra of the left and right channels are reconstructed by the synthesis module 414 by applying the parameters {ICLD ^q [ t -1, k ]} _{k = 0,..., 19} thus decoded by sub-band. This synthesis is carried out for example as follows: $$\{\begin{array}{c}\widehat{\mathrm{The}}\left[j\right]={\mathrm{vs}}_1\left[t,k\right]\mathrm{.}\hat{M}\left[j\right]\\ \hat{R}\left[j\right]={\mathrm{vs}}_2\left[t,k\right]\mathrm{.}\hat{M}\left[j\right]\end{array},j=B\left(k\right)...\left(k+1\right)-1$$

with $$\{\begin{array}{c}{\mathrm{vs}}_1\left[t,k\right]=\sqrt{\frac{2{\mathrm{vs}}^2\left[\mathrm{<figure-callout\; id="1"\; label="t\; k"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">t\; </figure-callout>},k,\right]}{1+{\mathrm{vs}}^2\left[t,k\right]}}\\ {\mathrm{vs}}_2\left[t,k\right]=\sqrt{\frac{2}{1+{\mathrm{vs}}^2\left[t,k\right]}}\end{array}$$

or $$\mathrm{vs}\left[t,k\right]={10}^{\mathit{ICLD}\left[t,k\right]/20}$$

Note that the calculation of scale factors above is given as an example. Other means of expressing scale factors exist and can be implemented for the present invention.

The left and right channels L ( n ) and R ( n ) are reconstructed by inverse discrete Fourier transform (blocks 406 and 409) of the respective spectra L [ j ] and R [ j ] and addition-overlap (blocks 408 and 411) with sinusoidal windowing (blocks 407 and 410).

• décodage (Q^-1) de paramètres d'information spatiale reçus pour une trame courante de longueur prédéterminée de signal décodé;
• mémorisation (Mem) des paramètres décodés pour la trame courante;
• obtention (Comp.P) des paramètres décodés et mémorisés d'au moins une trame précédente et association de ces paramètres à ceux décodés pour la trame courante;
• reconstruction (Synth.) du signal multicanal à partir du signal décodé et de l'association de paramètres obtenus pour la trame courante.

So the decoder described with reference to the figure 4 , in the particular embodiment of the decoding of stereo signals, uses a parametric decoding method of a multichannel digital audio signal comprising a step of decoding (G.722 Dec) a signal resulting from a reduction matrix of multichannel signal channels. The method further comprises the following steps:

• decoding (Q ^-1 ) received spatial information parameters for a current frame of predetermined length of decoded signal;
• storing (Mem) decoded parameters for the current frame;
• obtaining (Comp.P) decoded and stored parameters of at least one previous frame and associating these parameters with those decoded for the current frame;
• reconstruction (Synth.) of the multichannel signal from the decoded signal and the combination of parameters obtained for the current frame.

In the case of a division in more than two blocks spatial information parameters, for example in 4 blocks as in an embodiment variant described above, all the decoded parameter blocks for 4 decoded frames are obtained.

The bit rate of the stereo extension is therefore reduced and obtaining these parameters makes it possible to reconstruct a stereo signal of good quality.

It may also be noted that alternative techniques to parameter coding (ICLD, ICPD, ICC) can be adopted to implement the coding method according to the invention.

Thus, in an alternative embodiment, the module 314 of the parameter extraction block of the figure 3 differs.

This module in this embodiment makes it possible to obtain other stereo parameters by applying a principal component analysis (PCA) such as that described in the article by Manuel Briand, David Virette and Nadine Martin entitled "Parametric coding of stereo audio based principal component analysis "published in the DAFX conference, 1991.

Thus, a principal component analysis is performed by subbands. The left and right channels thus analyzed are then rotated to obtain a main component and a qualified environment sub component. The stereo analysis produces, for each sub-band, a rotation angle parameter (θ) and an energy ratio between the main component and the ambient signal ( PCAR which means Principal Component to Ambience Energy Ratio).

The stereo parameters then consist of the angle of rotation parameter and the energy ratio (θ and PCAR).

The figure 6 illustrates another embodiment of an encoder according to the invention.

Compared to the coder of the figure 3 here it is the block 303 for stamping or "downmix" which differs. In the example of the figure 3 the "downmix" operation has the advantage of being instantaneous and of minimal complexity.

However, this operation does not necessarily allow conservation of energy. An improvement of this "downmix" operation is possible in the time domain, for example with a calculation of the form M ( n ) = w ₁ L ( n ) + w ₂ R ( n ) and weights w ₁ and w ₂ adaptive, or in frequency as shown here with reference to the figure 6 .

The "downmix" operation here consists of the blocks 603a, 603b, 603c and 603d for the passage in the frequency domain.

où|.| représente l'amplitude (module complexe) et ∠(.) la phase (argument complexe).The calculation of the mono signal is carried out in block 603e of "downmix" in which the signal is calculated in the frequency domain by the following formula: $$M\text{'}\left[j\right]=\frac{\left|\mathrm{The}\text{'}\left[j\right]\right|+\left|R\text{'}\left[\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="imgf0004.png"\; state="\{\{state\}\}">j</figure-callout>}\right]\right|}{2}\mathrm{.}{e}^{\angle \mathrm{The}\text{'}\left(j\right)}$$

where |. | represents the amplitude (complex module) and ∠ (.) the phase (complex argument).

Blocks 603f, 603g and 603h make it possible to bring the mono signal back into the time domain in order to be coded by block 304 as for the encoder illustrated in FIG. figure 3 .

An offset of T '= 80 + T samples is then obtained, an offset of 80 + 80 + 22 = 182 samples.

This offset makes it possible to synchronize the time frames of the left / right channels and those of the decoded mono signal.

The invention has been described here in the case of a G.722 encoder / decoder. it can obviously apply in the case of a modified G.722 encoder, for example including noise reduction mechanisms ( English) or including a scalable extension of G.722 with additional information. The invention can also be applied in the case of another mono encoder than the G.722 type such as for example a G.711.1 type encoder. In the latter case, the delay T must be adjusted to take into account the delay of the G.711.1 encoder.

• un autre fenêtrage que le fenêtrage sinusoïdal pourrait être utilisé,
• un autre recouvrement que le recouvrement à 50% entre fenêtres successives pourrait être utilisé
• une autre transformée fréquentielle que la transformée de Fourier, par exemple la transformée en cosinus discret modifiée (MDCT en anglais) pourrait être utilisée.

Similarly, the time-frequency analysis of the embodiment described with reference to the figure 3 could be replaced according to different variants:

• another windowing than sinusoidal windowing could be used,
• another covering than the 50% overlap between successive windows could be used
• another frequency transform than the Fourier transform, for example the modified discrete cosine transform (MDCT) could be used.

The embodiments described previously dealt with the case of a multichannel signal of the stereo signal type, the embodiment of the invention also extends to the more general case of the coding of multichannel signals (with more than 2 audio channels) starting from a mono or even stereo downmix.

In this case the coding of spatial information involves the coding and transmission of spatial information parameters. This is for example the case of 5.1 channel signals including a left channel (L), right (R), center (C), left rear (Ls for Left surround), right rear (Rs for Right surround ), and subwoofer (LFE for Low Freguency Effects ). The spatial information parameters of the multichannel signal then take into account the differences or the coherences between the different channels.

Encoders and decoders as described with reference to figures 3 , 4 and 6 can be integrated in a multimedia equipment type decoder lounge, computer or communication equipment such as a mobile phone or personal electronic diary.

The figure 7a represents an example of such a multimedia equipment or coding device comprising an encoder according to the invention. This device comprises a PROC processor cooperating with a memory block BM having a storage and / or working memory MEM.

• d'obtention, par trame de longueur prédéterminée, de paramètres d'information spatiale du signal multicanal;
• de division des paramètres d'information spatiale en une pluralité de blocs de paramètres;
• de sélection d'un bloc de paramètres en fonction de l'indice de la trame courante;
• de codage du bloc de paramètres sélectionné pour la trame courante.

The memory block may advantageously comprise a computer program comprising code instructions for implementing the steps of the coding method in the sense of the invention, when these instructions are executed by the processor PROC, and in particular the steps:

• obtaining, by frame of predetermined length, spatial information parameters of the multichannel signal;
• dividing the spatial information parameters into a plurality of parameter blocks;
• selecting a parameter block according to the index of the current frame;
• encoding the selected parameter block for the current frame.

Typically, the description of the figure 3 takes the steps of an algorithm of such a computer program. The computer program can also be stored on a memory medium readable by a reader of the device or downloadable in the memory space of the equipment.

The device comprises an input module adapted to receive a multichannel signal S _m representing a sound scene, either by a communication network, or by reading a content stored on a storage medium. This multimedia equipment may also include means for capturing such a multichannel signal.

The device comprises an output module capable of transmitting the coded spatial information parameters P _c and a sum signal Ss resulting from the coding of the multichannel signal.

In the same way, figure 7b illustrates an example of multimedia equipment or decoding device comprising a decoder according to the invention.

This device comprises a PROC processor cooperating with a memory block BM having a storage and / or working memory MEM.

• décodage de paramètres d'information spatiale reçus pour une trame courante de longueur prédéterminée de signal décodé;
• mémorisation des paramètres décodés pour la trame courante;
• obtention des paramètres décodés et mémorisés d'au moins une trame précédente et association de ces paramètres à ceux décodés pour la trame courante;
• reconstruction du signal multicanal à partir du signal décodé et de l'association de paramètres obtenus pour la trame courante.

The memory block may advantageously comprise a computer program comprising code instructions for implementing the steps of the decoding method in the sense of the invention, when these instructions are executed by the processor PROC, and in particular the steps of:

• decoding received spatial information parameters for a current frame of predetermined length of decoded signal;
• storing the decoded parameters for the current frame;
• obtaining decoded and stored parameters of at least one previous frame and associating these parameters with those decoded for the current frame;
• reconstruction of the multichannel signal from the decoded signal and the combination of parameters obtained for the current frame.

Typically, the description of the figure 4 takes the steps of an algorithm of such a computer program. The computer program can also be stored on a memory medium readable by a reader of the device or downloadable in the memory space of the equipment.

The device comprises an input module able to receive the coded spatial information parameters P _c and a sum signal S _s originating, for example, from a communication network. These input signals can come from a reading on a storage medium.

The device comprises an output module capable of transmitting a multichannel signal decoded by the decoding method implemented by the equipment.

This multimedia equipment may also include speaker-type reproduction means or communication means capable of transmitting this multi-channel signal.

Obviously, such multimedia equipment may include both the encoder and the decoder according to the invention. The input signal then being the original multichannel signal and the output signal, the decoded multichannel signal.

Claims (9)

1. Parametric coding method for a multichannel digital audio signal comprising a coding step (G.722 Cod) for coding a signal from a channel reduction matrixing of the multichannel signal, characterized in that it also comprises the following steps:

   - frequency transformation (Fen., FFT) of the multichannel signal to obtain the spectra of the multichannel signal, for each frame;

   - subdivision (D), for each frame, of the spectra of the multichannel signal, into a plurality of frequency sub-bands;

   - obtaining, for each frame of predetermined length and for each frequency sub-band, of spatial information parameters;

   - division (Div.) of the spatial information parameters into two blocks of parameters interleaving the parameters of the different frequency sub-bands;

   - selection of the first or of the second block of parameters to be coded out of the two blocks obtained in the division step, according to whether the current frame to be coded is of even index or of odd index;

   - coding (Q) of the spatial information parameters of the block of parameters selected for the current frame.
2. Method according to Claim 1, characterized in that said spatial information parameters are defined as the energy ratio between the channels of the multichannel signal.
3. Method according to Claim 1, characterized in that the coding of the spatial information parameters of a block of parameters is performed by non-uniform scalar quantization.
4. Method according to Claim 1, characterized in that it also comprises a principal component analysis step to obtain the spatial information parameters comprising a rotation angle parameter and an energy ratio between a principal component and an ambience signal.
5. Parametric decoding method for a multichannel digital audio signal comprising a decoding step (G.722 Dec) for decoding a signal from a channel reduction matrixing of the multichannel signal, characterized in that it also comprises the following steps:

   - decoding (Q^-1) spatial information parameters received for a current frame of predetermined length of the decoded signal;

   - storing (Mem) the decoded parameters for the current frame;

   - obtaining (Comp.P) the decoded and stored parameters of at least one preceding frame and associating these parameters with those decoded for the current frame, the decoded and stored parameters of a preceding frame and the decoded parameters of the current frame corresponding to the interleaved parameters of different frequency sub-bands of the decoding frequency band;

   - reconstructing (Synth.) of the multichannel signal from the decoded signal and from the association of parameters obtained for the current frame.
6. Computer program comprising code instructions for implementing the steps of a coding method according to one of Claims 1 to 4, when they are executed by a processor.
7. Computer program comprising code instructions for implementing the steps of a decoding method according to Claim 5, when they are executed by a processor.
8. Parametric coder for coding a multichannel digital audio signal comprising a coding module (304) for coding a signal from a channel reduction matrixing of the multichannel signal, characterized in that it also comprises:

   - a module for frequency transformation (307, 310) of the multichannel signal to obtain the spectra of the multichannel signal, for each frame;

   - a module for subdividing (313), for each frame, the spectra of the multichannel signal, into a plurality of frequency sub-bands,

   - a module for obtaining (314), for each frame of predetermined length, and for each frequency sub-band, spatial information parameters of the multichannel signal;

   - a module for dividing (315) the spatial information parameters into two blocks of parameters interleaving the parameters of the different frequency sub-bands;

   - a module for selecting (316) the first or the second block of parameters to be coded out of the two blocks obtained by the division module, according to whether the current frame to be coded is of even index or of odd index;

   - a coding module(312) suitable for coding the spatial information parameters of the block of parameters selected for the current frame.
9. Parametric decoder for decoding a multichannel digital audio signal comprising a decoding module (401) for decoding a signal from a channel reduction matrixing of the multichannel signal, characterized in that it also comprises:

   - a decoding module (404) for decoding spatial information parameters received for a current frame of predetermined length of the decoded signal;

   - storage space (412) for storing the parameters for the current frame;

   - a module (413) for obtaining the decoded and stored parameters of at least one preceding frame and associating these parameters with those decoded for the current frame, the decoded and stored parameters of a preceding frame and the decoded parameters of the current frame corresponding to the interleaved parameters of different frequency sub-bands of the decoding frequency band;

   - a reconstruction module (414) for reconstructing the multichannel signal from the decoded signal and from the association of parameters obtained for the current frame.

Images