EP2042001B1 - Binaural spatialization of compression-encoded sound data - Google Patents

Abstract

The invention is aimed at improving the quality of the filtering by transfer functions of HRTF type of signals (L, R) compressed in a transformed domain, for binaural playing on two channels (L-BIN, R-BIN), using a combination of HRTF filters (hL,L, hL,R) including a decorrelated version (HRTF-C*, HRTF-E*) of a few of these filters. For this purpose, a decorrelation cue is given with spatialization parameters (SPAT) accompanying the compressed signals (L, R). The Decorrelation comprises applying a different phase shift to each subband of the input signal combined with addition of an overall delay. The invention makes it possible to improve the broadening in the binaural rendition of audio scenes initially in a multi-channel format.

Description

The invention relates to the processing of sound data for spatialized reproduction.

The three-dimensional sound spatialization (so-called "3D rendering") of compressed audio signals occurs especially during the decompression of a 3D audio signal, for example encoded in compression and represented on a number of channels, to a number of different channels (two for example to allow the rendering of 3D audio effects on a headset).

The term "binaural" aims at the reproduction on a stereophonic headphones of a sound signal with nevertheless effects of spatialization. The invention is however not limited to the aforementioned technique and applies, in particular, to techniques derived from the "binaural", such as the so-called technical rendering techniques TRANSAURAL (registered trademark), that is to say on remote speakers. Such techniques can then use a "crosstalk cancellation" (or "cross-talk cancellation" in English), which consists in canceling the crossed acoustic paths, so that a sound, thus processed and then emitted by the loudspeakers, speakers, can be perceived only by one of the two ears of a listener. Subsequently, these two binaural and transaural rendering techniques are commonly referred to as binaural restitution.

Thus, more generally, the invention relates to the transmission of multichannel audio signals and their conversion for a spatialized rendering (with 3D rendering) on two channels. The rendering device (simple headset with earflaps for example) is most often imposed by the equipment of a user. The conversion can aim for example the case of a reproduction of a sound scene initially in 5.1 multichannel format (or 7.1, or other) by a simple audio headset (in binaural technique).

Of course, the invention also relates to the restitution, in the frame of a game or a video recording, for example, of one or more sound samples stored in files, with a view to their spatialization.

Regarding the prior art, reference is made to the document US2005 / 047618 , which shows a sound data processing method for spatialized three-dimensional rendition on two rendering channels for the left and right ear of a listener using a transfer function.

Among the known techniques in the field of binaural sound spatialization, different approaches have been proposed.

• associer à chaque source sonore S_i (ou chaque canal du signal multicanal) une position dans l'espace,
• filtrer ces sources dans le domaine fréquentiel par les fonctions de transfert acoustiques gauche HRTF-I et droite HRTF-r correspondant à la direction (ou à la position) choisie, et définies par leurs coordonnées polaires (θ₁, ϕ₁ ).

In particular, bi-binaural synthesis consists, with reference to the figure 1 relating to the prior art, to:

• associating with each sound source S _i (or each channel of the multichannel signal) a position in the space,
• filtering these sources in the frequency domain by the left HRTF-I and right HRTF-r acoustic transfer functions corresponding to the direction (or position) chosen, and defined by their polar coordinates (θ ₁ , φ ₁ ).

These transfer functions, commonly known as "HRTF" functions (for "Head Related Transfer Functions"), represent the acoustic transfer functions between the positions of the space and the auditory canal of each ear of the listener. The term "HRIR" (for "Head Related Impulse Response") denotes their temporal form or impulse response. These HRIR functions may further include a room effect.

We obtain for each sound source S _i two signals (left and right) which are then added to the left and right signals from the spatialization of all other sound sources, to finally give the signals L and R which will be broadcast in the left and right ears of the listener through two respective speakers (headphones of a binaural technique or remote speakers in transaural technique).

If N denotes the number of sound sources or incident audio streams to be spatialized, the number of filters, or transfer functions, necessary for the binaural synthesis is 2xN for rendering in static binaural spatialization and 4xN for rendering in dynamic binaural spatialization ( with transitions of transfer functions).

The treatment described above with reference to the figure 1 and involving HRTF transfer functions is conventional. It is often used for 3D rendering from two speakers. It may be the basis of an embodiment implemented by the present invention, as will be seen below. It is for this reason that it is introduced here.

Nevertheless, the invention starts from another type of prior art.

There are techniques of compression, often in a transformed domain, of signals in a multichannel format to be able to convey these signals, in particular through telecommunication networks, on a limited number of channels, for example one or two channels only. Thus, for the transmission of a signal in a multichannel format comprising more than two channels (for example 5.1, 7.1 or other), an encoder compresses the multichannel signal on only one or two channels (typically according to the bit rate offered on the network telecommunication) and also provides spatialization information. This achievement is illustrated on the Figure 2A where, as an example for a signal in a multichannel 5.1 format, five channels (C for a center speaker, FL for a front left speaker, FR for a right front speaker, BL for a loudspeaker). left rear speaker and BR for a right rear speaker) are encoded in compression by an ENCOD module capable of delivering two compressed L and R channels, as well as SPAT spatialization information. The compressed channels L and R, as well as the spatialization information SPAT, are then conveyed through one or more telecommunication networks RES, on one or two channels according to the available bit rate ( Figure 2B ).

With reference to the Figure 2C at the reception of the compressed signal on both L and R channels, a decoder (DECOD) reconstructs the original signal in the initial multichannel format by means of the SPAT spatialization information delivered by the coder and, in the example of Figures 2A and 2C , there are still five channels, after decoding, feeding five speakers (HP-FL, HP-FR, HP-C, HP-BL and HP-BR) for a 5.1 format playback.

Many types of parametric encoders / decoders, including standardized ones, offer such possibilities.

Audio encoders (AAC, MP3) use time-frequency representations of signals to compress information. These representations are based on an analysis by filter banks or by time-frequency transformation of the MDCT type (for "Modified Discrete Cosine Transform"). In the case where a binaural spatialization must be performed after an audio decoding, the filtering operations are advantageously performed from the outset in the transformed domain.

• " A Generic Framework for Filtering in Subband Domain", A.Benjelloun Touimi, Proceeding IEEE - 9th Workshop on Digital Signal Processing, Hunt, Texas, USA, Octobre 2000 .

Recent work on filtering in the transformed subband domain has formalized the filtering architecture for a bank of filters commonly used in audio coders. We can usefully refer to the document:

• " A Generic Framework for Filtering in Subband Domain ", A.Benjelloun Touimi, Proceeding IEEE - 9th Workshop on Digital Signal Processing, Hunt, Texas, USA, October 2000 .

où :

• N_h est la longueur du filtre dans le domaine temporel,
• L_q /K_h +2, avec k_h =[N _h/ 64], la longueur des filtres en sous-bandes (pour 64 sous-bandes).

A more recent filtering technique in the transformed domain of complex QMFs (for "Quadrature Mirror Filters") has been proposed in the "MPEG Surround" standard. This technique aims to convert the impulse response (finite) of the temporal filter noted h (v) into a set of M complex filters denoted h _m (l), where M is the number of sub-frequency bands. The conversion is performed by analyzing the time filter h (v) by a complex filter bank similar to the QMF filterbank used for signal analysis. In an exemplary embodiment, the prototype filter q ( v ) used to generate the conversion filter bank may be of length 192. An extension with zeros of the temporal filter is defined by the following formula: $$\tilde{h}\left(v\right)=\{\begin{array}{ll}h\left(v\right),& v=0,1,...,{\mathrm{NOT}}_h-1;\\ 0,& \mathrm{if\; not},\end{array},$$

or :

• N _h is the length of the filter in the time domain,
• L _q / K _h + 2, with k _h = [ N _h / 64], the length of the filters in sub-bands (for 64 subbands).

avec :

• m = 0,1,...,63 , correspondant à l'indice de la sous-bande
• l =0,1,..., K_h +1 , correspondant à l'indice temporel dans le domaine décimé des sous-bandes.

The conversion is therefore given by the following formula: $$h_m\left(l\right)={\displaystyle \underset{v=0}{\overset{191}{\Sigma }}}\tilde{h}\left(v+64\left(l-2\right)\right)q\left(v\right)\exp \left(-j\frac{\pi }{64}\left(m+\frac{1}{2}\right)\left(v-95\right)\right),$$

with:

• m = 0.1, ..., 63, corresponding to the index of the sub-band
• l = 0.1 , ..., K _h +1 , corresponding to the time index in the decimated domain of the subbands.

In more generic terms, it will be understood that such processing, directly in the transformed domain, makes it possible to go from a representation of the compressed signal on two L, R channels to a representation of the signal on two reproduction channels L-BIN, R -BIN ( figure 3 ) with binaural or transaural widening. For this purpose, a transcoding (DECOD BIN module of the figure 3 ) which is based on an approach consisting in reconstructing, from the compressed signals L, R and SPAT spatialization information, the transfer functions, of the HRTF type, between an ear of a listener and each (virtual) loudspeaker which would have been powered by a given channel of the original multichannel format.

• une pour le chemin A entre le haut-parleur avant gauche HP-FL et l'oreille gauche OL de l'auditeur AU,
• une pour le chemin B entre le haut-parleur avant gauche HP-FL et l'oreille droite OR de l'auditeur AU,
• une pour le chemin C entre le haut-parleur arrière gauche HP-BL et l'oreille gauche OL de l'auditeur AU,
• une pour le chemin D entre le haut-parleur arrière gauche HP-BL et l'oreille droite OR de l'auditeur AU,
• une pour le chemin G entre le haut-parleur avant droit HP-FR et l'oreille droite OR de l'auditeur AU,
• une pour le chemin H entre le haut-parleur avant droit HP-FR et l'oreille gauche OL de l'auditeur AU,
• une pour le chemin F entre le haut-parleur arrière droit HP-BR et l'oreille droite OR de l'auditeur AU,
• une pour le chemin E entre le haut-parleur arrière droit HP-BR et l'oreille gauche OL de l'auditeur AU,
• une pour le chemin J entre le haut-parleur du centre HP-C et l'oreille gauche OL de l'auditeur AU, et
• une pour le chemin I entre le haut-parleur du centre HP-C et l'oreille droite OR.

So, with reference now to the figure 4 illustrating a "virtual" rendering in 5.1 format, thus from five speakers, the transcoding implemented by the DECOD BIN module of the figure 3 must consider ten transfer functions:

• one for path A between the left front speaker HP-FL and the left ear OL of the AU listener,
• one for path B between the left front speaker HP-FL and the right ear OR of the AU listener,
• one for path C between the left rear loudspeaker HP-BL and the left ear OL of the listener AU,
• one for the path D between the left rear loudspeaker HP-BL and the right ear OR of the listener AU,
• one for the G path between the right front speaker HP-FR and the right ear OR of the AU listener,
• one for the H path between the HP-FR right front speaker and the left ear OL of the AU listener,
• one for path F between the right rear speaker HP-BR and the right ear OR of the AU listener,
• one for the path E between the right rear speaker HP-BR and the left ear OL of the listener AU,
• one for the path J between the loudspeaker of the center HP-C and the left ear OL of the listener AU, and
• one for path I between the center speaker HP-C and the right ear OR.

Thus, the subband filters in the transformed domain are calculated for each ear and for each of the five positions of the loudspeakers. This technique is often called "virtual speaker technology".

Using the sub-band representation of the binaural filters determined as described above from the HRTF transfer functions, the binaural spatialization can then be advantageously performed by applying these binaural filters directly to the transformed domain at the heart of the decoder. DECOD BIN audio as shown on the figure 3 .

Thus, this type of DECOD BIN decoder uses a monophonic or stereophonic representation (compressed L, R channels) of the multichannel audio scene, representation to which are associated SPAT spatialization parameters (which may consist, for example, of energy differences between channels and correlation indices between channels). These SPAT parameters are used at decoding to reproduce the original multichannel sound scene.

In addition, when the original signal is encoded by a parametric encoder (for example in the sense of recent work in the "MPEG Surround" standard), in addition to the monophonic or stereophonic signal transmitted and spatialization information, the decoding can use representations decorrelated from these L, R signals (which are obtained, for example, by the application of filters, all-pass decorrelation or reverb filters). These signals are then adjusted in energy thanks to interchannel energy differences, then recombined to obtain the multichannel signal for restitution.

In particular, the parametric encoder (ENCOD - Figure 2A ) from the multichannel format to two compressed channels (stereo or mono) according to the draft standard "MPEG Surround" provides cross-channel decorrelation information in the initial multichannel format and this decorrelation information can be taken over by the peer parametric decoder (DECOD - Figure 2C ) when rendering in the initial multichannel format.

• « http://www.chiariglione.org/mpeg/technologies/mpd-mps/index.htm »

et des précisions quant à un tel codeur selon ce projet de norme peuvent être trouvées dans :

• " MPEG Spatial Audio Coding / MPEG Surround: Overview and Current Status", J.Breebaart et al., dans 119th Conv. Aud. Eng. Soc (AES), New York, NY, USA, Octobre 2005 .

A description of preparatory work for this standard is given at the URL:

• "Http://www.chiariglione.org/mpeg/technologies/mpd-mps/index.htm"

and details of such an encoder according to this draft standard can be found in:

• " MPEG Spatial Audio Coding / MPEG Surround: Overview and Current Status ", J.Breebaart et al., In 119th Conv. Aud. Eng. Soc (AES), New York, NY, USA, October 2005 .

In the case of a parametric audio decoder for a binaural restitution (DECOD BIN - figure 3 ), it is advantageously possible to simplify the filtering operations by combining the front and rear filters corresponding to the different left speakers (an equivalent treatment being worn also for the right speakers). This combination is performed according to the target energies of the audio channels given by the spatialization parameters. This combination, for the left ear and left front and left left channels, is performed in the transformed domain according to an expression (1) of the type: $${\mathrm{h}}_{\mathrm{The},\mathrm{The}}={\mathrm{boy\; Wut}}_{\mathrm{The},\mathrm{The}}{\sigma }_{\mathit{FL}}\exp \left(-j{\phi }_{\mathit{FL}\mathit{.}\mathit{BL}}^{\mathrm{The}}{\mathrm{<figure-callout\; id="2"\; label="\sigma \; BL"\; filenames="imgf0001.png"\; state="\{\{state\}\}">\sigma \; </figure-callout>}}_{\mathit{BL}}^{\mathit{}}\right){\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{FL}}+{\mathrm{boy\; Wut}}_{\mathrm{The}\mathrm{.}\mathit{The}}{\sigma }_{\mathit{BL}}\exp \left(j{\phi }_{\mathit{FL}\mathit{.}\mathit{BL}}^{\mathrm{The}}{\sigma }_{\mathit{<figure-callout\; id="2"\; label="FL"\; filenames="imgf0001.png"\; state="\{\{state\}\}">FL</figure-callout>}}^2\right){\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{BL}}$$

• h_L,_L est le filtre correspondant aux contributions des canaux avant et arrière gauche,
• g_L,_L est le gain associé à l'ensemble des canaux gauches,
• σ² _FL et σ² _BL sont les énergies utiles des canaux respectivement avant et arrière gauches,
• h_L,FL et h_L,BL sont les fonctions de transfert dans le domaine des sous-bandes entre l'oreille gauche et les haut-parleurs respectivement avant et arrière gauche (chemins A et C de la figure 4),
• Φ^L _FL,BL est le déphasage correspondant au retard entre les filtres temporels avant et arrière gauche h_L,FL et h_L,BL.

In this expression:

• h _L , _L is the filter corresponding to the contributions of the left front and rear channels,
• g _L , _L is the gain associated with all the left channels,
• σ ² _FL and σ ² _BL are the useful energies of the respectively front and left channels,
• h _{L, FL} and h _{L, BL} are the transfer functions in the field of the sub-bands between the left ear and the speakers respectively front and rear left (paths A and C of the figure 4 )
• Φ ^L _{FL, BL} is the phase shift corresponding to the delay between the forward and backward time filters h _{L, FL} and h _{L, BL} .

The purpose of this phase compensation, a function of the target energy of the channels, is to avoid a so-called "coloring" effect resulting from the addition of two filters shifted in time (comb filtering).

With reference to the Figure 5A , a decoder receives the SPAT spatialization parameters accompanying the signals compressed on two L and R channels in the example shown, and it has been illustrated on this same Figure 5A , how the above filter h _{L, L} applies to the compressed channel L to form a component of the L-BIN signal for binaural restitution. Nevertheless, as represented again on the Figure 5A , it is also necessary to take into account the compressed signal on the channel R which must, for its part, be filtered by a filter involving HRTF transfer functions (denoted H _{L, FR} and H _{L, BR} ) relating to the cross paths H and E of the figure 4 , always to the left ear. The filter corresponding to these crossed paths (denoted h _{L, R} ) is computed according to the gains, target energies and phase shifts, taken from the SPAT spatialization parameters, using an expression equivalent to the relation (1) given previously. . This filter h _{L, R} is finally applied to the compressed signal on the channel R. It is also necessary to take into account the "contribution" of the central loudspeaker in the construction of the signal intended for the binaural restitution L-BIN and, for this is done by applying a filter h _{L, C} ( Figure 5A ) to a combination (for example by addition) of the compressed signals of the two channels L and R to take into account here the path J to the left ear OL of the figure 4 .

• le signal compressé sur le canal R filtré par le filtre h_R,R représentant les fonctions HRTF des haut-parleurs de droite (chemins directs G et F de la figure 4) ;
• le signal compressé sur le canal L filtré par le filtre h_R,L représentant les fonctions HRTF des haut-parleurs de gauche (chemins croisés B et D de la figure 4) ; et
• une combinaison des signaux compressés L et R filtrée par le filtre h_R,C représentant les fonctions HRTF du haut-parleur de centre (chemin direct I de la figure 4).

Still referring to the Figure 5A , an equivalent treatment is provided for the construction of the R-BIN signal for restitution binaural for the right ear OD, with three contributions given by:

• the signal compressed on the R channel filtered by the filter h _{R, R} representing the HRTF functions of the right speakers (direct paths G and F of the figure 4 );
• the compressed signal on the channel L filtered by the filter h _{R, L} representing the HRTF functions of the left speakers (crossed paths B and D of the figure 4 ); and
• a combination of the compressed signals L and R filtered by the filter h _{R, C} representing the HRTF functions of the center loudspeaker (direct path I of the figure 4 ).

On the Figure 5B another example is shown, in which a decoder receives the compressed signal on a single channel M, accompanying the spatialization parameters SPAT. In the example shown, the M channel is duplicated in two L and R channels and the rest of the processing is strictly equivalent to the treatment shown in FIG. Figure 5A .

The two signals L-BIN and R-BIN resulting from these filterings can then be applied to two loudspeakers intended respectively for the left ear and the right ear of the listener after a transition from the transformed domain to the time domain.

However, a problem with this combination of filters for binaural playback is that it does not take into account any decorrelation between the front and rear channels. This information, however used in the decoding of a 5.1 scene of an encoder according to the aforementioned project of the MPEG Surround standard, is not exploited in the binaural decoding technique. Thus, when the sound scene includes decorrelating effects between the front and rear channels (for example for reverberated signals), this information is not used in the combination of the HRTF filters, which leads to a degradation of the spatialization quality. and in particular the enveloping effect of the 3D audio scene. Thus, the binaural format is not optimal.

The present invention improves the situation.

• obtenir, avec les données compressées sur ledit nombre réduit de canaux, des paramètres de spatialisation,
• pour chaque voie de restitution associée à une oreille de l'auditeur, former, à partir desdits paramètres de spatialisation, une combinaison de filtres représentatifs chacun de fonctions de transfert entre cette oreille de l'auditeur et des haut-parleurs susceptibles d'être alimentés par des canaux respectifs du format multicanal initial, et
• appliquer aux données compressées la combinaison de filtres associée à chaque voie de restitution.

It firstly aims at a sound data processing method for a spatialized restitution in three dimensions on two playback channels for the respective ears of a listener,
the sound data being initially represented in a multichannel format and then encoded in compression on a reduced number of channels (for example one or two channels),
said initial multichannel format consisting of providing more than two channels capable of supplying respective loudspeakers,
the process comprising the steps:

• obtain, with the compressed data on said reduced number of channels, spatialization parameters,
• for each playback channel associated with an ear of the listener, forming, from said spatialization parameters, a combination of filters representative each of transfer functions between this listener's ear and speakers capable of being powered by respective channels of the initial multichannel format, and
• apply to the compressed data the filter combination associated with each rendition channel.

• pour chaque voie de restitution associée à une oreille de l'auditeur, déterminer à partir desdits paramètres de spatialisation au moins une fonction de transfert d'un haut-parleur situé à l'arrière de l'oreille de l'auditeur et représentative d'une décorrélation entre les canaux du format multicanal respectivement associés au haut-parleur arrière et à au moins un haut-parleur situé devant l'oreille de l'auditeur, et
• pour chaque voie de restitution, intégrer ladite fonction de transfert représentative d'une décorrélation dans ladite combinaison de filtres associée à cette voie de restitution.

The process according to the invention further comprises the steps of:

• for each playback channel associated with an ear of the listener, determining from said spatialization parameters at least one transfer function of a loudspeaker located at the back of the listener's ear and representative of a decorrelation between the channels of the multichannel format respectively associated with the rear loudspeaker and at least one speaker located in front of the listener's ear, and
• for each rendering channel, integrating said transfer function representative of a decorrelation in said combination of filters associated with this rendering path.

Spatial restitution on two paths, within the meaning of the invention, can be in binaural or transaural format as well. The initial multichannel format can be of ambiophonic type (or "ambisonic" in English and aiming at the decomposition of the sound signal in a spherical harmonics base). Alternatively, it can be a type of 5.1 or 7.1, or 10.2. It will be understood that for the latter types of format using channels intended to respectively supply at least pairs of speakers front-left / left-back, on the one hand, and front-right / right-back, d On the other hand, the decorrelation information can be directed to the respective channels of the front / rear speakers preferably associated with the same ear (left or right).

According to one advantage provided by the invention, since this decorrelation information at the rear of a 3D scene is represented in the binaural or transaural restitution, a better representation of the ambiances is obtained, for example crowd noise or reverberation at the back of a scene, or otherwise, contrary to the achievements of the prior art.

• une fonction de transfert brute du haut-parleur situé à l'arrière, et
• une version de la fonction de transfert de ce haut-parleur, représentative de la décorrélation.

In a particular embodiment, the combination of filters comprises a weighting, according to a chosen coefficient, between:

• a function of raw transfer of the loudspeaker located at the back, and
• a version of the transfer function of this speaker, representative of the decorrelation.

This weighting advantageously makes it possible to favor the gross transfer function of this rear loudspeaker, or the decorrelated version of this raw transfer function, depending on whether the signal in the back channel of the initial multichannel format is correlated or not with at least one signal. from one of the front channels.

• la fonction de transfert d'un haut-parleur avant,
• la fonction de transfert d'un haut-parleur arrière, et
• la fonction de transfert représentative d'une décorrélation entre canaux,

et ces haut-parleurs avant et arrière sont situés sur un même côté par rapport à l'auditeur. Il peut s'agit par exemple des haut-parleurs avant et arrière situés tous les deux à gauche (ou tous les deux à droite) de l'auditeur au format 5.1 (comme illustré sur la figure 4). Dans une telle réalisation, lorsque la pondération entre la version décorrélée et la version brute des fonctions de transfert est prévue, il pourra être avantageux de privilégier la version décorrélée dans la combinaison de filtres des haut-parleurs de gauche pour la voie de restitution sur l'oreille droite (et inversement) et privilégier la version brute (non décorrélée) dans la combinaison de filtres des haut-parleurs de droite (gauche) pour la voie de restitution sur l'oreille droite (gauche).Moreover, in a particular embodiment, the combination of filters associated with a restitution channel comprises at least one filter grouping from:

• the transfer function of a front speaker,
• the transfer function of a rear speaker, and
• the representative transfer function of a decorrelation between channels,

and these front and rear speakers are located on the same side of the listener. This may include, for example, the front and rear speakers located to the left (or both to the right) of the 5.1 format listener (as shown in the figure below). figure 4 ). In such an embodiment, when the weighting between the uncorrelated version and the raw version of the transfer functions is provided, it may be advantageous to prefer the decorrelated version in the combination of filters of the left speakers for the playback channel on the left. right ear (and vice versa) and prefer the raw version (not decorrelated) in the right (left) speaker filter combination for the right ear (left) feedback channel.

Advantageously, the encoding in compression implements a parametric encoder delivering, in the compressed stream including the spatialization parameters, cross-channel decorrelation information of the multichannel format, from which the aforementioned weighting can be determined dynamically. .

Thus, in this embodiment, for a transcoding between a multichannel format and a binaural format, the aforementioned combination of transfer functions takes advantage of the information already present concerning the correlation between channel signals in multichannel format, this information simply being provided by the coder parametric, with the aforementioned spatialization parameters.

By way of example, it is recalled that the parametric decoder according to the draft MPEG Surround standard delivers such cross-channel decorrelation information in multichannel 5.1 format.

• les figures 6A et 6B illustrent à titre d'exemple un traitement par filtrage de données compressées (sur deux canaux dans l'exemple représenté), le filtrage étant déterminé par la mise en oeuvre du procédé au sens de l'invention pour délivrer des signaux L-BIN et R-BIN destinés à alimenter respectivement les voies gauche et droite d'un dispositif de restitution binaurale tel qu'un casque à deux oreillettes, en tenant compte d'une décorrélation avant/arrière, et
• la figure 7 illustre schématiquement la structure d'un module mettant en oeuvre le procédé au sens de l'invention.

Other advantages and characteristics of the invention will appear on reading the detailed description given below by way of example, and on the observation of the appended drawings, in which, in addition to the FIGS. 1, 2A, 2B, 2C, 3 and 4, 5A and 5B commented above:

• the Figures 6A and 6B illustrate by way of example a filtering treatment of compressed data (on two channels in the example shown), the filtering being determined by the implementation of the method according to the invention for delivering L-BIN and R signals -BIN for respectively feeding the left and right channels of a binaural rendering device such as a headset with two earpieces, taking into account a forward / backward decorrelation, and
• the figure 7 schematically illustrates the structure of a module implementing the method in the sense of the invention.

With reference to the Figure 6A , the compressed signal, often in the transformed domain, is first recovered on two L and R channels in the example represented, as well as the SPAT spatialization parameters provided by an encoder such as the ENCOD module of the Figure 2A previously described. From the SPAT spatialization parameters, transfer functions are determined to construct a combination of filters ("+" sign of the Figure 6A ), each filter being to be applied to a channel L (filter h _{L, L} of the Figure 5A ), or R (filter h _{L, R} of the Figure 5A ), or a combination of these channels (filter h _{L, C} of the Figure 5A ) to construct a signal feeding one of two binaural L-BIN recovery channels. These transfer functions, of the HRTF type, are representative of the disturbances experienced by an acoustic wave on a path between a loudspeaker that would have been fed by a channel of the initial multichannel format and an ear of the listener. For example, if the audio content is initially in 5.1 format, as described above with reference to the figure 4 a total of ten HRTF transfer functions are determined, five HRTF functions for the right ear (on paths B, D, G, F and I of the figure 4 ) and five HRTF functions for the left ear (on the A, C, H, E and J paths). It indicates that the speaker of the center is treated separately in the binaural spatialization and obtaining the corresponding filter h _{L, C} or hic will not be described here, it being understood that it does not involve, a priori in the example described, the object of the 'invention.

Thus, in general terms, the HRTF functions of front and rear speakers, located on the same side of the listener are thus grouped to construct each filter of a filter combination specific to a playback channel on an ear of a listener. A grouping of HRTF functions to construct a filter is for example an addition, by means of multiplicative coefficients, an example of which will be described later.

For the purposes of the invention, from the recovered SPAT spatialization parameters, a decorrelated version of the HRTF functions of the loudspeakers located at the rear of the listener (paths C, D, E and F of the listener) is also determined. figure 4 ) and this decorrelated version is integrated with each grouping to form a filter to be applied to a compressed channel.

• la fonction HRTF-A (pour le haut-parleur avant gauche selon un chemin direct vers l'oreille gauche OL de la figure 4),
• la fonction HRTF-C (pour le haut-parleur arrière gauche selon un chemin direct vers l'oreille gauche),
• et la version décorrélée de cette fonction HRTF-C, notée HRTF-C*, pour former le filtre à appliquer au canal compressé L.

Un second regroupement comporte :

• la fonction HRTF-H (pour le haut-parleur avant droit selon un chemin croisé vers l'oreille gauche),
• la fonction HRTF-E (pour le haut-parleur arrière droit selon un chemin croisé),
• et la version décorrélée de cette fonction HRTF-E, notée HRTF-E*, pour former le filtre à appliquer au canal compressé R.

L'addition des deux signaux résultant de tels filtrages sera une composante du signal alimentant la voie de restitution binaurale L-BIN associée à l'oreille gauche.By way of a purely illustrative example, the initial sound data can be in multichannel format 5.1 and, with reference to FIG. Figure 6A a first grouping includes:

• the function HRTF-A (for the front left speaker in a direct path to the left ear OL of the figure 4 )
• the HRTF-C function (for the left rear speaker in a direct path to the left ear),
• and the decorrelated version of this HRTF-C function, denoted HRTF-C *, to form the filter to be applied to the compressed channel L.

A second grouping includes:

• the HRTF-H function (for the right front speaker in a cross path to the left ear),
• the HRTF-E function (for the right rear speaker in a cross path),
• and the decorrelated version of this HRTF-E function, denoted HRTF-E *, to form the filter to be applied to the compressed channel R.

The addition of the two signals resulting from such filtering will be a component of the signal feeding the L-BIN binaural restoring channel associated with the ear left.

A similar treatment is planned to construct the signal intended to feed the other binaural R-BIN restitution channel of the Figure 6B . Here, we take into account the HRTF functions of the paths leading to the right ear OD of the listener AU ( figure 4 ). A first grouping includes the HRTF-G functions (for the right front speaker in a direct path), HRTF-F (for the right rear speaker in a direct path) and the decorrelated version HRTF-F * of the HRTF-F function to form the filter to be applied to the compressed channel R. A second grouping includes the function HRTF-B (for the front left speaker according to a cross path), the function HRTF-D (for the rear speaker left in a cross path) and the decorrelated version, denoted HRTF-D *, of the HRTF-D function, to form the filter to be applied to the compressed channel L.

Finally, the filter combinations integrating the decorrelated versions of the HRTF functions of the rear loudspeakers are applied to the L and R compressed channels to deliver the L-BIN and R-BIN rendering channels, for spatial binaural rendering with 3D rendering.

In the example shown on the Figures 6A and 6B , the received sound data is encoded in compression on two stereo channels L and R as illustrated in the example of the Figure 5A . As a variant, they could be encoded in compression on a single monophonic channel M, as shown in FIG. Figure 5B , in which case the filter combinations are applied to the monophonic channel (duplicated) as shown in FIG. Figure 5B , to supply two signals supplying the two L-BIN, R-BIN reproduction channels respectively.

In an advantageous embodiment, the initial sound data is in the 5.1 multichannel format and is encoded in compression by a parametric encoder according to the aforementioned draft standard MPEG Surround. More particularly, during such encoding, it is possible to obtain, among the spatialization parameters provided, decorrelation information between the right rear channel and the right front channel (respective loudspeakers HP-BR and HP-FR of the figure 4 ), as well as homologous decorrelation information between the left rear channel and the left front channel (respective loudspeakers HP-FR and HP-BR of the figure 4 ).

This decorrelation information, in a 5.1 format, is intended to make the reproduction of the rear speakers as independent as possible from the reproduction of the front speakers, to enrich, in 5.1 format, the enveloping effect by noise of reverb or audience for concert recordings for example. It is recalled that this enrichment of the 3D envelopment has not been proposed in binaural restitution and an advantage of the invention is to take advantage of the availability of the decorrelation information among the SPAT spatialization parameters to build uncorrelated versions of the functions. HRTF that integrate advantageously with filter combinations for binaural restitution.

According to another advantage, these filter combinations can be calculated directly in the transformed domain, for example in the field of sub-bands, and the filters representing the decorrelated versions of the HRTF functions of the rear loudspeakers can be obtained for example by applying the initial HRTF functions a phase shift function of the sub-frequency band considered.

More generally, the decorrelation filters can be so-called "natural" reverb filters (recorded in a particular acoustic environment such as a concert hall for example), or "synthetic" (created by summation of multiple reflections of decreasing amplitudes in the time). The application of a decorrelated filter can therefore return to apply to the signal broken down into frequency subbands a different phase difference in each of the subbands, combined with the addition of a global delay. In the case of a parametric decoder of the aforementioned type (formula (1) given previously in the description of the prior art), this amounts to multiplying each sub-frequency band by a complex exponential, of different phase in each sub-band. . These decorrelation filters can therefore correspond to syntheses of phase-shifting all-pass filters.

avec les mêmes notations explicitées précédemment et où h ^Decorr _L,BL représente la version décorrélée de la fonction de transfert du haut-parleur arrière gauche. Bien entendu, on prévoit un même type de relations donnant les autres filtres h _L,R , h_R,R et h _R,L (figures 5A et 5B).
Par exemple pour le filtre h _L,R propre aux chemins croisés vers l'oreille gauche, on a : $${\mathrm{h}}_{L\mathrm{.}R}=g_{L\mathrm{.}R}{\sigma }_{\mathit{FR}}\exp \left(-j{\phi }_{\mathit{FR}\mathit{.}\mathit{BR}}^L{\sigma }_{\mathit{BR}}^2\right){\mathrm{h}}_{L\mathrm{.}\mathit{FR}}+g_{L\mathrm{.}\mathit{R}}{\sigma }_{\mathit{BR}}\exp \left(j{\phi }_{\mathit{FR}\mathit{.}\mathit{BR}}^L{\sigma }_{\mathit{BR}}^2\right)\left(\alpha {\mathrm{h}}_{L\mathrm{.}\mathit{BR}}+\left(1-\alpha \right){\mathrm{h}}_{L\mathrm{.}\mathit{BR}}^{\mathit{Decorr}}\right)$$

Advantageously, a weighting is provided between the transfer function of a rear loudspeaker and its uncorrelated version in the same group forming a filter. Thus, by taking again the formula (1) given previously for the calculation of a filter for example h _{L, L} for the left ear, we introduce weighting coefficients α and (1-α) and the decorrelated version of a transfer function as follows: $${\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathrm{The}}={\mathrm{boy\; Wut}}_{\mathrm{The}\mathrm{.}\mathrm{The}}{\sigma }_{\mathit{FL}}\exp \left(-j{\phi }_{\mathit{FL}\mathit{.}\mathit{BL}}^{\mathrm{The}}{\mathrm{<figure-callout\; id="2"\; label="\sigma \; BL"\; filenames="imgf0001.png"\; state="\{\{state\}\}">\sigma \; </figure-callout>}}_{\mathit{BL}}^{\mathit{}}\right){\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{FL}}+{\mathrm{boy\; Wut}}_{\mathrm{The}\mathrm{.}\mathit{The}}{\sigma }_{\mathit{BL}}\exp \left(j{\phi }_{\mathit{FL}\mathit{.}\mathit{BL}}^{\mathrm{The}}{\sigma }_{\mathit{<figure-callout\; id="2"\; label="FL"\; filenames="imgf0001.png"\; state="\{\{state\}\}">FL</figure-callout>}}^2\right)\left(\alpha {\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{BL}}+\left(1-\alpha \right){\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{BL}}^{\mathit{Decorr}}\right)$$

with the same notation explained above and where h ^Decorr _{L, BL} represents the decorrelated version of the transfer function of the left rear loudspeaker. Of course, the same type of relations is provided giving the other filters h _{L, R} , h _{R, R} and h _{R, L} ( Figures 5A and 5B ).
For example, for the filter H _{L, R} specific to paths crossed to the left ear, we have: $${\mathrm{h}}_{\mathrm{The}\mathrm{.}R}={\mathrm{boy\; Wut}}_{\mathrm{The}\mathrm{.}R}{\sigma }_{\mathit{EN}}\exp \left(-j{\phi }_{\mathit{EN}\mathit{.}\mathit{BR}}^{\mathrm{The}}{\sigma }_{\mathit{<figure-callout\; id="2"\; label="BR"\; filenames="imgf0001.png"\; state="\{\{state\}\}">BR</figure-callout>}}^2\right){\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{EN}}+{\mathrm{boy\; Wut}}_{\mathrm{The}\mathrm{.}\mathit{R}}{\sigma }_{\mathit{BR}}\exp \left(j{\phi }_{\mathit{EN}\mathit{.}\mathit{BR}}^{\mathrm{The}}{\sigma }_{\mathit{<figure-callout\; id="2"\; label="BR"\; filenames="imgf0001.png"\; state="\{\{state\}\}">BR</figure-callout>}}^2\right)\left(\alpha {\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{BR}}+\left(1-\alpha \right){\mathrm{h}}_{\mathrm{The}\mathrm{.}\mathit{BR}}^{\mathit{Decorr}}\right)$$

More specifically, weighting is provided by different coefficients α ₁ , (1-α ₁ ) and α ₂ , (1-α ₂ ) depending on whether the rear loudspeaker is on the same side as the considered ear (α = α ₁ giving the filters h _{L, L} and h _R , _R ) or not (α = α ₂ giving the filters h _L , _R and h _{R, L} ). Preferably, the decorrelated version is preferred for crossed paths (right rear speaker for the left ear and left rear speaker for the right ear), so that, in general, the coefficient α ₁ , may often be greater than the coefficient α ₂ .

$$\mathrm{\alpha }=\mathrm{sqrt}\left({{\mathrm{\sigma }}_{\mathrm{BL}}}^2\right),\mathrm{sinon},$$

où la notation « sqrt » vise la fonction « racine carrée », la notation « abs » vise la fonction « valeur absolue » et le terme ICC_L représente l'information de décorrélation (appelée autrement « indice de corrélation ») entre le canal avant et le canal arrière du même côté gauche et fait partie des paramètres de spatialisation transmis par l'encodeur selon le projet de norme MPEG Surround indiqué ci-avant. Comme décrit ci-avant, le terme σ_BL représente l'énergie cible du canal arrière gauche lorsqu'il s'agit de déterminer le coefficient α pour calculer le filtre h _L,L (α=α₁). Bien entendu, une expression équivalente peut être appliquée pour calculer le coefficient de pondération α intervenant dans le filtre homologue h _R,R propre aux chemins acoustiques directs vers l'oreille droite. Néanmoins, pour les filtres h _L,R et h _R,L propres aux chemins croisés, par exemple pour le filtre h _L,R propre aux chemins croisés vers l'oreille gauche, le coefficient α=α₂ peut préférentiellement s'écrire : $${\mathrm{\alpha }}_2=\mathrm{abs}\left({\mathrm{ICC}}_{\mathrm{R}}\right),\mathrm{si\; abs}\left({\mathrm{ICC}}_{\mathrm{R}}\right)>{{\mathrm{\sigma }}_{\mathrm{BR}}}^2,$$

$${\mathrm{\alpha }}_2={{\mathrm{\sigma }}_{\mathrm{BR}}}^2,\mathrm{sinon},$$

le terme σ_BR représentant l'énergie cible du canal arrière droit et le terme ICC_R représentant l'indice de corrélation entre le canal avant droit et le canal arrière droit.
On relèvera que la fonction « sqrt » ne s'applique plus pour les chemins croisés et pour le calcul du coefficient correspondant α₂, dans l'exemple décrit. En effet, les énergies cibles et les indices de corrélation sont des termes compris entre 0 et 1 de sorte que le coefficient α₂ est généralement inférieur au coefficient α₁.In practice, the coefficients α (α ₁ or α ₂ ) are given by variable weighting functions so as to dynamically favor the raw version of the HRTF function of the rear loudspeaker or its uncorrelated version depending on whether the back signal is correlated or no with the signal before. This gives a better representation of the ambiances (crowd noise, reverb, or other) in the 3D rendering.
The weighting function α can be defined dynamically by means of the decorrelation information provided with the spatialization parameters. as a non-limitative example, as follows: $$\mathrm{\alpha }=\mathrm{sqrt}\left(\mathrm{abs}\left({\mathrm{ICC}}_{\mathrm{The}}\right)\right),\mathrm{if\; abs}\left({\mathrm{ICC}}_{\mathrm{The}}\right)>{{\mathrm{<figure-callout\; id="2"\; label="\sigma \; BL"\; filenames="imgf0001.png"\; state="\{\{state\}\}">\sigma \; </figure-callout>}}_{\mathrm{BL}}}^2$$

$$\mathrm{\alpha }=\mathrm{<figure-callout\; id="2"\; label="sqrt\; \sigma \; BL"\; filenames="imgf0001.png"\; state="\{\{state\}\}">sqrt\; </figure-callout>}\left({{\mathrm{\sigma }}_{\mathrm{BL}}}^{}\right)\left({{}_{}}^2\right),\mathrm{if\; not},$$

where the notation "sqrt" refers to the function "square root", the notation "abs" refers to the function "absolute value" and the term ICC _L represents the information of decorrelation (otherwise called "correlation index") between the front channel and the rear channel of the same left side and is part of the spatialization parameters transmitted by the encoder according to the draft MPEG Surround standard indicated above. As described above, the term σ _BL represents the target energy of the left rear channel when it is a question of determining the coefficient α to calculate the filter h _{L, L} (α = α ₁ ). Of course, an equivalent expression can be applied to calculate the weighting coefficient α involved in the homologous filter h _{R, R} specific to direct acoustic paths to the right ear. Nevertheless, for the filters h _{L, R} and h _{R, L} specific to the crossed paths, for example for the filter h _{L, R} specific to the paths crossed to the left ear, the coefficient α = α ₂ may preferentially be written: $${\mathrm{<figure-callout\; id="2"\; label="\alpha "\; filenames="imgf0001.png"\; state="\{\{state\}\}">\alpha </figure-callout>}}_2=\mathrm{abs}\left({\mathrm{ICC}}_{\mathrm{R}}\right),\mathrm{if\; abs}\left({\mathrm{ICC}}_{\mathrm{R}}\right)>{{\mathrm{<figure-callout\; id="2"\; label="\sigma \; BR"\; filenames="imgf0001.png"\; state="\{\{state\}\}">\sigma \; </figure-callout>}}_{\mathrm{BR}}}^2,$$

$${\mathrm{<figure-callout\; id="2"\; label="\alpha "\; filenames="imgf0001.png"\; state="\{\{state\}\}">\alpha </figure-callout>}}_2={{\mathrm{<figure-callout\; id="2"\; label="\sigma \; BR"\; filenames="imgf0001.png"\; state="\{\{state\}\}">\sigma \; </figure-callout>}}_{\mathrm{BR}}}^2,\mathrm{if\; not},$$

the term σ _BR representing the target energy of the right rear channel and the term ICC _R representing the correlation index between the right front channel and the right rear channel.
It will be noted that the "sqrt" function no longer applies for the crossed paths and for the calculation of the corresponding coefficient α ₂ , in the example described. Indeed, the target energies and the correlation indices are terms between 0 and 1 so that the coefficient α ₂ is generally lower than the coefficient α ₁ .

The global filter combination, for the L-BIN pathway, has many groupings of HRTF functions forming filters h _{L, L} and h _{L, R} obtained by the formulas given above, and in each grouping, the HRTF function is well involved. a front speaker, the HRTF function of a rear speaker and a decorrelated version of this Last function HRTF, which allows to represent a decorrelation between the front and back channels directly in the combination of filters, and thus directly in the binaural synthesis.

Recall that, since the sound data L, R (or M) are encoded in compression in a transformed domain, the combination of filters can be directly applied in the transformed domain as a function of the target energies (σ _FL , σ _BL , σ _FR , σ _BR ) associated with the channels of the multichannel format, these target energies being determined from the SPAT spatialization parameters. In this embodiment, of course, the transformation from the transformed domain to the time domain is then again planned for the actual reproduction in binaural context (TRANS modules). Figures 6A and 6B ).

• une entrée E pour recevoir les canaux compressés et les paramètres de spatialisation,
• une mémoire de travail MEM et un processeur PROC pour construire les combinaisons de filtres à partir des paramètres SPAT et appliquer ces combinaisons respectivement aux canaux compressés L et R,
• et une sortie S pour délivrer les signaux compressés et filtrés pour une restitution binaurale spatialisée sur les voies de restitution respectives L-BIN et R-BIN.

The present invention also aims at a decoding module DECOD BIN as represented by way of example on the figure 7 , for a three-dimensional spatialized rendering on two L-BIN and R-BIN rendering channels, and comprising in particular sound data processing means (compressed channels L, possibly R in stereophonic mode and SPAT spatialization parameters) for the implementation of the method described above. These means can typically comprise:

• an input E to receive the compressed channels and the spatialization parameters,
• a working memory MEM and a processor PROC for constructing the filter combinations from the parameters SPAT and applying these combinations respectively to the compressed channels L and R,
• and an output S for delivering the compressed and filtered signals for binaural spatialized reproduction on the respective L-BIN and R-BIN restitution channels.

The present invention also aims at a computer program, intended to be stored in a memory of a decoding module, such that the MEM memory of the DECOD-BIN module of the figure 7 , for spatialized three-dimensional reproduction on two L-BIN and R-BIN rendering channels. The program then comprises instructions for executing the method according to the invention and, in particular, to construct the filter combinations integrating the uncorrelated versions as illustrated on the Figures 6A and 6B described above. As such, one or the other of these figures can constitute a flowchart representing the algorithm underlying the program.

Claims (10)

1. Method of processing sound data for three-dimensional spatialized reproduction on two reproduction tracks for the respective ears of a listener,
   the sound data being initially represented in a multichannel format and then compression encoded (ENCOD) on a reduced number of channels (L,R),
   said multichannel format consisting in providing for more than two channels able to feed respective loudspeakers,
   the method comprising the steps:

   - obtaining, with the data compressed on said reduced number of channels, spatialization parameters (SPAT),

   - forming, on the basis of said spatialization parameters, for each reproduction track associated with an ear of the listener, a combination of filters each representative of transfer functions (HRTF) between this ear of the listener and loudspeakers able to be fed by respective channels of the initial multichannel format, and

   - applying to the compressed data the filter combination (h_L,_L, h_L,R, h_L,C: h_R,R, h_R,_L, h_R,C) associated with each reproduction track (L-BIN; R-BIN),

   characterized in that the method furthermore comprises the steps:

   - determining on the basis of said spatialization parameters, for each reproduction track associated with an ear of the listener, at least one transfer function of a loudspeaker situated at the rear of the ear of the listener and representative of a decorrelation between the channels of the multichannel format which are respectively associated with the rear loudspeaker and with at least one loudspeaker situated in front of the ear of the listener, and

   - integrating, for each reproduction track, said transfer function representative of a decorrelation in said filter combination associated with this reproduction track.
2. Method according to Claim 1, characterized in that the filter combination associated with a reproduction track (L-BIN) comprises at least one first grouping, forming a first filter (h_L,L), on the basis:

   - of the transfer function of a front loudspeaker (HRTF-A),

   - of the transfer function of a rear loudspeaker (HRTF-C), and

   - of a version (HRTF-C*) of the transfer function of the rear loudspeaker, representative of a decorrelation between channels,

   and in that the front and rear loudspeakers are situated on one and the same first side with respect to the listener.
3. Method according to Claim 2, characterized in that said grouping comprises a weighting, according to a chosen coefficient (α1; α2), between:

   - the transfer function of the loudspeaker situated at the rear, and

   - the version representative of a decorrelation of this transfer function of the rear loudspeaker.
4. Method according to Claim 3, characterized in that the compression encoding implements a parametric coder (ENCOD) delivering an information cue regarding the decorrelation between channels of the multichannel format, and in that the weighting coefficient is represented by a function that can vary dynamically as a function of the decorrelation information cue (ICC_L; ICC_R) delivered by the parametric coder.
5. Method according to one of Claims 2 to 4, in which the sound data are compression encoded on two channels (L,R),
   characterized in that the filter combination associated with said reproduction track (L-BIN) comprises, in addition to said first filter-forming grouping (h_L,_L) of one of the compressed channels (L), a second filter-forming grouping (h_L,R) of the other of the compressed channels (R) on the basis:

   - of the transfer function of a front loudspeaker (HRTF-H) situated on a second side, opposite from the first side with respect to the listener,

   - of the transfer function of a rear loudspeaker (HRTF-E) situated on said second side, and

   - of a version (HRTF-E*) of the transfer function of this rear loudspeaker, representative of a decorrelation between channels.
6. Method according to one of the preceding claims, characterized in that, the sound data being compression encoded in a transformed domain, the combination of filters is applied in the transformed domain as a function of target energies associated with the channels of the multichannel format, these target energies being determined on the basis of said spatialization parameters.
7. Method according to one of the preceding claims, characterized in that said transfer functions of the loudspeakers are of HRTF type and represent acoustic disturbances on paths between each loudspeaker and an ear for a reproduction track associated with this ear.
8. Method according to Claims 6 and 7, in which the transformed domain is the sub-band domain, characterized in that the decorrelated versions of the HRTF functions of the rear loudspeakers are obtained by applying to the initial HRTF functions of the rear loudspeakers a phase shift which is dependent on each sub-band of frequencies.
9. Decoding module (DECOD BIN) for three-dimensional spatialized reproduction on two reproduction tracks, characterized in that it comprises means for processing sound data for the implementation of the method according to one of the preceding claims.
10. Computer program, intended to be stored in a memory of a decoding module for three-dimensional spatialized reproduction on two reproduction tracks, characterized in that it comprises instructions for the execution of the method according to one of Claims 1 to 8.

Images