EP2957110B1 - Method and device for generating feed signals intended for a sound restitution system - Google Patents

Description

Background of the invention

The invention relates to the general field of acoustic processing and spatialization of sounds.

It relates more particularly to a method of sound reproduction based on a transaural technique.

The invention applies for example in a restricted space, such as for example in a passenger compartment of a car or of another vehicle (eg an airplane) equipped with at least two pairs of loudspeakers, in order to reproduce a wide stereophonic sound scene (ie spatialized) around two known listening positions to which listeners are likely to find themselves. In the example of the passenger compartment of the car, these two positions are for example the position of the driver of the car and that of his passenger.

The invention is however not limited to this single application, and can easily be applied to other sound environments having a similar speaker configuration and in which two different listening positions are aimed.

In the state of the art, there exist binaural sound reproduction techniques which rely on the physiological characteristics of human hearing, via the use of transfer functions also called HTRF (for Head Related Transfer Functions) which translate the set of spatial acoustic indices received at the level of a listener's eardrums for a source emitted from a given point in space.

These binaural techniques make it possible to obtain a sound spatialization very close to natural listening. However, they are limited to a sound reproduction via a headset.

Transaural techniques aim to adapt binaural synthesis to a speaker device.

In the simple case of a two-speaker device used to render a stereophonic signal to a single listener, these techniques are used to broadcast, via the two speakers of the device, the right binaural signal only on the right ear of the listener, and the left binaural signal only on his left ear. For this purpose, unwanted signals between the loudspeakers and the ears of the listener are eliminated via so-called cross-cycle cancellation techniques (CTCs), also known as datalogging techniques. echo cancellation. In other words, the sound signal emitted by the right speaker on the left ear, and the sound signal emitted by the left speaker on the right ear are eliminated or at least reduced by an appropriate signal processing.

The document J. Bauck et al., "Generalized Transaural Stereo and Applications," Journal of Audio Engineering Society, Vol. 44 No. 9, Sept. 1996 (hereinafter D1), described a generalized transaural technique that can be applied to any number of loudspeakers and listeners.

The figure 1 illustrates the system envisaged in D1: N sound programs (ie, N multichannel sound signals) are used to generate M signals supplying respectively M loudspeakers of a playback system intended for L / 2 listeners (in other words, L human ears) . The processing for generating the M power signals of the loudspeakers from the N sound programs is a cross-path cancellation process.

• Z désigne la matrice acoustique de transfert entre les N programmes sonores d'origine et les L signaux reçus respectivement par les oreilles des auditeurs,
• Y désigne la matrice de traitement CTC entre les N programmes sonores d'origine et les M signaux d'alimentation des M haut-parleurs (Y représente les filtres transauraux), et
• X désigne la matrice de propagation acoustique entre les M haut-parleurs et les L oreilles des auditeurs,

l'équation matricielle suivante doit être résolue : $$Z=\mathit{XY}$$

This system can be modeled in matrix form. So, if:

• Z denotes the acoustic transfer matrix between the N original sound programs and the L signals received respectively by the listeners' ears,
• Y denotes the CTC processing matrix between the N original sound programs and the M feed signals of the M loudspeakers (Y represents the transaural filters), and
• X denotes the acoustic propagation matrix between the M loudspeakers and the ears of the listeners,

the following matrix equation must be solved: $$Z=\mathit{XY}$$

In this matrix equation, the matrices X and Z are known: in fact, the matrix Z characterizes the desired sound reproduction at the level of the listeners while the matrix X is measurable. We therefore seek a processing matrix Y satisfying equation (1).

où ^H définit l'opérateur hermitien.The matrix X is of dimensions LxM: it is not necessarily square. Moreover, this matrix is not always well conditioned. It is not always invertible. Under such conditions, it is customary to use the pseudo-inverse matrix of the X matrix, denoted by X ⁺ , and defined by: $$X^{+}={\left(X^{H}X\right)}^{-1}{X}^H$$

where ^H defines the Hermitian operator.

The solution to equation (1) is then given by: $$Y=X^{+}Z$$

The technique proposed in D1 is general and is theoretically suitable for any values of N, M and L.

However, this technique only works in a specific geometric configuration of loudspeakers and listeners' positions, and is particularly sensitive to the acoustic environment of the listeners. It also requires the inversion of the acoustic matrix X, which for many reasons, as mentioned above, can be problematic (in the event of poor packaging of the matrix), or even impossible for certain frequency bands of the audible band. .

The document EP 1 370 115 A2 describes a device similar to that of D1 for rendering a signal to two listeners simultaneously using four loudspeakers.

The document US 6,956,954 B1 describes a device applying the same signal resulting from a binaural processing with two loudspeakers situated symmetrically on either side of a central loudspeaker, to achieve symmetrical spatialization effects at two different listening points using three speakers.

There is a need for a method of spatially reproducing a sound signal intended for two given spatial positions, and via a feedback system equipped with least two pairs of speakers, which is robust and does not have the aforementioned drawbacks of the state of the art.

Object and summary of the invention

The invention responds in particular to this need by proposing a method of generating power supply signals intended for a sound reproduction system comprising two sets of loudspeakers situated on either side of a separation plane, these signaling signals. power supply being generated for sound reproduction of a multichannel sound signal to two spatial listening positions, each listening space position being associated with one of the speaker sets and being located on a parallel plane to the separation plane so that at least one loudspeaker of this set is positioned on either side of this parallel plane, the generation method comprising, for each set of loudspeakers, a generating step, to from an input signal derived from the multichannel sound signal for this set, supply signals for supplying the loudspeakers of this set, this generation step comprising the applicati one of a technique of transaural restitution to the input signal, this technique creating a virtual sound source for the listening position associated with the set on a first line obtained by symmetry, with respect to the parallel plane on which is located the listening space position associated with this set of a second straight connecting the listening space position to the speaker more remote from this position belonging to the other set of speakers of the playback system.

Correlatively, the invention also provides a device for generating supply signals intended for a sound reproduction system comprising two sets of loudspeakers located on either side of a separation plane, said signals being generated for a sound reproduction of a multichannel sound signal to two spatial listening positions, each listening spatial position being associated with one of the speaker sets and being located on a plane parallel to the separation plane so that at least one loudspeaker of this set is positioned on either side of this parallel plane, this generation device comprising means able to generate for each set of loudspeakers, from a signal of input derived from the multichannel sound signal for this set, power signals for supplying the speakers of this set, these means being able to apply a technique of restitution tra nsaural to the input signal, this technique creating a virtual sound source for the listening position associated with the set on a first straight line obtained by symmetry, with respect to the parallel plane on which is located the listening spatial position associated with this set of a second straight connecting the listening space position to the speaker more distant from this position belonging to the other set of speakers of the rendering system.

The loudspeakers of the two sets of speakers of the rendering system can be mounted in or on two disjoint physical entities, or alternatively, on the same physical entity but separated by a "virtual" vertical plane called separation.

They are oriented towards the listening positions with which they are associated, that is, each listening position is in the frontal area of the speakers of the set with which it is associated.

Thus, the invention allows a spatialized sound reproduction of a multichannel sound signal, from a speaker array comprising at least two sets of speakers, and simultaneously for two listening positions defined respectively with respect to these two sets of speakers.

These listening positions are not necessarily punctual: they may indeed be actual listening zones, in which case the listening position on a plane parallel to the separation plane means that the geometric center of this listening position is on the parallel plane. These listening spatial zones are preferentially defined by a radius around the geometric center of less than 15-20 cm.

Each set of speakers of the reproduction system envisaged for the sound reproduction of the multichannel signal thus comprises a loudspeaker on either side of the parallel plane on which the listening position is located. The listening position is for example centered with respect to the set of speakers to which it is attached.

To allow a spatialized sound reproduction, the invention advantageously proposes to treat separately each set of speakers and the listening position attached thereto.

Specifically, a separate transaural technique is used for each set of speakers to generate the speaker power signals of this set. This transaural technique is implemented according to the invention, so as to create a virtual source outside the speaker array, in the opposite direction with respect to the listening position associated with the set of loudspeakers. considered, to the speaker of the network farthest from this position (and belonging to the other set of speakers).

In this way, the invention combines, on the one hand, the advantages of a transaural reproduction applied to a small number of loudspeakers (resulting from the separate treatment performed on each set of loudspeakers), and on the other hand, the precedence principle or law of the first wavefront, well known to those skilled in the art. According to this principle, the location of a sound object is given by the direction from which the sound comes first to the ear (i.e. first wavefront).

To better illustrate the invention, consider a multichannel sound signal of the stereo type, which is to be restored via a frontal array of loudspeakers consisting of two sets of two loudspeakers, these sets being respectively facing two spatial positions. listening.

The invention returns, according to this example, to spatialize the left part of the sound stage (given by the left channel of the stereo signal) for the listening position to the left of the separation plane (called left listening position) and to spatialize the right part of the sound stage (given by the right channel of the stereo signal) for the listening position to the right of the separation plane, (the right listening position), the right part of the sound stage being then naturally given for the left listening position by the position of the set of speakers located to the right of it, and the left part of the sound stage being naturally given for the right listening position by the position of all speakers located to the left of it.

This results in a spatialized sound reproduction technique that is relatively simple to implement: the matrices to be reversed in the transaural techniques implemented according to the invention are in fact of reduced dimensions (2x2 matrices in the example envisaged above), since they are limited to each set of speakers. In this way, they suffer little or nothing from the disadvantages of the state of the art (matrices of large dimensions difficult to reverse and poorly conditioned). The rendering technique proposed by the invention is thus more robust to the environment of the rendering system.

Moreover, for a reproduction system configuration in which the loudspeakers of each set are slightly spaced apart (typically, for a maximum spacing between the speakers of 70 cm), the sound reproduction proposed by the invention is less sensitive to the so-called "sweet-spot" phenomenon that the sound reproduction proposed in D1.

In known manner, the "sweet-spot" defines an ideal listening area for which the sound reproduction is designed and optimized: this ideal listening area generally corresponds to a single position in the center of the device.

In D1, when a listener moves outside this ideal listening area, he perceives not only a change in the spatial image of the sound signal rendered, but also a change of timbre (i.e. a coloration of the restored signal).

On the contrary in the invention, a listener placed at a listening position associated with a set of loudspeakers, and which is thereby outside the "sweet-spot" of the associated rendering technique to the other set of speakers and the other listening position, does not hear any unwanted spatialization effect. It only hears a coloring effect of the input signal reproduced by the other set of speakers, but it is perceived on distant speakers (ie on the speakers of the other set of loudspeakers). speakers).

In a particular embodiment of the invention, the loudspeakers of the two sets of restitution are substantially aligned on a horizontal axis of the rendering system.

Such a configuration of speakers has a preferred application in a small space such as a car interior for example.

In this particular embodiment, during the generation step, the virtual source is preferentially created at the intersection of the first line and the horizontal axis of the rendering system, so as to optimize the spatial reproduction of the signal. multichannel sound.

In a particular embodiment, the generation method further comprises a spectral equalization step during which, for each supply signal intended to supply a loudspeaker, an equalization filter adapted to this particular signal is applied individually. loud speaker.

This spectral equalization advantageously makes it possible to take into account the environment in which the rendering system is located.

In general, and more specifically when the rendering system is in a closed environment such as for example the passenger compartment of a car, it is to be expected that the timbre of the multichannel sound signal reproduced will not be identical to the two listening positions. Spectral equalization can be implemented to correct this effect. At the same time, for each listening position, it corrects the staining effect due to the transaural technique applied to the set of speakers opposite to the listening position.

Different criteria can therefore be envisaged to derive these equalization filters.

According to a first criterion, each equalization filter adapted to a loudspeaker is chosen so as to obtain an amplitude spectrum of a frequency response established for this loudspeaker from measurements made in a soundproof environment, in accordance with a first predetermined target spectrum (eg flat amplitude spectrum).

The application of this first criterion aims to individually correct the defects of each speaker, and reduce the differences that may appear between them. The equalization aims to harmonize the amplitude spectra of the frequency responses of the different loudspeakers.

According to a second variant, the equalization filters are chosen so as to obtain an average amplitude spectrum evaluated from the impulse responses of the loudspeakers of the sets of the rendering system, according to a second predetermined target spectrum.

This second criterion aims at a global average equalization of all the power supply signals of the loudspeakers, for example to improve the tone (or coloration) of the restored signal.

In a particular embodiment, the generation method further comprises a step of spatial analysis of the multichannel sound signal comprising the extraction of at least one sound object and the determination of a spatial position of this sound object, each signal input device used to generate the speaker power signals of a set of loudspeakers being developed from at least one sound object extracted during this spatial analysis step and selected according to its position space relative to the separation plane.

By means of this spatial analysis, the invention makes it possible to process any multichannel sound signal and to optimize the rendering of this sound signal, by extracting sound objects from this system.

In another embodiment, the rendering system further comprises a third set of loudspeakers, at least one loudspeaker of this third set being placed on either side of the separation plane.

This third set of speakers makes it possible to reproduce the sound objects of the multichannel signal whose spatial position determined during the spatial analysis step is in the center.

In a preferred embodiment, each set of speakers of the rendering system comprises a pair of loudspeakers.

In a particular embodiment, the various steps of the generation method are determined by instructions of computer programs or microprocessors.

Consequently, the invention also relates to a computer or microprocessor program on an information carrier, this program being capable of being implemented in a generation device or more generally in a computer or by a microprocessor, this program comprising instructions adapted to the implementation of the steps of a generation method as described above.

This program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code, such as in a partially compiled form, or in any other form desirable shape.

The invention also relates to a computer-readable information medium, comprising instructions of a computer program as mentioned above.

The information carrier may be any entity or device capable of storing the program. For example, the medium may comprise storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or a magnetic recording medium, for example a diskette (floppy disc) or a disk hard.

On the other hand, the information medium may be a transmissible medium such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention can be downloaded in particular on an Internet type network.

Alternatively, the information carrier may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

• au moins deux ensembles de haut-parleurs ; et
• un dispositif de génération de signaux d'alimentation destinés aux haut-parleurs desdits deux ensembles de haut-parleurs selon l'invention.

According to another aspect, the invention also provides a sound reproduction system comprising:

• at least two sets of speakers; and
• a device for generating supply signals for the loudspeakers of said two sets of loudspeakers according to the invention.

In another embodiment, the sound reproduction system comprises at least two sets of loudspeakers located in a front plane with respect to the listening positions and at least two sets of speakers located in a rear plane with respect to the listening positions.

This embodiment is particularly well suited to the rendering of multichannel signals comprising rear signals, for example signals having a 5.1 format.

It can also be envisaged, in other embodiments, that the generation method, the generation device and the system according to the invention present in combination all or part of the aforementioned characteristics.

Brief description of the drawings

• la figure 1 déjà décrite représente, de façon schématique, une technique de restitution transaurale généralisée de l'état de la technique ;
• la figure 2 représente un système de restitution conforme à l'invention comprenant un dispositif de génération de signaux d'alimentation destinés à des ensembles de haut-parleurs, conforme à l'invention ;
• la figure 3 illustre l'architecture matérielle du dispositif de génération de la figure 2 ;
• la figure 4 représente un système de restitution et un dispositif de génération conformes à l'invention dans un premier mode particulier de réalisation de l'invention ;
• la figure 5 illustre les principales étapes du procédé de génération selon l'invention telles qu'elles sont mises en oeuvre par le dispositif de génération de la figure 4 dans le premier mode de réalisation ;
• les figures 6A et 6B illustrent comment peuvent être déterminés les filtres d'égalisation pouvant être utilisés lors du procédé de génération représenté à la figure 5 ;
• la figure 7 illustre une variante du premier mode particulier de réalisation de l'invention ;
• la figure 8 représente un système de restitution et un dispositif de génération conformes à l'invention dans un deuxième mode particulier de réalisation de l'invention ;
• la figure 9 illustre les principales étapes du procédé de génération selon l'invention telles qu'elles sont mises en oeuvre par le dispositif de génération de la figure 8 dans le deuxième mode de réalisation ; et
• la figure 10 représente un système de restitution et un dispositif de génération conformes à l'invention dans un troisième mode particulier de réalisation de l'invention.

Other features and advantages of the present invention will emerge from the description given below, with reference to the accompanying drawings which illustrate embodiments having no limiting character. In the figures:

• the figure 1 already described represents, schematically, a technique of generalized transaural restitution of the state of the art;
• the figure 2 represents a rendering system according to the invention comprising a device for generating supply signals for sets of loudspeakers, according to the invention;
• the figure 3 illustrates the hardware architecture of the device generating the figure 2 ;
• the figure 4 represents a reproduction system and a generation device according to the invention in a first particular embodiment of the invention;
• the figure 5 illustrates the main steps of the generation process according to the invention as they are implemented by the device for generating the figure 4 in the first embodiment;
• the Figures 6A and 6B illustrate how the equalization filters that can be used in the generation process shown in FIG. figure 5 ;
• the figure 7 illustrates a variant of the first particular embodiment of the invention;
• the figure 8 represents a reproduction system and a generation device according to the invention in a second particular embodiment of the invention;
• the figure 9 illustrates the main steps of the generation process according to the invention as they are implemented by the device for generating the figure 8 in the second embodiment; and
• the figure 10 represents a reproduction system and a generation device according to the invention in a third particular embodiment of the invention.

Detailed description of the invention

We will now describe, with reference to Figures 2 to 10 , three particular embodiments of the invention.

In these three embodiments, it is envisaged, as illustrated schematically in FIG. figure 2 , the reproduction of a multi-channel sound signal S via a reproduction system 1 according to the invention, comprising a plurality of sets of loudspeakers E1,..., EN, (N greater than or equal to 2), and a device 2 for generating power signals to these sets of loudspeakers, according to the invention.

Each set of loudspeakers includes at least one pair of loudspeakers and is associated with a determined listening position, this listening position being centered here with respect to this set of loudspeakers, so that at least a loudspeaker of the set is located on both sides of the listening position.

More generally, the loudspeakers of each set are positioned on either side of a vertical plane passing through the listening position and parallel to a vertical separation plane situated between the two sets of loudspeakers.

However, the assumption that each listening position is centered relative to the set of speakers with which it is associated is not limiting in itself: the speakers of a set of speakers can be arranged asymmetrically with respect to the listening position.

The sets of speakers E1, ..., EN, can be mounted on separate physical entities or on the contrary on the same entity while delimiting different areas of space. Similarly, the generation device 2 is not necessarily located on the same physical entity as the speaker sets.

In the embodiments described here, the generation device 2 according to the invention has the hardware architecture of a computer, as schematically illustrated in FIG. figure 3 .

It comprises in particular a processor 3, a read-only memory 4, a random access memory 5, a non-volatile memory 6 and communication means 7, on the one hand, with a source supplying it with the multichannel sound signal S (not shown on FIG. figure 2 and on the other hand, with the loudspeakers of the loudspeaker assemblies E1,..., EN to which the generating device 2 supplies the power signals which it has generated. These communication means may be for example wired communication means, or alternatively, wireless communication means using WiFI (Wireless FIdelity) for example or Bluetooth ™.

The read-only memory 4 of the generation device 2 constitutes a recording medium in accordance with the invention, readable by the processor 3 and on which is recorded a computer program according to the invention, comprising instructions for the execution of the steps of a generation method according to the invention described later with reference to the Figures 5 and 9 , in different particular embodiments. This computer program defines, in an equivalent way, functional modules of the generation device 2 (ie module of generation, generation module and if appropriate sound signal analysis module S and equalization).

For the sake of simplicity and clarity in the remainder of the description, the similar elements are referenced using identical reference numerals indexed by "," or "" according to the embodiment envisaged.

We will now describe with reference to Figures 4 to 7 a first particular embodiment of the invention.

To illustrate this first embodiment, it is contemplated figure 4 a system of reproduction 1 'of a stereo signal S' comprising two sets of loudspeakers E1 'and E2', each set comprising a pair of loudspeakers (H11 ', H12' for the set E1 'and H21' , H22 'for the set E2').

The two sets E1 'and E2' are located on either side of a vertical separation plane D '. In the example shown in figure 4 , the vertical separation plane D 'defines a plane of symmetry for the two sets of speakers E1' and E2 '.

The loudspeakers H11 ', H12', H21 ', H22' are, in the example illustrated in FIG. figure 4 , aligned or substantially aligned on the same horizontal axis Δ 'of the rendering system.

The invention is however not limited to such an aligned configuration of the loudspeakers. Thus, by way of illustration, the loudspeakers H11 'and H22' may be on the axis Δ 'while H12' is positioned above the axis Δ 'and H21' below, and so on.

Each set of speakers E1 ', E2' is positioned facing a privileged listening position P1 ', P2'. These listening positions are located on vertical planes L1 'and L2' respectively, which are parallel to the separation plane D '.

It should be noted that the listening positions P1 'and P2' are not necessarily one-off. In the sense of the invention P1 'and P2' are located on a plane parallel to the separation plane since their geometric center is located on this parallel plane.

According to the example shown in figure 4 the listening position P1 'is centered with respect to the set of loudspeakers E1', in other words, the plane L1 'is a median plane of the set of loudspeakers E1', orthogonal to the axis Δ.

Similarly, the listening position P2 'is centered relative to the set of speakers E2', in other words, the plane L2 'is a median plane of the set of loudspeakers E2', orthogonal to the speaker. 'axis Δ'.

It is assumed that the listening positions P1 'and P2' are occupied by two listeners.

Thus, according to this configuration, the listeners (i.e. listening positions) and the loudspeakers are placed symmetrically with respect to the separation plane D '.

However, this hypothesis is not limiting and it can be envisaged alternatively that the two sets of speakers E1 'and E2' are placed asymmetrically with respect to the separation plane D '.

In the example envisaged at figure 4 , the signal S 'is a stereo signal composed of two left and right channels, denoted respectively SL' and SR '.

We will now describe with reference to the figure 5 , the main steps of the generation method according to the invention, in the first embodiment, when it is implemented by the generating device 2 'for supplying power signals to the loudspeakers of the sets E1' and E2.

On receiving the sound signal S ', the generating device 2' implements a spatial analysis of this signal (step E10), in order to identify the components (ie sound objects) of this signal intended to be restored on the different sets of speakers.

The sound signal S 'being a stereo signal in the example envisaged at the figure 4 , this spatial analysis is limited here to associate the first set of speakers E1 'located to the left of the separation plane D' in the sound reproduction space, the left channel SL 'of the stereo signal, and the second set of high speakers E2 'located to the right of the separation plane D' in the sound reproduction space, the right channel SR 'of the stereo signal. The signals SL 'and SR' are input signals in the sense of the invention for the set E1 'and for the set E2' respectively.

Then according to the invention, the generation device 2 'applies two different transaural techniques to each of the input signals SL' and SR '. Each set of speakers E1 'and E2' is thus treated separately according to the invention.

More precisely, during the step E20, the generation device 2 'applies a first transaural technique T1' to the input signal SL 'so as to create a virtual source V1' in a direction symmetrical with respect to the plane L1 ' passing through the listening position P1 'to the direction of the speaker of the set E2' furthest from the listening position P1 ', in other words to the direction of the loudspeaker H22'.

This virtual source V1 'is created, in the example envisaged at the figure 4 at the intersection of the axis Δ 'and of the straight line d11' (first line within the meaning of the invention) passing through the listening position P1 'and symmetrical with the straight line d12' (second straight line in the sense of the invention) connecting the listening position P1 'to the loudspeaker H22'. In other words, in the example of figure 4 , V1 'is the symmetrical (in distance and direction) of the loudspeaker H22' with respect to the plane L1 '.

In this way, the listener at the listening position P1 'has a sound stage extending between the direction given by the virtual source V1' (that is to say by the line d11 ') and the direction given by the loudspeaker H22 '(i.e., by the line d12').

According to another configuration of loudspeakers in which the loudspeakers of the sets E1 'and E2' are not aligned on the same axis Δ ', the virtual source V1' can also be created on the right d11 'so as to be the symmetrical direction and distance of the speaker H22 'relative to the plane L1'.

However, it should be noted that the implementation of the invention does not require that a symmetry in terms of distance is imperatively respected. It is enough that the source virtual V1 'is created on the right d11', that is to say in a direction symmetrical to the direction d12 'relative to the plane L1', the speakers are aligned or not on the same axis.

Step E20 is implemented by the generation device 2 'by filtering the signal SL' by transaural filters determined so as to create the virtual source V1 '. In other words, these transaural filters are defined in such a way that the listener at the listening position P1 'receives on his two ears two binaural signals which define the virtual source V1'.

• Z1' est une matrice acoustique de transfert entre le signal d'entrée SL' et les deux signaux reçus respectivement par les oreilles de l'auditeur situé à la position d'écoute P1' (Z1' est en fait un vecteur à deux composantes dans l'exemple envisagé ici),
• Y1' est une matrice de traitement transaural entre le signal d'entrée SL' et les deux signaux d'alimentation des haut-parleurs H11' et H12' (Y1' représente les filtres transauraux et est également un vecteur à deux composantes dans l'exemple envisagé ici), et
• X1' désigne une matrice de propagation acoustique entre les deux haut-parleurs H11' et H12' et les oreilles de l'auditeur placé à la position d'écoute P1' (X1' est de dimension 2x2 ici),

la matrice Y1' contenant les filtres transauraux appliqués au signal SL' vérifie l'équation matricielle suivante : $$\mathrm{<figure-callout\; id="1"\; label="Z"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">Z</figure-callout>}1\prime =\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">X</figure-callout>}1\prime \mathrm{<figure-callout\; id="1"\; label="Y"\; filenames="imgf0001.png,imgf0002.png"\; state="\{\{state\}\}">Y</figure-callout>}1\prime$$

More specifically, if we assume, by repeating the notations previously introduced with reference to the figure 1 , than :

• Z1 'is an acoustic transfer matrix between the input signal SL' and the two signals received respectively by the ears of the listener at the listening position P1 '(Z1' is in fact a two-component vector in the example envisaged here),
• Y1 'is a transaural processing matrix between the input signal SL' and the two power supply signals of the loudspeakers H11 'and H12' (Y1 'represents the transaural filters and is also a two-component vector in the example considered here), and
• X1 'denotes an acoustic propagation matrix between the two loudspeakers H11' and H12 'and the ears of the listener placed at the listening position P1' (X1 'is of dimension 2x2 here),

the matrix Y1 'containing the transaural filters applied to the signal SL' satisfies the following matrix equation: $$Z1\text{'}=X1\text{'}Y1\text{'}$$

In other words, the transaural filters applied to the signal SL 'are obtained by inversion of the matrix X1' and multiplication of the inverse matrix thus obtained by the matrix Z1 '.

As mentioned previously, the matrices Z1 'and X1' are known: Z1 'contains the binaural filters which correspond to the creation of the virtual source V1', and X1 'contains the impulse response of the loudspeakers H11' and H12 '.

The determination of the binaural filters for the creation of the virtual source V1 'does not pose any difficulty in itself for the skilled person and will not be described further here. We can refer in particular to the document WO 95/23493 .

• XI' peut être mesurée directement aux oreilles de l'auditeur en posant des microphones binauraux au niveau de ses oreilles ;
• X1' peut être simulée en utilisant une base de données de fonctions HRTF et en extrayant les HTRF dans la direction des haut-parleurs H11' et H12' par rapport à l'auditeur placé à la position d'écoute P1' ; ou
• X1' peut être simulée grâce à des fonctions HTRF correspondant à un modèle de tête sphérique bien connues de l'homme du métier.

The matrix X1 'is composed of four acoustic paths, namely the path between the loudspeaker H11' and the left ear of the listener placed at P1 ', the path between the loudspeaker H11' and the loudspeaker H11 '. the right ear of the listener, the path between loudspeaker H12 'and the listener's left ear and finally the path between speaker H12' and the right ear of the listener. It can be determined in different ways. For example, as examples:

• XI 'can be measured directly in the ears of the listener by placing binaural microphones at the level of his ears;
• X1 'can be simulated using a database of HRTF functions and extracting the HTRFs in the direction of the loudspeakers H11' and H12 'relative to the listener at the listening position P1'; or
• X1 'can be simulated by HTRF functions corresponding to a spherical head model well known to those skilled in the art.

The application of the transaural filters of the matrix Y1 'thus determined to the signal SL' results in two supply signals S11 'and S12' intended to supply respectively the two loudspeakers H11 'and H12' of the assembly E1 '.

In the first embodiment described herein, these power signals S11 'and S12' are equalized by means of individual equalizing filters F1 ', F2' before being supplied to the loudspeakers H11 'and H22' (step E30). This spectral equalization step is intended to compensate for the possible defects of the speakers H11 'and H12' (for example, the timbres of the speakers), and to reduce the differences that may exist between these different speakers.

The equalization filters F1 'and F2' respectively applied to the signals S11 'and S12' are of course predetermined during a preliminary step, and stored in the nonvolatile memory of the generation device 2 '.

One method that can be implemented during this preliminary step to determine the filters F1 'and F2' is as follows.

Initially, measurements are made for each loudspeaker H11 'and H12' in a soundproof environment (ie treated acoustically) with an omnidirectional microphone placed in front of each loudspeaker and known reference signals (eg white noise ). These measurements make it possible to obtain the impulse responses of each loudspeaker H11 'and H12'.

Then for each loudspeaker, a time average of the different impulse responses measured for this loudspeaker is obtained so as to obtain a mean impulse response.

The average impulse response is then transformed into a frequency response, in a manner known per se, for example by means of a Fourier transformation. This average frequency response is then optionally smoothed by frequency band (eg octave bands or auditory bands).

Then the amplitude spectrum of this frequency response is determined.

The equalization filter is then chosen so as to obtain an amplitude spectrum of the optionally smoothed frequency response in accordance with a predetermined target amplitude spectrum SPcible (first target spectrum in the sense of the invention). This target amplitude spectrum is for example a flat spectrum, that is to say constant whatever the frequency.

In other words, if SP1 'and SP2' respectively designate the amplitude spectrum of the frequency response thus determined of the loudspeaker H11 'and of the loudspeaker H12' of the set E1 ', the equalization filter F1 ', respectively F2', is chosen so as to minimize the difference between SP1 'and SPcible, respectively between SP2' and SPCible, within a limit for example of plus or minus 5dB.

This ensures that, on average, the amplitude spectrum of the frequency response of each speaker of the set of speakers E1 is identical or similar to the amplitude spectrum SPCible and therefore between them. It should be noted that in the example considered here, only the amplitude of the impulse response of each speaker is corrected, it is not interested in the phase itself.

The filtering of the signal S11 'by the filter F1' results in a filtered feed signal S11f ', which is supplied by the device 2' to the loudspeaker H11 'for reproduction (step E40).

Similarly, the filtering of the signal S12 'by the filter F2' results in a filtered feed signal S12f ', which is supplied by the device 2' to the loudspeaker H12 'for reproduction (step E40).

Similarly, and in parallel with steps E20-E40, the generation device 2 'applies a second transaural technique T2' to the input signal SR 'so as to create a virtual source V2' in a direction symmetrical with respect to the plane L2 'passing through the listening position P2' to the speaker direction of the set E1 'furthest from the listening position P2', ie the speaker H11 '(step E50)

This virtual source V2 'is created, in the example envisaged at the figure 4 at the intersection of the axis Δ 'and of the straight line d22' (first straight line within the meaning of the invention) passing through the listening position P2 'and symmetrical with the straight line d12' (second right in the direction of the invention) connecting the listening position P2 'to the loudspeaker H11'.

In this way, the listener at the listening position P2 'has a sound stage extending between the direction given by the virtual source V2' (that is to say, by the line d22 ') and the direction given by the loudspeaker H11 '(i.e., by the line d12').

The step E50 is carried out by the generation device 2 'by filtering the signal SR' by transaural filters determined so as to create the virtual source V2 '. These transaural filters are determined similarly or identically to the transaural filters applied during step E20 for the creation of the virtual source V1 'and will therefore not be described again here.

The application of the transaural filters to the signal SR 'results in two supply signals S21' and S22 'intended to supply respectively the two loudspeakers H21' and H22 'of the set E2'.

In the first embodiment described here, these supply signals S21 'and S22' are equalized with the aid of individual equalization filters F3 ', F4' before being supplied to the loudspeakers H21 'and H22' (step E60). This spectral equalization step is intended to compensate for any speaker faults (for example, the tone of the speakers), and to reduce the differences that may exist between the different speakers (ie between H21 'and H22 ' right here).

The equalization filters F3 'and F4', respectively applied to the signals S21 'and S22' are predetermined during a preliminary step, similarly or identically to the equalization filters F1 'and F2', and are stored in the non-volatile memory of the generation device 2 '.

The filtering of the signal S21 'by the filter F3' results in a filtered feed signal S21f 'supplied to the loudspeaker H21' for reproduction (step E70).

Similarly, the filtering of the signal S22 'by the filter F4' results in a filtered feed signal S22f 'supplied to the loudspeaker H22' for reproduction (step E70).

In the first embodiment described here, the filters F1 ', F2', F3 ', F4' are determined so as to compensate for possible defects of the loudspeakers H11 ', H12', H21 'and H22' respectively, and to harmonize the timbre of these speakers.

In an alternative embodiment, it is possible to apply to the signals S11 ', S12', S21 'and S22' obtained after applying the transaural techniques T1 'and T2', spectral equalization filters (a filter for each loudspeaker). speaker) to improve the timbre or coloration of the system 1 'restitution as a whole, that is to say, including the two sets of speakers E1' and E2 '.

These filters can then be determined as follows.

At first, several measurements are made in the real environment in which the speakers of the reproduction system 1 '(eg passenger compartment of the car if the reproduction system 1' is included in such a cabin) are positioned. the impulse response of these loudspeakers on a network of microphones distributed on the target listening area and which includes the listening positions P1 'and P2' (that is, the target listening area is defined near the listening positions P1 'and P2'). The Figure 6A illustrates a network of four microphones m1, m2, m3 and m4 duly positioned for the system 1 '.

The impulse responses thus measured are averaged over all the speakers for each microphone so as to obtain an average impulse response for each microphone.

Then the average impulse responses are transformed into frequency responses using a Fourier transform for example.

Frequency smoothing of the frequency responses thus obtained can be achieved (for example by octave or by auditory band).

The frequency average of the amplitude spectra of the possibly smoothed frequency responses associated with the microphones is then evaluated. The average amplitude spectrum thus obtained thus includes the contribution of all the speakers of the reproduction system 1 'on all the microphones m1, m2, m3 and m4.

The equalization filters to be applied to the supply signals before they are supplied to the loudspeakers are then chosen so as to obtain a spectrum of average amplitude (in a tolerance for example of +/- 5dB) with a spectrum of predetermined target amplitude SPcible '(second frequency response in the sense of the invention). This target amplitude spectrum is for example a flat spectrum, as illustrated in FIG. Figure 6B , That is to say constant regardless of the frequency.

It should be noted that the filters making it possible to improve the timbre of the system 1 as a whole can be applied alternatively to the filters F1 ', F2', F3 ', F4' previously described, or in addition to these filters (on applies the two sets of filters successively or equivalent filters to the supply signals).

In the example envisaged at figure 4 the speaker assemblies E1 'and E2' each comprise a pair of loudspeakers. However, the invention can also be applied in a similar way to a different configuration of speaker assemblies E1 'and E2', as illustrated in FIG. figure 7 .

So, according to the example of the figure 7 each set of loudspeakers E1 'and E2' comprises three loudspeakers H11 ', H12', H13 'and H21', H22 ', H23' respectively.

The virtual source V1 'is then created at a position symmetrical with respect to the plane L1' of the position of the loudspeaker H23 ', whereas the virtual source V2' is created at a position symmetrical with respect to the plane L2 'of the position of the speaker H11 '.

We will now describe with reference to figures 8 and 9 , a second embodiment of the invention.

To illustrate this second embodiment, it is contemplated figure 8 a system for reproducing a stereo signal S "comprising two sets of loudspeakers E1" and E2 "arranged in a similar manner to the sets E1 'and E2' of the figure 4 , on both sides of a vertical separation plane D ".

Each set of speakers E1 ", E2" is placed facing a preferred listening position P1 ", P2" defined in a manner similar to that previously described for P1 'and P2', and comprises a pair of high speakers (H11 ", H12" for the set E1 "and H21", H22 "for the set E2"). P1 "and P2" are located respectively on vertical planes L1 "and L2" parallel to the separation plane D. "A loudspeaker of the assembly E1" is thus positioned on either side of L1 "and the listening position P1 ", and a speaker of the set E2" is positioned on either side of L2 "and the listening position P2".

The reproduction system 1 "further comprises a third set E3" of two loudspeakers H31 "and H32" placed in the center of the reproduction system 1 ", between the two sets E1" and E2 ", as illustrated in FIG. figure 8 . More precisely, the set of loudspeakers E3 "is positioned at the level of the separation plane D": at least one loudspeaker of the set E3 "is placed on either side of the separation plane D".

The speakers H11 ", H12", H21 ", H22", H31 "and H32" are in the example illustrated in FIG. figure 8 , Aligned or substantially aligned on the same horizontal axis Δ "of the rendering system However, as mentioned above, this assumption is not limiting and no limitation is strictly attached to the placement of the loudspeakers of the sets E1" , E2 "and E3" with respect to this horizontal axis.

We will now describe with reference to the figure 9 , the main steps of the generation method according to the invention, in the second embodiment, when it is implemented implemented by the generation device 2 "to supply power signals to the loudspeakers of the sets E1", E2 "and E3".

On reception of the sound signal S ", the generation device 2" implements a spatial analysis of this signal (step F10), in order to identify the components (ie sound objects) of this signal which will be restored on the different sets of speakers.

For this purpose, in the second embodiment described here, the analysis step F10 comprises a decomposition of the sound signal S "into several frequency sub-bands (eg in octave, or into auditory bands depending on the available processing power to implement the invention).

Then the device 2 "extracts on each frequency subband the sound objects contained in the signal S", and determines for each sound object, its spatial position.

It uses for this purpose techniques known to those skilled in the art, such as for example the determination of a Gerzon vector representative of the signal S "as described in the document US 2007/0269063 , or a decomposition in spherical harmonics as described in the document WO 2012/025580 .

The device 2 "then generates from the sound objects thus extracted three input signals denoted SL", SR "and SC" for the three sets of speakers E1 ", E2" and E3 ".

• les objets sonores se trouvant à gauche du plan de séparation D" sont alors intégrés dans le signal SL" dérivé pour l'ensemble E1",
• les objets sonores se trouvant à droite du plan de séparation D" sont alors intégrés dans le signal SR" dérivé pour l'ensemble E2", et
• les objets sonores se trouvant à proximité du plan de séparation D" sont alors intégrés dans le signal SC" dérivé pour l'ensemble E3".

The sound objects selected for each input signal depend here on their spatial positions with respect to the separation plane D "(and more generally predetermined analysis criteria which determine to which set of loudspeakers to send which sound object according to its spatial position) In the embodiment envisaged here:

• the sound objects to the left of the separation plane D "are then integrated in the signal SL" derived for the set E1 ",
• the sound objects to the right of the separation plane D "are then integrated in the signal SR" derived for the set E2 ", and
• the sound objects in the vicinity of the separation plane D "are then integrated in the signal SC" derived for the set E3 ".

Then according to the invention, the generation device 2 "applies two different transaural techniques to each of the input signals SL" and SR ", similarly to what has been described above with reference to the figure 5 . The steps F20 and F50 of application of a transaural technique resulting in the creation of two virtual sources V1 "and V2", the steps F30 and F60 of spectral equalization by the filters F1 ", F2", F3 "and F4" , and the steps F40 and F70 for generating the power supply signals to the loudspeakers of the assemblies E1 "and E2" being identical respectively to the steps E20 and E50, to the steps E30 and E60, and to the steps E40 and E70, they are not not described again here.

In the second embodiment described here, the generation device 2 "also applies a spatial rendering technique T3" to the input signal SC "in order to generate the power signals of the speakers H31 "and H32" of the third central speaker set E3 "(step F80).

This rendering technique is chosen for example so as to directly broadcast the signal SC "on the speakers H31" and H32 ".

The supply signals S31 "S32" generated by applying the technique T3 "to the input signal SC" are then filtered using equalization filters (which can be determined according to a method similar to that explained previously to determine the filters F1 ', F2', F3 'and F4') (step F90).

Then the filtered signals S31f "and S32f" are provided by the generating device 2 "to the loudspeakers H31" and H32 "respectively (step F100).

As a variant, it can be envisaged that the spatial reproduction technique T3 "applied to the input signal SC" generates supply signals for the loudspeakers of the three sets E1 ", E2" and E3 ", using a technique of holophonic rendering such as for example a WFS technique (for Wave Field Synthesis) Thus, by means of this technique, one can for example create a virtual source in the center, placed on the separation plane D "at a distance far enough from the listening positions in terms of wavefront (typically at a distance greater than 10m), so that the listeners placed at the listening positions P1 "and P2" perceive the center of signals restored in front of them.

Of course, this second embodiment can be declined according to the different variants previously envisaged for the first embodiment.

On the other hand, to illustrate the first and second embodiments, sound signals S and S are considered stereo.

However, the invention also applies to other types of multichannel signals, for example to a multi-channel 5.1 signal.

The figure 10 illustrates a reproduction system 1 '"according to a third embodiment, allowing the reproduction of a sound signal S'" 5.1, composed of a left channel (SL "'), a right channel (SR"') , a left rear channel (SLs '"), a right rear channel (SRs"'), a center channel (SC "'), and a bass channel (SLfe"').

The reproduction system 1 '"comprises two subsystems" front "A and" rear "B, respectively placed in front of and behind the listening positions P1" and P2 "" (with respect to the positioning of the listeners at these positions d 'listening).

The subsystem A comprises three sets of loudspeakers EA1 "', EA2'" and EA3 '"arranged with respect to the listening positions P1"' and P2 ", and with respect to the vertical separation plane D" 'of identical to the sets of speakers E1 ", E2" and E3 "considered previously to illustrate the second embodiment.

The subsystem B further comprises two sets of loudspeakers EB1 '"and EB2'" arranged behind the listening positions P1 '"and P2'" identically to the loudspeaker assemblies E1 'and E2 'considered previously to illustrate the first embodiment. EB1 '"and EB2 '' are placed on either side of the vertical separation plane D '", and are respectively centered with respect to the vertical planes L1"' and L2 "'which are parallel to the separation plane D"' and pass through P1 '"and P2'" respectively.

The reproduction system 1 '"thus comprises a frontal network of loudspeakers comprising the sets EA1"', EA2 '"and EA3'" which lies in a front plane with respect to the listening positions P1 "'and P2" ', and a rear speaker array comprising the sets EB1 "and EB2"' which lies in a rear plane with respect to the listening positions P1 "'and P2'".

• les canaux SL'" et SR'" sont utilisés pour générer des signaux d'alimentation des ensembles de haut-parleurs EA1'" et EA2'" respectivement, comme décrit précédemment dans le premier mode de réalisation, résultant en la création de deux sources virtuelles VA1'" et VA2"' ;
• le canal SC'" est utilisé pour générer des signaux d'alimentation des haut-parleurs du troisième ensemble EA3'" de haut-parleurs du sous-système A, en appliquant une technique de restitution spatiale directement sur ces canaux, comme décrit précédemment pour le signal d'entrée SC" dans le deuxième mode de réalisation ;
• les canaux SLs'" et SRs'" sont utilisés pour générer des signaux d'alimentation des ensembles de haut-parleurs EB1'" et EB2'" respectivement, comme décrit précédemment dans le premier mode de réalisation, résultant en la création de deux sources virtuelles VB1'" et VB2"', comme illustré à la figure 10 ; et
• et le canal SLfe'" est utilisé pour générer un signal d'alimentation d'un caisson de grave non représenté sur la figure 10.

The sound signal S '"is then restored via the system 1"' as follows:

• SL '' and SR '' channels are used to generate power supply signals from speaker sets EA1 '"and EA2'" respectively, as previously described in the first embodiment, resulting in the creation of two sources virtual VA1 '"and VA2"';
• the channel SC '' is used to generate speaker power signals of the third set of loudspeaker speakers EA3 '' of subsystem A, by applying a spatial rendering technique directly on these channels, as previously described for the input signal SC "in the second embodiment;
• the channels SLs '"and SRs'" are used to generate power signals for the speaker assemblies EB1 '"and EB2'", respectively, as previously described in the first embodiment, resulting in the creation of two sources VB1 '"and VB2"', as illustrated in figure 10 ; and
• and the channel SLfe '' is used to generate a power signal of a subwoofer not shown on the figure 10 .

It should be noted that the "rear" subsystem B may be an exact replica of the "forward" subsystem A, symmetrical with respect to the listening positions P1 "and P2"'. Alternatively, different loudspeaker configurations may be considered for both subsystems (eg, three sets of two speakers for A and two sets of two speakers for B, as illustrated in FIG. figure 10 ).

Claims (14)

1. A method for generating feed signals intended to a sound reproduction system (1'; 1"; 1"') comprising two sets of loudspeakers (E1', E2'; E1", E2"; EA1 "', EA2"', EB1"', EB2"') located on either side of a separation plane (D'; D"; D"'), said feed signals being generated in view of sound reproduction of a multichannel sound signal (S'; S"; S"') to two spatial listening positions (P1', P2'; P1", P2"; P1"', P2"'), each spatial listening position being associated to one of the sets of loudspeakers and being located on a plane (L1', L2'; L1", L2"; L1"', L2"') parallel to the separation plane such that at least one loudspeaker of this set is positioned on either side of this parallel plane,
   the generation method comprising, for each set of loudspeakers, a generation step (F20-F30, F50-F60) from an input signal derived from the multichannel sound signal for this set, of feed signals intended to feed the loudspeakers of this set, this generation step comprising application (F20, F50) of a transaural reproduction technique to said input signal, this technique creating a virtual sound source (V1', V2'; V1", V2"; VA1"', VA2"', VB1"', VB2"') for the listening position associated to the set on a first straight line, (d11', d22'; d11", d22"; d11"', d22", d22'", d11"') obtained by symmetry, relative to the parallel plane (P1', P2'; P1", P2"; P1"' P2"') on which the spatial listening position associated to this set is located, of a second straight line, (d12', d21' ; d12", d21"; d12"', d21"', d21"', d12"') connecting the spatial listening position to the loudspeaker farthest away from this position belonging to the other set of loudspeakers of the reproduction system.
2. The method according to claim 1 wherein the loudspeakers of the two sets of loudspeakers are substantially aligned on a horizontal axis (Δ', Δ") of the reproduction system.
3. The method according to claim 2 wherein during the generation step, the virtual source is created at the intersection of the first straight line and of the horizontal axis of the reproduction system.
4. The method according to any one of claims 1 to 3, further comprising a spectral equalisation step during which an equalisation filter adapted to a loudspeaker is applied (F30, F60) individually to each feed signal intended to feed this loudspeaker.
5. The method according to claim 4 wherein each equalisation filter adapted to a loudspeaker is selected so as to obtain an amplitude spectrum of a frequential response set up for this loudspeaker from measurements taken in a noiseless environment, according to a first preset target spectrum.
6. The method according to claim 4 or 5 wherein the equalisation filters are selected so as to obtain an average amplitude spectrum evaluated from the pulse responses of the loudspeakers of the sets of the reproduction system, according to a second preset target spectrum.
7. The method according to any one of claims 1 to 6 further comprising a spatial analysis step (F10) of the multichannel sound signal comprising extraction of at least one sound object and determination of a spatial position of this sound object, each input signal used for generating the feed signals of the loudspeakers of a set of loudspeakers being prepared from at least one sound object extracted during this spatial analysis step and selected as a function of its spatial position relative to the separation plane.
8. The method according to any one of claims 1 to 7 wherein the reproduction system further comprises a third set (E3") of loudspeakers, at least one loudspeaker of this third set being placed on either side of the separation plane.
9. The method according to any one of claims 1 to 8 wherein each set of loudspeakers of the reproduction system comprises a pair of loudspeakers.
10. A program comprising instructions for execution of the steps of the reproduction method according to any one of claims 1 to 9 when said program is executed by a computer or by a microprocessor.
11. A recording medium readable by a computer on which a computer program is registered comprising instructions for executing the steps of the generation method according to any one of claims 1 to 9.
12. A device for generating (2', 2", 2"') feed signals intended for a sound reproduction system (1'; 1"; 1"') comprising two sets of loudspeakers located on either side of a separation plane, said feed signals being generated in view of sound reproduction of a multichannel sound signal to two spatial listening positions, each spatial listening position being associated to one of the sets of loudspeakers and being located on a plane parallel to the separation plane so that at least one loudspeaker of this set is positioned on either side of this parallel plane,
    the generation device comprising means capable of generating for each set of loudspeakers, from an input signal derived from the multichannel sound signal for this set, feed signals intended to feed the loudspeakers of this set, these means being capable of applying a transaural reproduction technique to said input signal, this technique creating a virtual sound source for the listening position associated to the set on a first straight line obtained by symmetry, relative to the parallel plane on which the spatial listening position associated to this set is located, of a second straight line connecting the spatial listening position to the loudspeaker farthest away from this position belonging to the other set of loudspeakers of the reproduction system.
13. A sound reproduction system (1'; 1"; 1"') comprising:

    - at least two sets of loudspeakers; and

    - a device for generating (2'; 2"; 2"') according to claim 12 feed signals intended for the loudspeakers of said two sets of loudspeakers.
14. A sound reproduction system (1"') according to claim 13 comprising at least two sets (EA1"', EA2"') located in a frontal plane relative to the listening positions (P1"', P2"') and at least two sets (EB1"', B2"') located in a rear plane relative to the listening positions.

Images