EP1994526A1

EP1994526A1 - Joint sound synthesis and spatialization

Info

Publication number: EP1994526A1
Application number: EP07731685A
Authority: EP
Inventors: Grégory PALLONE; Marc Emerit; David Virette
Original assignee: France Telecom SA
Current assignee: Orange SA
Priority date: 2006-03-13
Filing date: 2007-03-01
Publication date: 2008-11-26
Anticipated expiration: 2027-03-01
Also published as: ES2335246T3; EP1994526B1; JP5051782B2; JP2009530883A; US20090097663A1; WO2007104877A1; US8059824B2; ATE447224T1; PL1994526T3; DE602007002993D1

Abstract

The invention concerns a process for joint synthesis and spatialization of multiple sound sources in associated spatial positions, including: a) a step of assigning to each source at least one parameter (pi) representing an amplitude; b) a step of spatialization consisting in implementing an encoding into a plurality of channels, wherein each amplitude (pi) is duplicated to be multiplied to a specialization gain (gim), each spatialization gain being determined for one encoding channel (pgm) and for a source to be spatialized (Si); c) a step of grouping (R) the parameters multiplied by the gains (Pim), in respective channels (pg1,..., pgM), by applying a sum of said multiplied parameters (pim) on all the sources (Si) for each channel (pgm), and d) a step of parametric synthesis (SYNTH(I ),..., SYNTH(M)) applied to each of the channels (pgm).

Description

Joint sound synthesis and spatialization

The present invention relates to audio processing and, more particularly, to three-dimensional spatialization of synthetic sound sources.

Currently, the spatialization of a synthetic sound source is often performed without taking into account the sound production mode, that is to say, the way in which the sound is synthesized. Thus, many models, including parametric, have been proposed for synthesis. At the same time, many spatialization techniques have also been proposed, without proposing a cross-check with the technique chosen for a synthesis.

Among the synthetic techniques, the so-called "non-parametric" methods are known. No particular parameter is used a priori to modify samples previously stored in memory. The best-known representative of these methods is classical wave table synthesis.

This type of technique is opposed by "parametric" synthesis methods, which rely on the use of a model that makes it possible to manipulate a small number of parameters, compared to the number of signal samples produced in the sense of the non-parametric methods. Parametric synthesis techniques typically rely on additive, subtractive, source / filter or non-linear models.

Among these parametric methods, those which make it possible to manipulate parameters corresponding to different sound sources, so as to use only one synthesis process, nevertheless for all sources, are called "mutuals". In so-called "sinusoidal" methods, typically, a frequency spectrum is constructed from parameters such as the amplitude and frequency of each component. partial of the overall sound spectrum of the sources. Indeed, an implementation by inverse Fourier transform, followed by an addition / overlap, ensures an extremely efficient synthesis of several sound sources simultaneously.

As far as the spatialization of sound sources is concerned, different techniques are currently known. Some techniques (such as "transaural" or "binaural") are based on the consideration of HRTFs ("Head Related Transfer Function") transfer functions representing the disturbance of acoustic waves by the morphology of an individual, these HRTFs functions being specific to this individual. The sound reproduction is carried out in a way adapted to the HRTFs of the listener, typically on two remote speakers ("transauraf") or from the two earpieces of a headset ("binaural") Other techniques (for example Vamphonic "or the" multichannel "(5.1 to 10.1 or more) provide rather a restitution on more than two speakers.

More specifically, some HRTFs techniques use the separation of the frequency and position variables of the HRTFs, thus giving a set of p basic filters (corresponding to the first p eigenvalues of the covariance matrix of the HRTFs whose variables statistics are the frequencies), these filters being weighted by spatial functions (obtained by projection of the HRTFs on basic filters). The spatial functions can then be interpolated, as described in US-5,500,900.

Spatialization of many sound sources can be achieved through a multichannel implementation applied to the signal of each of the sound sources. The gains of the spatialization channels are applied directly to the sound samples of the signal, often described in the time domain (but possibly also in the frequency domain). These samples sound are processed by a spatial isation algorithm (with gain applications that depend on the desired position), irrespective of the origin of these samples. Thus, the proposed spatialization could apply to both natural and synthetic sounds.

On the one hand, each sound source must be synthesized independently (with a temporal or frequency signal), in order to then be able to apply independent spatialization gains. For N sound sources, it is therefore necessary to perform N synthesis calculations. On the other hand, the application of the gains to sound samples, whether they come from the time or frequency domain, requires at least as many multiplications as there are samples. For a block of Q samples, it is therefore necessary to apply at least N. M. Q gains, M being the number of intermediate channels (surround channels for example) and N being the number of sources.

Thus, this technique requires a high calculation cost in the case of the spatialization of many sound sources.

Among the ambiophonic techniques, the so-called "virtual loudspeakers" method makes it possible to encode the signals to be spatialized by applying them in particular gains, the decoding being done by convolution of the signals encoded by pre-calculated filters (Jérôme Daniel, "Representation of acoustic fields, application to the transmission and reproduction of complex sound scenes in a multimedia context", PhD Thesis, 2000).

A very promising technique, combining synthesis and spatialization, has been presented in document WO-05/069272.

It consists in determining amplitudes to be assigned to signals representing sound sources to define, at the same time, the loudness (by example a "volume") of a source to synthesize and a gain of spatialization of this source. This document notably discloses a binaural spatialization taking into account delays and gains (or "spatial functions") and, in particular, a mixing of the sources synthesized in the encoding part of spatialization.

More particularly still, an exemplary embodiment which is referred to in this document WO-05/069272 and in which the sources are synthesized by associating amplitudes at frequencies constituting a "sound tone" (for example a fundamental frequency and its harmonics ), plans to group synthesized signals by identical frequencies, with a view to subsequent spatialization operating on the frequencies.

This exemplary embodiment is illustrated in FIG. 1. In a SYNTH synthesis block (represented in dashed lines), the frequencies fo, fi, f2,..., F _P of each source to be synthesized Si are assigned to. . S _N respective amplitudes a ₀ ¹ , ai ¹ , ..., a _p ¹ , ..., a /, ..., a ₀ ^N , a-Λ ..., a _p ^N , where, in the general notation al, j is a source index between 1 and N and i is a frequency index between 0 and p. Of course, certain amplitudes of a set ao ^j , a-ι ^j , ..., a _p ^j to be assigned to the same source j can be zero if the corresponding frequencies are not represented in the sound signal of this source j .

The amplitudes ai ¹ , ..., ai ^N relative to each frequency fi are grouped ("mixed") to be applied, frequency by frequency, to the SPAT spatialization block for encoding operating on the frequencies (in binaural for example, in anticipating interaural delay to be applied to each source). The signals of the channels Ci,..., C _k , originating from the spatialization block SPAT, are then intended to be transmitted through one or more networks, or else stored, or other, for the purpose of a subsequent restitution (preceded by where appropriate, a suitable spatialization decoding). This technique, although very promising, still deserves some optimizations.

In general, current methods require significant computing power to spatialize many synthesized sound sources.

The present invention improves the situation.

To this end, it proposes a method for jointly synthesizing and spatialising a plurality of sound sources in associated positions of the space, the method comprising: a) a step of assigning to each source at least one parameter representative of a amplitude, b) a spatialization step implementing an encoding in a plurality of channels, in which each amplitude parameter is duplicated to multiply by a spatialization gain, each spatialization gain being determined, on the one hand, for an encoding channel and, secondly, for a source to be spatialised, c) a step of grouping the parameters multiplied by the gains, in respective channels, by applying a sum of said multiplied parameters to all the sources for each channel and d) a parametric synthesis step applied to each of the channels.

Thus, the present invention proposes for this purpose to first apply a spatialization encoding, then a "pseudo-synthesis", the term "pseudo" aiming at the fact that the synthesis applies in particular to the encoded parameters, derived from spatialization and not to usual synthetic sound signals.

Indeed, a feature that the invention proposes is the spatial encoding of some synthesis parameters, rather than performing a spatial encoding of the signals corresponding directly to the sources. This spatial encoding applies more particularly to synthesis parameters which are representative of an amplitude and it advantageously consists in applying to these few synthesis parameters spatialization gains which are calculated as a function of respective desired positions of the sources. It will thus be understood that the parameters multiplied by the gains in step b) and grouped in step c) are not really sound signals, as in the general prior art described above.

The present invention then uses a mutual parametric synthesis where one of the parameters has the dimension of an amplitude. Unlike techniques of the prior art, it thus takes advantage of the advantages of such a synthesis to perform the spatialization. The combination of synthesis parameter sets obtained for each of the sources advantageously makes it possible to globally control the encoded blocks of mutual parametric synthesis.

The present invention then makes it possible to spatialize simultaneously and independently of numerous synthesized sound sources from a parametric synthesis model, the spatialization gains being applied to the synthesis parameters rather than to the samples of the time or frequency domain. This embodiment thus ensures a substantial saving of the computing power required because it implies a low calculation cost.

According to one of the advantages provided by the invention, since the number of steps in the synthesis is made independent with respect to the number of sources, only one intermediate channel synthesis can be applied. Regardless of the number of sound sources, only a constant number M of synthesis calculations is planned. Typically, since the number of sources N becomes larger than the number M of intermediate channels, the technique in the sense of the invention requires fewer calculations than the usual techniques in the sense of the prior art. For example, at the surround order 1 and in two dimensions (ie three intermediate channels), the invention already allows a calculation gain for only four sources to spatialize.

The present invention also makes it possible to reduce the number of gains to be applied. Indeed, the gains are applied to the synthesis parameters and not to the sound samples. Updating parameters such as the volume is generally less frequent than the sampling frequency of a signal, a calculation economy is thus achieved. For example, for a frequency of updating parameters (such as the volume in particular) of 200 Hz, a substantial saving in multiplication is achieved for a sampling frequency of the signal of 44100 Hz

(according to a ratio of about 200).

The fields of application of the present invention may concern both the musical field (including polyphonic ringtones of mobiles), the field of multimedia (including video game sound systems), the field of virtual reality (rendering of sound scenes). , simulators (synthesis of engine noise), or others.

Other features and advantages of the invention will appear on examining the detailed description below, and the accompanying drawings in which, in addition to Figure 1 relating to the prior art and described above:

FIG. 2 illustrates the general processing of spatialization and synthesis provided for in a method according to the invention; FIG. 3 illustrates a processing of the spatialized and synthesized signals for spatial decoding with a view to restitution,

FIG. 4 illustrates a particular embodiment in which several amplitude parameters are assigned to each source, each parameter being associated with a frequency component, FIG. 5 illustrates the steps of a method in the sense of the invention, and may correspond to a flowchart of a computer program for the implementation of the invention.

With reference to FIG. 2, at least one parameter Pi, representative of an amplitude, is assigned to a source Si from among a plurality of sources Si, ..., S _N to be synthesized and spatialized (i being between 1 and N ). Each parameter pi is duplicated in as many spatialization channels provided in the spatialization block SPAT. In the example shown where M encoding channels are provided for the spatialization, each parameter pi is duplicated M to apply respective spatialization gains gΛ ..., where ^M (i being, as a reminder, an index of source Si). We then obtain N.M parameters each multiplied by a gain: p-igΛ ...,

Pi9i ^M , ---, PigΛ ---, Pi9i ^M , ---, PNgN ¹ , ..., PNgN ^M. These multiplied parameters (reference R of FIG. 2) are then grouped by spatialization channels (M channels in all), namely:

- pigΛ ..., pigΛ ..., PNgN ¹ grouped in a first spatialization channel

P ₉ ¹ . and this, up to: - pigi ^M , ..., Pigi ^M , ..., PNgN ^M grouped in a ^same channel of spatialization

the letter g of the index designating the term "globaf.

Thus, new parameters p ™ (i varying from 1 to N and m varying from 1 to M) are calculated by multiplying the parameters pi by the encoding gains g ™, obtained from the position of each of the sources. The parameters Pi ^m are combined (by summation in the example described) to provide the parameters p _g ^m which feed M mutual parametric synthesis blocks.

These M blocks (referenced SYNTH (I) to SYNTH (M) in FIG. 2) constitute the synthesis module SYNTH, which delivers M time or frequency signals ss ^m (m varying from 1 to M), obtained by synthesis from of the parameters p _g ^m . These signals ss ^m can then feed a conventional block of spatial decoding, as will be seen below with reference to FIG.

In a particular embodiment, the synthesis used is an additive synthesis with application of an inverse Fourier transform (IFFT).

For this purpose, a set of N sources is characterized by a plurality of parameters pi, _k representing the amplitude in the frequency domain of the kth ^frequency component for the ^ith source Sj. The time signal Si (n) that would correspond to this source If, if it were synthesized independently of the other sources, would be given by:

^s i (») = Σ ^c α ( ⁿ )" ^{with c} α ( ⁿ ) = Pa (») ^cos [ ² ^ α ^ ^{nj F} e + V ₁ * ( ⁿ )] k = 1 where Pi, _k is the amplitude of the component of frequency fi, _k and whose phase is given by φ _ijk for the source Si, at time n. It is possible to carry out the additive synthesis in the frequency domain from the only given parameters pi, _k , fi, k and φi, _k , using, for example, the technique described in the document FR-2 679 689.

The parameter Pi, _k represents the amplitude of a given frequency component k for a given source Si. We thus deduce the parameters p ^m i, _k for each source and each of the M channels by means of the relation: p ^m i, k = g ^m ι - Pι, k _> rn varying from 1 to M.

G gains are predetermined for a desired position for the source Si and according to the selected spatialization encoding.

In the case of an ambiophonic encoding for example, these gains correspond to the spherical harmonics and can be written g ™ = Y _m (θi, δi), where:

Y _m is a spherical harmonic of order m,

θi and δi are respectively the azimuth and the desired site for the Si source. The parameters p ^m i, k are then combined frequency by frequency, so as to obtain a single global parameter:

NOT

P ¹ V = ΣP ^m ' ^k _> ^{or k <} describes all the frequencies fi, _k present in all

1 = 1 sources Sj.

In practice, the value of k ¹ is less than ki because common frequencies can characterize several sources at once. In one embodiment, it may be provided to associate the same global set of frequencies to all sources, even if certain amplitude parameters for certain source frequencies are zero.

In this case, the values of k and k 'are equal and the preceding relation is written simply:

NOT

P S. ^k = 2 ^ P ^> - ^k

1 = 1

The synthesis step consists of using these parameters p ^m _g , _k (m varying from 1 to M) to synthesize each of the M frequency spectra ss ^m (ω) from the SYNTH synthesis module. It may be provided for this purpose to apply the technique described in FR-2,679,689, iteratively adding spectral envelopes corresponding to the Fourier transform of a time window (for example Hanning), these spectral envelopes being previously sampled , tabulated, centered at frequencies fk and then weighted by p ^m _g , _k , which is written as:

ss θ ^m) = approx _k (ω), where env _k (ω) is the spectral envelope centered at the frequency f _k .

This embodiment is illustrated in FIG. 4. K amplitude parameters p _{iik are assigned} to each source Sj. The index i, of source, is between 1 and N. The index k, of frequency, is between 1 and K. For each source Si, these K parameters are duplicated, M times, to be multiplied each by one gain Spatialization g ™. The index m, of spatialization encoding channel, is between 1 and M.

In each channel m, we group, frequency by frequency, the K results of products gΛpi. _k , according to the expression given above:

pV = ΣP ^m α. with p ^m i, _k = g ^m i. pi, _k ,

1 = 1 where k varies from 1 to K in each channel m, and m varies globally from 1 to M. It will thus be understood that in each channel m, sub-channels p ^m _g , _k each associated with a component are provided. frequency k, the index g designating, as a reminder, the term "globaf.

The processing then continues by multiplying the global parameter of each subchannel p ^m _g , k associated with a frequency fk by a spectral envelope envk (ω) centered at this frequency fk, and this, for all the K subchannels ( k between 1 and K), and globally, for all M channels (m being between 1 and M). Then the sub-channels K are summed in each channel m, according to the following relation:

ss ^m (<") = env _k {ω), for m ranging from 1 to M channels in total.

The signals ss, m ^m (/ ω) encoded for their spatialization and synthesized in the sense of the invention are then obtained. They are expressed in the frequency domain.

To bring back these M signals in the time domain (then noted SS ^m (n)), we can then apply them an inverse Fourier transform (IFFT): SS ^m (n) = IFFT (ss ^m (ω)). successive frames can be achieved by a conventional technique of addition / overlap. Each of the M time signals SS ^m (n) can then be supplied to a spatialization decoding block.

For this purpose, it is possible, for example, to provide a pair of matched filters Fg ^m (n), Fd ^m (n) to be convolutionally applied to each signal SS ^m (n), as shown in FIG. adaptation of an ambiophonic encoding to a two-way left and right binaural rendition. These filters for such an ambiophonic / binaural transition can be obtained by applying the virtual speaker technique mentioned above.

The processing performed by the spatial decoding DECOD block of FIG. 3 can be of the type:

SS ^m _g (n) = (SS ^m * Fg ^m ) (n)

SS ^m _d (n) = (SS ^m * Fd ^m ) (n) After filtering, all the signals for the left and right ears are summed respectively, thus obtaining a pair of binaural signals:

M

S _d (n) = ΣSS ™ (n) m = 1 which then supply the headphones of a headset with two earpieces.

However, a more advantageous variant is described below. The adaptation filters from the surround format to the binaural format can be applied directly in the frequency domain, thus avoiding convolution in the time domain and a corresponding calculation cost.

For this purpose, each of the M frequency spectra ss ^m (ω) is directly multiplied by the respective Fourier transforms of the temporal filters, noted Fg ^m (ω) and Fd ^m (ω) (adapted if necessary to have a coherent number of points), which is written: ss ^m _g (ω) = ss ^m (ω) Fg ^m (ω) ss ^m _d (ω) = ss ^m (ω) Fd ^m (ω) The spectra are then summed by ear before performing the inverse Fourier transform and the addition / recovery operation, ie:

s _g (ω) = Σs ^m Aω) m = l M s _d (ω) = Σs ^m _d (ω)

Then, to express the signals supplying the rendering device in the time domain, the inverse Fourier transform is applied: S _g (n) = IFFT (S _g (ω))

S _d (n) = IFFT (s _d (ω))

The present invention also relates to a computer program product, whether it is stored in a memory of a central unit or a terminal, or on a removable support adapted to cooperate with a reader of this central unit.

(CD-ROM, floppy disk or other), or downloadable via a telecommunications network. This program comprises in particular instructions for the implementation of the method described above and a flowchart can be illustrated by way of example in Figure 5, summarizing the steps of such a method.

Step a) is aimed at assigning the parameters representative of an amplitude to each source Sj. In the example shown, a parameter pi, _k is assigned by frequency component f _k as described above. Step b) aims at the duplication of these parameters and their multiplication by the gains g ™ of the encoding channels.

Step c) relates to the grouping of the products obtained in step b), with in particular the calculation of their sum on all Si sources. Step d) targets the parametric synthesis with multiplication by a spectral envelope env _k as described above, followed by a grouping of the subchannels by applying, in each channel, a sum over all the frequency components (d index k ranging from 1 to K). Step e) aims at decoding the spatialization of the signals ss ^m originating from the respective channels, synthesized, spatialised and represented in the frequency domain, for a reproduction on two loudspeakers, for example in binaural format.

The present invention also provides a device for generating synthetic and spatialized sounds, comprising in particular a processor, and in particular a working memory adapted to store instructions of the computer program product defined above.

Of course, the present invention is not limited to the embodiment described above by way of example; it extends to other variants.

Thus, it has been described above as an example encoding space spatialization in surround format made by the SPAT module of Figure 2, followed by an adaptation of the surround format to the binaural format. As a variant, it may be provided, for example, to directly apply an encoding to the binaural format.

Moreover, the spectral envelope multiplication of the parametric synthesis is described above by way of example, other models that can be provided alternatively.

Claims

claims

A method for synthesizing and spatially co-ordinating a plurality of sound sources in associated positions of the space, comprising: a) a step of assigning to each source at least one parameter (pi) representative of an amplitude, b ) a spatialization step implementing an encoding in a plurality of channels, in which each amplitude parameter (pi) is duplicated to multiply it to a spatialization gain (g ™), each spatialization gain being determined, and on the one hand, for an encoding channel (p _g ^m ) and, on the other hand, for a source to be spatialized (Si), c) a grouping step (R) of the parameters (pi ^m ) multiplied by the gains, in respective channels (p _g ¹ ,..., p _g ^M ), by applying a sum of said multiplied parameters (pi ^m ) to all the sources (Si) for each channel (p _g ^m ), and d) a step of parametric synthesis (SYNTH (I), ..., SYNTH (M)) applied to each of the channels (p _g ^m ).

The method according to claim 1, wherein: a) each source (Si) is assigned a plurality of representative parameters (p _ijk ), each of an amplitude of a frequency component

(f _k ), b) we duplicate each amplitude parameter (pi, _k ) representative of a frequency component (f _k ) to multiply it to a spatialization gain (gi ^m ), each spatialization gain being determined, d on the one hand, for an encoding channel (p _g ^m ) and, on the other hand, for a source to be spatialised (S ₁ ), c) in each channel, the frequency component by frequency component is grouped together. parameters (p _ijk ) by the gains (g ™), in subchannels (p _g , k ^m ) each associated with a frequency component (fk).

3. Method according to claim 2, in which the synthesis is carried out, in each channel, by: d1) multiplying the output of each subchannel associated with a frequency component (fk) by a spectral envelope (env _k ) centered on a frequency corresponding to said frequency component (f _k ), d2) and grouping, by a sum on the frequency components (f _k ), the products resulting from the operation d1), to obtain, following the operation d2), a signal (ss ^m ) from each channel, encoded in space isation and synthesized.

4. Method according to one of the preceding claims, wherein the spatialization is conducted by surround encoding and the representative parameters of an amplitude which are assigned to the sources correspond to spherical harmonics amplitudes (Y _m ).

The method of claim 4, taken in combination with claim 3, wherein, to switch from surround encoding to decoding for binaural spatialization rendering, frequency domain processing is directly applied to the results. products from the respective channels after the operation d2).

6. Computer program product, stored in a memory of a central unit or a terminal, and / or on a removable support adapted to cooperate with a reader of said central unit, and / or downloadable via a telecommunications network , characterized in that it comprises instructions for implementing the method according to one of claims 1 to 5.

7. module for generating spatialised synthetic sounds, comprising in particular a processor, characterized in that it further comprises a working memory storing instructions of the computer program product according to claim 6.