EA042232B1 - ENCODING AND DECODING AUDIO USING REPRESENTATION TRANSFORMATION PARAMETERS - Google Patents

Description

The present invention relates to the field of signal processing, and in particular, it discloses a system for the efficient transmission of audio signals containing spatial representation components, i.e. audio components associated with different spatial locations.

Background of the invention

Any discussion of the current state of the art in this specification should in no way be construed as an admission that the state of the art is widely known or forms part of the public knowledge in the field.

Content creation, encoding, distribution and playback of audio is typically done in a channel-based format, i.e. one specific playback target system is provided for content across the entire content ecosystem. Examples of such target playback system audio formats are mono, stereo, 5.1, 7.1, and so on.

If the content is to be played by a different playback system than the intended one, downmixing or upmixing may be applied. For example, 5.1 content can be reproduced by a stereo playback system by using special downmix equations. Another example is the playback of stereo-encoded content on a 7.1 speaker system, which may include a process called upmixing, which may or may not be controlled by the information present in the stereo signal. A system capable of upmixing is Dolby Pro Logic from Dolby Laboratories Inc (Roger Dressier, Dolby Pro Logic Surround Decoder, Principles of Operation, www.Dolby.com).

An alternative sound format system is the sound object format, such as that provided by Dolby Atmos, see Robinson, C. Q., Mehta, S., Tsingos, N. (2012) Scalable format and tools to extend the possibilities of cinema audio Motion Imaging Journal , SMPTE, 121(8), 63-69. In this type of format, objects are given specific locations around the listener, which may be time-varying. In such an object-based format, content is represented in a manner that is invariant to a particular rendering system. Therefore, a specialized presentation process is required to convert content into a representation suitable for a particular playback system, such as a speaker system or headphones.

When stereo, multi-channel, or object-based content is to be played back through headphones, it is often necessary to simulate a multi-channel speaker system (for channel-based content) or a set of virtual sound sources (for object-based content) via head-related impulse response. , HRIR) or binaural room impulse responses (BRIR), which simulate the acoustic path from each speaker to the eardrums in an anechoic or echoic environment (simulated), respectively. In particular, audio signals can be convolved with HRIR or BRIR to reconstruct inter-aural level differences (ILD), inter-aural time differences (ITD), and spectral labels, allowing the listener to locate each individual channel. . Achieving a certain perceived distance is also facilitated by simulating the acoustic environment (reverberation). Let us turn to Fig. 1, which shows a schematic overview of the progress of data processing when presenting two object or channel signals xi 10, 11 read from the content memory 12 for processing 4 with four HRIRs, for example, 14. The output of the HRIRs is then summed 15, 16 for each channel signal to produce a headphone output for playback to a listener through headphones 18. The basic principle of HRIR is explained, for example, in Wightman, F. L., Kistler, D. J. (1989b). Headphone simulations of free-field listening. I. Stimulus synthesis, J. Acoust. soc. Am. 85, 858-867. The resulting headphone stereo signal 15, 16 is often referred to as a binaural signal, binaural presentation, or headphone (binaural) presentation. Moreover, such a binaural presentation is designed (or specifically designed) for playback through headphones, as opposed to a speaker presentation intended for playback by a speaker system that matches the channels represented in the speaker presentation signal(s). These different playback systems are called modalities, for example, one playback modality consists of headphones, while another playback modality contains one or more loudspeakers. Regardless of the playback modality, based on an input stream, such as a multi-channel or object-based content format, various representations (stereo, binaural, 5.1) can be presented (generated). Ideally, to ensure that the artistic intent is correctly communicated to the listener, representations are generated, or presented, for specific rendering modalities. In the case of headphone playback, this involves using HRIR or BRIR to create a binaural presentation, while amplitude panning techniques are typically used for loudspeakers. Thus, such a presentation process can be applied to both channel-based input content (5.1, 7.1, etc.) and object-based immersive content such as Dolby Atmos. In the latter case, amplitude panning (in loudspeaker representations) or BRIR (in headphone representations) is typically used independently for each input object, followed by summing the individual object contributions to the resulting binaural signal.

The convolution process to obtain a binaural representation intended for headphone playback can be constructed such that the audio source location cues present in the (anechoic) HRTFs (head-related transfer functions) are reconstructed for each input independently, depending on (intended, perceived) position of the input channel or object, while the simulated echo environment can be, at least in part, shared by a common algorithm for two or more inputs. To this end, one or more inputs are mixed, or combined, into one or more environment simulation algorithm inputs, which are subsequently processed to generate environment simulation outputs that can be combined with the output of an anechoic convolution process using HRTF. The environment simulation algorithm may simulate early reflections, late reverbs, or both, and may be implemented by known techniques such as convolution, delays, feedback delay networks, all-pass filters, and the like.

The HRIR/BRIR convolution approach comes with several drawbacks, one of which is the significant amount of convolution data processing required for headphone playback. The HRIR or BRIR convolution needs to be applied to each input feature or channel individually, and thus the complexity typically increases linearly with the number of channels or features. Because headphones are often used in conjunction with battery-powered portable devices, high computational complexity is undesirable as this can significantly reduce battery life. Moreover, when presenting object-based audio content, which may contain, for example, more than 100 simultaneously active objects, the complexity of convolution with HRIR can be significantly higher than for traditional channel-based content.

One solution to reduce the computational load is to apply convolution processes upstream in the data processing chain. For example, during the content creation or coding phase. In this particular case, called binaural preview, the resulting binaural signal, or binaural representation, generated during the preview step contains all of the location cues intended to be played back by the headphones, and no further data processing is required in the playback device. The disadvantage of this method is that the input sound source location marks that are present in HRIR (such as interaural time differences (ITD, interaural time differences), interaural level differences (ILD, interaural level differences), spectral marks and reverberation) degrade perceived quality when this particular binaural presentation is played back by the loudspeakers, since said location cues will actually be applied twice: once algorithmically, during the pre-presentation stage, and once acoustically, as a consequence of the acoustic path between the loudspeakers and the listener's ears.

Brief description of the essence of the invention

The aim of the invention in its preferred embodiment is to provide an improved embodiment of the encoding and decoding of audio signals for playback.

According to a first aspect of the present invention, there is provided a method for encoding an input audio stream comprising one or more audio components, each audio component being associated with a spatial location, and the method comprising the steps of obtaining a first playback stream representation for the input audio stream, the first playback stream representation being a set of M1 signals intended to be reproduced by the first sound reproduction system, obtaining a second representation of the playback stream for the input audio stream, while the second representation of the reproduction stream is a set of M2 signals intended to be reproduced by the second sound reproduction system, determining a set of conversion parameters suitable for conversion the intermediate representation of the playback stream to the approximation of the second representation of the playback stream, and the intermediate representation of the playback stream The work is one of the first

- 2 042232 setting the playback stream, the result of the downmix of the first representation of the playback stream and the result of the upmix of the first representation of the playback stream, wherein the transformation parameters are determined by minimizing the amount of difference between the approximation of the second representation of the playback stream and the second representation of the playback stream, and encoding the first representation of the playback stream and a set of transformation parameters for transmission to the decoder.

According to a second aspect of the present invention, there is provided a method for decoding playback stream representations from a data stream, the method comprising the steps of receiving and decoding a first playback stream representation, wherein the first playback stream representation is a set of M1 signals to be reproduced by a first audio playback system, receiving and decoding a set of transformation parameters suitable for converting the intermediate representation of the playback stream to an approximation of the second representation of the playback stream, wherein the second representation of the playback stream is a set of M2 signals intended for reproduction by the second audio playback system, while the intermediate representation of the playback stream is one of the first playback stream representation, the downmix result of the first playback stream representation, or and a result of upmixing the first playback stream representation, wherein the transformation parameters ensure that the amount of difference between the approximation of the second reproduction stream representation and the second reproduction stream representation is minimal, and applying the transformation parameters to the intermediate representation of the playback stream to obtain an approximation of the second representation of the playback stream.

In some embodiments, the first audio playback system may include a number of speakers at fixed spatial locations, and the second audio playback system may include a set of headphones adjacent to the listener's ears. The first or second representation of the playback stream may be an echo or anechoic binaural presentation.

The conversion parameters are preferably time-varying and frequency-dependent.

The transformation parameters are preferably determined by minimizing the amount of difference between the result of applying the transformation parameters to the first representation of the playback stream and the second representation of the playback stream.

In accordance with yet another aspect of the present invention, a method is provided for encoding audio channels or audio objects as a data stream, comprising the steps of: receiving N input audio channels or objects; calculating a set of M signals, where M < N, by forming combinations of N input audio channels or objects, while the set of M signals is intended to be reproduced by the first sound reproduction system; calculating a set of time-varying transform parameters W converting a set of M signals to be reproduced by the first audio reproduction system into approximate reproduction by the second audio reproduction system, wherein the approximation reproduces approximately any spatial presentation effects obtained by reproducing the N input audio channels, or objects by the second playback system; and combining the M signals and the transform parameters W into a data stream to be transmitted to the decoder.

In some embodiments, the transform parameters form an M1xM2 gain matrix that can be applied directly to the first playback stream representation to form said approximation of the second playback stream representation. In some embodiments, M1 is equal to M2, i. e. both views, first and second, contain the same number of channels. In a particular case, both representations, the first and second, are stereo representations, i.e. M1=M2=2.

Those skilled in the art will appreciate that the first presentation stream encoded in the encoder may be a multi-channel speaker presentation, such as a surround or immersive (3D) speaker presentation, such as a 5.1, 7.1, 5.1.2 presentation. , 5.1.4, 7.1.2 or 7.1.4. In such a situation, in order to avoid or minimize computational complexity, in accordance with one embodiment of the present invention, the step of determining the set of transformation parameters may include downmixing the first representation of the playback stream into an intermediate representation with fewer channels.

In a particular example, this intermediate representation is a two-channel representation. In this case, the conversion parameters are thus suitable for converting the intermediate two-channel representation into the second representation of the playback stream. The first representation of the playback stream may be ambient or immersive.

- 3 042232 live presentation for loudspeakers.

The data stream may further comprise dialog signal evaluation parameters, the method further comprising: applying the dialogue signal evaluation parameters to the signals to be reproduced by the first audio playback system to obtain one or more evaluation dialogue signals; subtracting one or more estimated dialogue signals from the signals intended to be reproduced by the first sound reproduction system to obtain an intermediate signal with suppressed dialogue; applying the dialogue-suppressed intermediate signal to the acoustic environment simulation process to obtain one or more simulated acoustic environment signals; and combining one or more of the simulated acoustic environment signals with an audio stream suitable for playback by the second audio playback system.

The data stream may further comprise parameters W (W _F , WE) of generating input signals of the process of simulating the acoustic environment, while the method further includes: applying the parameters W (W _F , WE) of generating input signals of the process of simulating the acoustic environment to the signals intended to reproduce the first a sound reproduction system for receiving one or more input signals of the process of simulating the acoustic environment; applying one or more acoustic environment simulation process inputs to the acoustic environment simulation process to produce one or more simulated acoustic environment signals; and combining one or more of the simulated acoustic environment signals with an audio stream suitable for playback by the second audio playback system.

One or more simulated acoustic environment signals may preferably comprise one or more of early reflection signals and late reverberation signals. The acoustic environment simulation process may include one or more of an early reflection simulation process and a late reverberation simulation process. The early reflection simulation process may include processing one or more input signals of the acoustic environment simulation process through a delay element. In some embodiments, the late reverberation simulation process may include processing one or more acoustic environment simulation process inputs via a feedback delay network.

The data stream may preferably contain additional parameters W (W _F , WE) of generating input signals of the process of simulating the acoustic environment, while the method further includes the steps of: applying additional parameters W (W _F , W _E ) of generating input signals of the process of simulating the acoustic environment to the signals, intended to be reproduced by the first sound reproduction system, to receive one or more additional input signals of the process of simulating the acoustic environment; applying one or more additional acoustic simulation process inputs to the additional acoustic environment simulation process to produce one or more additional acoustic simulation process inputs; and combining one or more additional simulated acoustic environment signals with one or more of the one or more simulated acoustic environment signals and an audio stream suitable for playback by the second audio playback system.

The acoustic environment simulation process may be configured in accordance with one or more parameters, these parameters being dependent on one or more of the user's preferences and information included in the data stream.

In accordance with yet another aspect of the present invention, an encoder is provided for encoding an input audio stream comprising one or more audio components, wherein each audio component is associated with a spatial location, the encoder comprising: a first presentation unit for representing a first representation of the playback stream of the input audio stream, wherein the first playback stream representation is a set of M1 signals to be reproduced by the first audio playback system, the second presentation unit to represent the second playback stream representation of the input audio stream, the second playback stream representation is a set of M2 signals to be reproduced by the second audio playback system , a transform parameter determiner for determining a set of transform parameters suitable for transforming the intermediate representation of the playback stream approximation of the second representation of the playback stream, while the intermediate representation of the playback stream is one of the first representation of the playback stream, the result of the downmix of the first representation of the playback stream and the result of the upmix of the first representation of the playback stream, while the transformation parameters are determined by minimizing the amount of difference between the approximation a second representation of the playback stream and a second representation of the playback stream, and an encoding unit for encoding the first representation of the playback stream and a set of transformation parameters for transmission to the decoder.

In accordance with another aspect of the present invention, a decoder is provided for decoding representations of a playback stream from a data stream, comprising a base decoder block,

- 4 042232 configured to receive and decode a first representation of the playback stream, wherein the first representation of the playback stream is a set of M1 signals intended to be reproduced by the first audio playback system, and to receive and decode a set of conversion parameters suitable for converting the intermediate representation of the playback stream approximation of the second playback stream representation, wherein the second playback stream representation is a set of M2 signals to be reproduced by the second audio playback system, wherein the intermediate playback stream representation is one of the first playback stream representation, the downmix result of the first playback stream representation and the result of the upmix of the first representation of the playback stream, wherein the transformation parameters ensure that The difference between the approximation of the second representation of the playback stream and the second representation of the playback stream is minimal.

The decoder further comprises a matrix multiplier for applying transform parameters to the intermediate representation of the playback stream to obtain an approximation of the second representation of the playback stream.

According to a further aspect of the present invention, a decoder is provided for decoding a number of audio channels and/or audio objects from a data stream, the data stream comprising a set of M signals for playback by a first audio playback system and conversion parameters W adapted to convert the M signals for playback. a second audio playback system, the decoder comprising: a base decoder unit for separating the M signals and W transform parameters from the data stream, wherein the M signals are separated into at least high and low frequency bands; a matrix multiplier for applying the W transform parameters to the M signals to obtain a set of frequency division output signals; and a deconverter adapted to convert the set of frequency division output signals into a series of time domain output signals suitable for reproduction by the second audio reproduction system.

In some embodiments, the decoder may further comprise: a reverb block adapted to add reverb to the set of frequency division outputs before being converted by the inverse transform block.

In some embodiments, the first sound reproduction system may comprise a set of speakers and the second sound reproduction system may comprise a set of headphones, where the transform parameters W provide binauralization of the set of frequency division outputs in the sense that the second representation of the playback stream is an echo or an anechoic binaural performance.

According to a further aspect of the present invention, an encoder is provided for encoding an input audio stream comprising one or more audio components, each audio component being associated with a spatial location, the system comprising a first coding unit for encoding the input audio stream for a first playback modality, which outputs a first representation of the playback stream; a transformation parameter determination unit for determining a number of transformation parameters for mapping the first representation of the playback stream to the second representation of the reproduction stream; and a second encoding unit for encoding the first representation of the playback stream and the transformation parameters into an output encoding stream.

The transform parameter determiner determines the transform parameter set by minimizing the absolute value of the amount of difference between the desired second playback stream representation and the result of applying the transform parameter series to the first playback stream representation. The term series may refer to the property of having time-varying transform parameters and/or frequency-dependent transform parameters. The second representation of the playback stream may contain binauralized audio for playback with headphones.

In accordance with a further aspect of the present invention, there is provided a method for obtaining an audio signal for presentation by headphones, the method comprising the steps of receiving a data stream containing an encoded anechoic binaural signal and generating parameters W (W _F , WE) of an acoustic environment simulation process input signal; decoding the encoded anechoic binaural signal to obtain a decoded anechoic binaural signal; applying parameters W (WF, WE) of generating acoustic environment simulation process inputs to the decoded anechoic binaural signal to obtain one or more acoustic environment simulation process inputs; applying one or more acoustic simulation process inputs to the acoustic environment simulation process to produce one or more simulated acoustic environment signals and combining one or more simulated acoustic environment signals and a decoded anechoic binaural signal to produce an audio signal for presentation by headphones.

- 5 042232

In some embodiments, one or more of the simulated acoustic environment signals are preferably one or more of early reflection signals and late reverberation signals. The acoustic environment simulation process may include one or more of an early reflection simulation process and a late reverberation simulation process. The early reflection simulation process may include processing one or more input signals of the acoustic environment simulation process through a delay element. The late reverberation simulation process may include processing one or more inputs of the acoustic environment simulation process through a feedback delay network.

The data stream may preferably contain additional acoustic environment simulation process input parameters W (W _F , WE), wherein the method may further comprise the steps of applying the acoustic environment simulation additional process input parameters W (W _F , WE) to the decoded anechoic binaural a signal for receiving one or more additional inputs to the process of simulating the acoustic environment; applying one or more additional acoustic simulation process inputs to the additional acoustic environment simulation process to produce one or more additional simulated acoustic environment signals and combining one or more additional simulated acoustic environment signals with one or more simulated acoustic environment signals and/or decoded anechoic binaural signal.

In accordance with a further aspect of the present invention, a decoder is provided for obtaining an audio signal for presentation by headphones, wherein the decoder comprises one or more processors configured to receive a data stream containing an encoded anechoic binaural signal and parameters W (W _F , WE) of generating input signals the process of simulating the acoustic environment; decoding the encoded anechoic binaural signal to obtain a decoded anechoic binaural signal; applying parameters W (WF, WE) of generating acoustic environment simulation process inputs to the decoded anechoic binaural signal to obtain one or more acoustic environment simulation process inputs; applying one or more acoustic simulation process inputs to the acoustic environment simulation process to produce one or more simulated acoustic environment signals and combining one or more simulated acoustic environment signals and a decoded anechoic binaural signal to produce an audio signal for presentation by headphones.

Brief description of graphic materials

Embodiments of the invention will be described below with reference to the accompanying drawings, in which FIG. 1 shows a schematic overview of the HRIR devolution process for two source objects, in which each channel or object is processed by an HRIR/BRIR pair.

In FIG. 2 schematically shows binaural pre-rendered content played back by loudspeakers (prior art);

in fig. 3 schematically shows binaural pre-rendered content played back by loudspeakers;

in fig. 4 schematically shows the derivation of w coefficients for processing the presentation for loudspeakers for headphone playback;

in fig. 5 schematically shows the W(WE) coefficients used to reconstruct an anechoic signal and one early reflection (with an additional volume delay step);

in fig. 6 schematically shows the process of using the coefficients W (W _F ) used to reconstruct the anechoic signal and the FDN input signal from the base decoder output.

In FIG. 7 schematically shows the acquisition and processing of w coefficients for processing anechoic representation for headphones and loudspeakers.

In FIG. 8a, 8b show block diagrams of an encoder/decoder in accordance with a further embodiment of the present invention.

In FIG. 9a is a block diagram of a decoder in accordance with a further embodiment of the present invention.

In FIG. 9b is a block diagram of a simplified version of the decoder shown in FIG. 9a.

Detailed description

Embodiments provide a low bit rate, low complexity channel and/or object based audio presentation method that is suitable for (binaural) loudspeaker and headphone playback. This method is performed by (1) creating and encoding a representation intended for playback by a particular playback system (eg, without limitation, speakers), and (2) adding additional metadata that makes it possible to convert this particular presentation into a modified representation suitable for another playback system. (e.g. headphones

- 6 042232 kov). This particular view may be referred to as the first playback view of the audio stream, while the modified view may be referred to as the second playback view of the audio stream. The first view may contain a set of M1 channels, while the second view may contain a set of M2 channels. The number of channels may be equal (M1=M2 ) or different. The metadata may take the form of a set of parameters, possibly variable in time and frequency.

In one implementation, the transformation metadata provides a means of converting a stereo speaker representation to a binaural headphone representation, with the ability to include early reflections and late reverberation. In addition, for object-based audio content, virtual acoustic attributes, in particular the (relative) late reverberation level and/or the level, spectral, and timing characteristics of one or more early reflections, can be controlled for each object.

Embodiments are directed to eliminating artifacts and/or improving playback quality and preserving artistic intent by using metadata that controls playback by one or more playback systems. In particular, embodiments include metadata with object, channel, or hybrid signal representations that enhance playback quality when the layout of the playback system does not match the layout envisaged during content creation. As such, the application and/or effect resulting from the application of the metadata will depend on the intended and actual rendering systems.

Binaural pre-rendered content played through loudspeakers

As described in the background section, loudspeaker playback of binaural pre-rendered content may result in an unnatural timbre due to the spectral labels inherent in HRIR or BRIR being applied twice: once during pre-rendering and again during playback. in an acoustic environment. In addition, each playback of the binaural pre-rendered content will inherently include azimuth location marks also applied twice, which causes spatial mismapping and location errors.

In FIG. 2 shows this embodiment of processing 20. Channel or object 21 is initially convolved 22 with HRIR 23 before encoding 25. Thus, before encoding, channel or object based content is subjected to simulated speaker playback through HRIR or BRIR processing. Subsequently, the processed signal is encoded 25, decoded 26 and reproduced by loudspeakers 27, which introduces the aforementioned artifacts.

Spectral artifacts resulting from the dual application of the acoustic path from the speakers to the eardrums can be at least partially compensated for by applying a frequency-dependent gain or attenuation during decoding or playback. These gain or attenuation parameters may subsequently be encoded and included in the content. For headphone playback, these parameters can be discarded, while for speaker playback, coded gains are applied to the signals before playback.

One embodiment of a suitable sequential data processing path 30 is shown in FIG. 3. In this scheme, when playback is for loudspeakers, gain metadata is precalculated 31 when the presentation is created. The metadata is encoded with binaural processed signals. During decoding, the metadata information is also decoded 32. It is then used to apply a gain factor 33 to the decoded signal to reduce the significance of artifacts. For headphone playback, on the other hand, steps 31-33 are not required (excluded), and the decoded information can be applied directly to headphone playback.

Implementation example

In one implementation, to calculate gain metadata 31, input signals xi[n] with discrete time index n and input index i are analyzed in time-frequency tiles. To compile time-frequency mosaics, each of the input signals x _i [n] can be divided into time frames, and each frame, in turn, can be divided into frequency bands. The frequency bands can be obtained, for example, by means of a filter bank such as a quadrature mirror filter (QMF), a discrete Fourier transform (DFT), a discrete cosine transform (DCT), or any other means for dividing the input signals into a plurality of frequency bands. The result of this transformation is that the input signal xi[n] for input data with index i and discrete-time index n is represented by subband signals xi[k,b] for slot (or frame) k and subband b. The short-term energy in the frequency-time mosaic (K,B) has the form:

- 7 042232 σ 2. (К, В) = х j [k, b] xf [k, b], bl ev, k where B, K are sets of frequency (b) and time (k) indices corresponding to the required mosaic.

The representation of binaural signals yl[n], y _r [n] in the discrete time domain for the left and right ear, respectively, has the form:

Y1Y = ^X; [u] * bc [η] y _r W = Y _t Xj [η] * h _r j [n] i

where hl,i, h _r ,i - HRIR or BRIR corresponding to the input index i for the left and right ears, respectively. In other words, a pair of yl[n], y _r [n] binaural signals can be created by combining the convolution and summation of the input i. Subsequently, these binaural signals can be converted into time-frequency mosaics using the same process as the process applied to the xi[k, b] signals. Thus, for given binaural signals in the frequency domain, the short-term energy in the frequency-time mosaic (K, B) can be calculated as:

OU. (K, B) \u003d udkL]y / [k, b]

ЪL ev, k

Further, based on the conservation of energy in each frequency-time mosaic summed over input objects i in the numerator and over binaural signals j in the denominator, gain metadata w(K, B) can be compiled:

The metadata w(K,B) may then be quantized, encoded and included in the audio codec bitstream. The decoder will then apply the metadata w(K, B) to frame K and band B of both signals y _l and y _r (input representation) to obtain an output representation. This use of common metadata w(K, B) applied to both yl and y _r ensures that the stereo balance of the input representation is preserved.

In addition to the above-described method in which the binaural signals yl[n], y _r [n] are created by convolution in the time domain, the binaural presentation process is also applicable in the frequency domain. In other words, instead of initially calculating the binaural signals yl[n], y _r [n] in the time domain, you can instead convert the input signals xi[n] into a frequency domain representation and apply a convolution process with HRIR in the frequency domain to generate a representation of the binaural signals yj[k, b] in the frequency domain, for example, using fast convolution methods in the frequency domain. With this approach, the frequency domain representation of the binaural signals yj[k, b] is obtained without the need to generate these signals in the time domain, and it does not require the use of a filter bank or transformation to be applied to the binaural signals in the time domain.

Headphone-playable stereo content, including anechoic binaural presentation

In this implementation, a stereo signal intended for speaker playback is encoded in combination with side data to facilitate playback of the signal for speakers through headphones. Given a set of input features or channels xi[n], then a set of speaker signals zs[n] is typically generated by amplitude panning gains gi,s representing the gain of feature i at speaker s:

^z sW = 2 g _Ls X|[n]

For channel-based content, the gains gi, _s of the amplitude panning gains are typically constant, while for object-based content in which the estimated position of the object is represented by time-varying metadata, the gains will correspondingly be time-varying.

Given signals z _s [n] to be encoded and decoded, then a set of coefficients w must be found such that, if these coefficients are applied to signals z _s [n], then the resulting modified signals yl, y _r , composed as

- 8 042232 exactly matched the binaural representation of the original input signals xi[n] according to the expressions ⁼ ^ ^W ^ ^Z s

Y1Y = 2 _Xi [η] * h _y [n]

^The coefficients w _can be found by minimizing the ^L2 -criterion E between the required and actual binaural _{representations}

A solution to minimize the error E can be obtained using closed-form solutions, gradient descent methods, or any other suitable iterative method for minimizing the error function. As one example of such a solution, one can write the various stages of representation in a matrix notation:

Y = XH

Z=XG

Y=XGW=ZW

This matrix notation is based on a single-channel frame containing N discrete values represented as a single column x, = and matrices as combinations of several channels i = {1, ..., I}, each of which is represented in the matrix by a single column vector

The solution for W that minimizes E is

W = (G^X^XG + el) ^-1 G*X*XH where (*) is the complex conjugate transposition operator,

I is the identity matrix and e is the regularization constant.

This solution differs from the gain-based approach in that the Y signal is generated by a matrix rather than a W scalar applied to the Z signal, which allows for the possibility of having crossover terms (e.g., a second Y signal being (partially) reconstructed from the first Z signal) .

Ideally, the coefficients w are determined for each time-frequency tile to minimize the error E in each time-frequency tile.

In the above fragments of the description, the criterion of the minimum standard error (L2-criterion) was used to determine the matrix coefficients. Without loss of generality, instead of or in addition to the principle of minimum mean square error, other well-known criteria or methods for calculating matrix coefficients can similarly be used. For example, matrix coefficients can be computed using higher order terms or by minimizing an Ll test (eg, Least Absolute Deviation test). In addition, a variety of techniques may be used, including non-negative decomposition or optimization techniques, non-parametric estimators, maximum likelihood estimators, and the like. In addition, matrix coefficients can be calculated using iterative or gradient descent processes, interpolation methods, heuristic methods, dynamic programming, machine learning, fuzzy optimization, simulated hardening, or closed-form solutions and analysis-by-synthesis techniques can be used. Last but not least, the estimation of matrix coefficients can be limited in various ways, for example, by limiting the range of values, regularizing terms, superposition of energy conservation requirements, and so on.

- 9 042232

In practical applications of HRIR or BRIR hy,^ will include frequency dependent delays and/or phase shifts. Accordingly, the coefficients w may be complex-valued with an imaginary component substantially different from zero.

One embodiment of data processing in this embodiment is shown at 40 in FIG. 4. Audio content 41 is processed by a hybrid complex quadrature mirror filter (HCQMF) analysis unit 42 into subband signals. Then, 43 HRIRs 44 are applied to the output data of the filter bank to generate binaural Y signals. In parallel, the input data is presented 45 for playback by loudspeakers, resulting in Z loudspeaker signals. In addition, coefficients (or weights) w are calculated 46 from the Z speaker signals and the Y binaural signals and included in the core encoder bitstream 48 . Various base encoders can be used, such as, for example, MPEG-1 Layer 1, 2, 3, disclosed in Brandenburg, K., & Bosi, M. (1997). Overview of MPEG audio: Current and future standards for low bit-rate audio coding. Journal of the Audio Engineering Society, 45(1/2), 4-21 or Riedmiller, J., Mehta, S., Tsingos, N., & Boon, P. (2015). Immersive and Personalized Audio: A Practical System for Enabling Interchange, Distribution, and Delivery of NextGeneration Audio Experiences. Motion Imaging Journal, SMPTE, 124(5), 1-23, which are incorporated by reference. If the core encoder does not have the ability to use subband signals as input, these subband signals may first be converted to the time domain using a synthesis hybrid complex quadrature mirror filter (HCQMF) block 47 .

On the decoding side, if the decoder is capable of playing back headphones, the coefficients are extracted 49 and applied 50 to the core decoder signals before HCQMF synthesis 51 and playback 52. If the core encoder does not produce signals in the HCQMF domain, an optional HCQMF analysis filter bank 54 may be required as shown in FIG. 4. Thus, the signals encoded by the core encoder are intended to be reproduced by the loudspeakers, while the conversion coefficients of the loudspeaker signals to binaural signals are determined in the encoder and applied in the decoder. The decoder can be further equipped with user equalization functionality, and thus, in headphone playback mode, the user can choose to play normal speaker signals rather than binaurally processed signals on the headphones. In this case, the decoder ignores the weights. Finally, if the decoder is configured for loudspeaker playback, the weights can be ignored and the base decoder signals can be played back through the loudspeakers either directly or after upmixing or downmixing to suit the loudspeaker playback system layout.

It will be apparent that the methods described in the preceding paragraphs are not limited to quadrature mirror filter banks, as other filter bank designs as well as windowed short-term discrete Fourier transforms can be used with equal success.

This scheme has various advantages over traditional approaches. These advantages include: 1) the complexity of the decoder is only marginally greater than the complexity of conventional stereo playback, since the addition to the decoder consists only of a simple (time and frequency dependent) matrix driven by bitstream information; 2) This approach is suitable for both channel-based and object-based content, and is independent of the number of objects or channels present in the content; 3) HRTFs become encoder settings, i.e. they may be modified, improved, changed or adapted at any time, regardless of decoder compatibility. For decoders present in the art, HRTFs can still be optimized or customized according to user requirements without the need to modify the processing steps on the decoder side; 4) The bit rate is extremely low compared to the bit rates required for multi-channel or object-based content because only a small number of speaker signals (typically one or two) with additional data needs to be sent from the encoder to the decoder (with low bit rate) for coefficients w; 5) The same bit stream can be correctly reproduced by loudspeakers and headphones; 6) The bitstream can be composed in a scalable manner; if, in the context of a particular service, only loudspeakers are guaranteed at the endpoint, then the coefficients w can be removed from the bitstream without affecting the traditional loudspeaker representation; 7) Features of advanced codecs that apply to speaker presentations, such as volume control, dialogue enhancement, etc., will continue to work as intended (when played back by speakers); 8) By scaling the w factors, the loudness of the binaural presentation can be processed independently of the loudspeaker reproduction loudness;

- 10 042232

9) Listeners using headphones can choose to listen to a binaural or traditional stereo performance instead of being forced to listen to one or the other.

Extension with early reflections

Often, in order to increase the realism of the binaural presentation, it is necessary to include one or more early reflections in the binaural signals, used due to the presence of a floor, walls or ceiling. If the reflection is of a specular nature, then it can itself be interpreted as a binaural presentation, in which the corresponding HRIRs contain the effect of surface absorption, increased delay, and reduced overall level due to increased acoustic path length from the sound source to the eardrums.

These properties can be registered with a modified circuit, such as the circuit shown 60 in FIG. 5 which is a modification of the circuit shown in FIG. 4. In encoder 64, W coefficients are determined for (1) reconstructing the anechoic binaural representation from the speaker representation (WY coefficients) and (2) for reconstructing the binaural reflection representation from the loudspeaker representation (WE coefficients). In this case, the anechoic binaural presentation is determined by HRIR H _a of the binaural presentation, resulting in a pair of Y anechoic binaural signals, while the early reflection is determined by HRIR H _e , resulting in a pair of early reflection signals E. In order to enable parametric early reflection reconstruction from the mixed stereo signal, it is important that the delay caused by the longer early reflection path length be removed from HRIR H _e in the encoder and that this particular delay be applied in the decoder.

The decoder will generate a pair of anechoic signals and a pair of early reflection signals by applying coefficients W (W _Y ; WE) to the loudspeaker signals. To simulate a longer path length for early reflection, the early reflection is subsequently processed by a delay step 68 . The block delay parameter 68 may be included in the encoder bitstream, or may be a user-defined parameter, or may be made independent of the simulated acoustic environment or dependent on the actual acoustic environment in which the listener is located.

Late reverb extension

A late reverberation algorithm such as a feedback delay network (FDN) can be used to include a late reverberation simulation in the binaural representation. The FDN takes as input one or more objects or channels and produces (in the case of a binaural reverb) two late reverb signals. The output of the decoder (or the result of its downmix) can be used as input to the FDN in the traditional algorithm. This approach has a significant drawback. In many use cases, it may be necessary to adjust the amount of late reverb for each object. For example, reducing the amount of late reverb increases the intelligibility of dialogue.

In one alternative embodiment, a per-object or per-channel reverberation amount control can be created in a manner similar to how an anechoic binaural representation or a binaural early reflection representation is composed from a stereo downmix signal.

As shown in FIG. 6, various modifications to the previous circuits can be made to further accommodate the late reverberation. The encoder 81 calculates 82 the input signal F FDN, which may be a weighted combination of the input data. These weights may be content dependent, such as manual markup during content creation, or automatic classification using intelligent media analysis algorithms. The FDN input signal itself is discarded by the weight estimator 83, but the coefficient data WF enabling estimation, reconstruction, or approximation of the FDN input signal from the speaker representation is included 85 in the bitstream. At the decoder 86, the FDN input signal is reconstructed 88, processed by the FDN itself, and included 89 in the binaural output signal for the listener 91.

In addition, the FDN can be composed such that it allows multiple (two or more) inputs, and thus the spatial magnitudes of the input signals are stored at the output of the FDN. In these cases, coefficient data is included in the bitstream, making it possible to evaluate each FDN input signal from the speaker representation.

In this case, it may be necessary to control the determination of the spatial location of the object or channel relative to the FDN input.

In some cases, it may be possible to generate late reverb simulation inputs (eg, FDN) in response to parameters present in the data stream for another destination (eg, parameters not specifically designed to be applied to the underlying signals to generate FDN inputs). For example, in one exemplary dialog amplification system, a dialog signal is reconstructed from a set of base signals by applying to those base signals

- 11 042232 dialogue gain parameter signals. The dialogue signal is then enhanced (eg, amplified) and mixed back into the base signals (thus amplifying the dialogue components relative to the rest of the base signals). As described above, it is often desirable to compose the input FDN signal in such a way that it does not contain dialogue components. Thus, in systems where dialogue gain parameters are already available, it is possible to reconstruct the desired FDN input without dialogue (or at least dialogue suppressed) by first reconstructing the dialogue signal from the base signal and dialogue gain parameters, and then subtracting (eg, canceling) the dialog signal from the underlying signals. In such a system, specialized parameters for reconstructing the FDN input signal from the base signals may not be necessary (since dialogue gain parameters may be used instead) and thus may be omitted, resulting in a reduction in the required parameter data rate without loss of functionality. opportunities.

Combination of early reflections and late reverbs

Although the preceding sections have indicated extensions of the anechoic performance through early reflection(s) and late reverberation, combinations of these are also possible. For example, the system may contain: 1) WY coefficients to determine the anechoic representation from the loudspeaker representation; 2) additional WE coefficients to determine a certain number of early reflections from the loudspeaker view; 3) additional WFs to determine one or more late reverb inputs from the loudspeaker view, allowing you to control the amount of late reverb for each object.

Anechoic performance as first performance

While using the speaker representation as the first representation to be encoded by the core encoder has the advantage of being backward compatible with decoders lacking the ability to interpret or process the transform data w, said first representation is not limited to the speaker reproduction representation. In FIG. 7 shows a schematic representation of a method 100 for encoding and decoding audio content 105 for playback by headphones 130 or speakers 140. Encoder 101 receives input audio content 105 and processes these signals with an HCQMF filter bank 106 . Subsequently, based on the HRIR/HRTF database 104, an anechoic Y representation is generated by the HRIR convolution element 109. In addition, an element 108 calculating and applying the speaker panning matrix G generates a Z representation for the speakers. In addition, the element 107 generates the input mixed signal F FDN.

The anechoic signal Y is optionally transformed into the time domain using the HCQMF synthesis filter bank 110 and encoded by the core encoder 111. The transform estimator 114 calculates parameters WF(112) enabling the reconstruction of the FDN input signal F from the anechoic Y representation, as well as the parameters W _Z (113) to reconstruct the Z representation for loudspeakers from the anechoic Y representation. Parameters 112 and 113 are included in the core encoder bitstream. Alternatively or in addition, although not shown in FIG. 7, the transform estimator may compute parameters W _E enabling reconstruction of the early reflection signal E from the anechoic representation Y.

The decoder has two modes of operation, shown as decoder mode 102 intended for listening 130 through headphones and decoder mode 103 intended for playback 140 by loudspeakers. In the case of headphone playback, the core decoder 115 decodes the anechoic Y representation and decodes the transform parameters WF. Subsequently, the WF transform parameters are applied to the anechoic representation of Y by a matrix block 116 to obtain an estimated FDN input signal, which is subsequently processed by the FDN 117 to obtain a late reverb signal. This late reverb signal is mixed with an anechoic Y representation by an adder 150 followed by an HCQMF synthesis filter bank 118 to obtain a headphone representation 130. If W _E parameters are also present, the decoder may apply these parameters to the Y anechoic representation to obtain an early reflection estimate, which is subsequently processed through a delay and mixed with the Y anechoic representation.

In the case of speaker playback, the decoder operates in mode 103, in which the base decoder 115 decodes the anechoic Y representation as well as the W _Z parameters. Subsequently, a matrixing step 116 applies the W _Z parameters to the anechoic Y representation to obtain an estimate, or approximation, of the Z representation for the loudspeakers. This signal is later converted to the time domain by the HCQMF synthesis filter bank 118 and reproduced by the speakers 140.

Finally, it should be noted that the system shown in FIG. 7 may optionally be operated in the absence of defining and transmitting parameters W _Z . In this mode of operation, the Z representation for loudspeakers cannot be generated from the anechoic Y representation. However, due to

- 12 042232 that the parameters WE and/or W _F are determined and transmitted, a headphone representation containing early reflections and/or late reverb components from the anechoic representation can be generated from the anechoic representation.

Multi-channel presentation for loudspeakers

Those skilled in the art will appreciate that the first representation of the playback stream encoded in the encoder may be a multi-channel representation, such as a surround or immersive presentation for speakers such as 5.1, 7.1, 7.1.4, etc. . Discussed above, for example, with reference to FIG. 4, embodiments of the invention in which the second representation of the playback stream is a stereo representation will operate in a similar manner, although the size of the matrices will be corrected. For example, while a 2x2 parameter matrix is sufficient to convert from one stereo view to another stereo view, a 5x2 matrix is required to convert a five-channel surround view to a stereo view, and a 5x2 matrix is required to convert from a 5.1 surround view (five full-bandwidth channels). and the low-frequency effects (LFE) channel into a stereo representation - a 6x2 matrix. As a consequence, the amount of side information needed to represent the transform parameters will increase along with the number of channels in the speaker representation, and the computational complexity of the decoding process will also increase accordingly.

To avoid or minimize this increase in computational complexity when converting a first representation from M1 channels to a second representation from M2 channels, where M1>M2, for example, when a surround or immersive presentation for loudspeakers is converted to a binaural stereo representation, it may be advantageous to determine the transformation parameters before downmixing the first view into an intermediate view. For example, a 5.1 surround presentation may be downmixed to stereo for 2.0 speakers.

In FIG. 8a shows an encoder 200 in which the audio content 201 is represented by a presentation unit 202 into a 5.1 speaker surround representation S, which is encoded by a base encoder 203. ) representation of Z. For example, the left channel, Z (ZL), can be expressed as the weighted sum of the left channel (SL), left side channel (SLS), center channel (SC), and low frequency effects channel (SLFE) of the surround representation S, according to with the following equation:

ZL = (SL + a*SC + b*SLS + c*SLFE) where a, b and c are the respective constants, e.g. a=b=sqrt(0.5)=0.71 and c=0.5 .

Audio content is also input to a binaural presentation unit 205 configured to present an anechoic binaural Y signal. A parameter calculation unit 206 receives an anechoic Y signal and a downmixed stereo Z signal, and calculates stereo-to-anechoic conversion parameters W _Y . Compared to the above-described FIG. 4, the presentation block 202 is a multi-channel version of the presentation block 45, since the output data in both cases is delivered to the base encoder 203/48. Blocks 205 and 206 are basically identical to blocks 43 and 46.

In addition, the encoder may also include a block 207 (corresponding to block 82 shown in FIG. 6) for representing the input FDN signal, in which case block 206 can be configured to calculate also a set of W _F FDN parameters (corresponding to block 83 in FIG. 6).

In FIG. 8b shows a decoder 210 in which the base decoder 211 receives and decodes a 5.1 surround representation S and parameter sets W _Y and W _F . The surround representation S is converted to the downmixed two-channel (stereo) signal Z by the downmixer 212, operating in the same manner as its equivalent 204 in the encoder. The first matrixer 213 applies the W _Y parameters to the Z stereo representation to create a reconstructed anechoic Y signal. The second matrixer 214 applies the W _F parameters to the Z stereo representation to create the reconstructed FDN input signal. The FDN input signal is used in the FDN 215 to create a late reverb signal, which is added 216 to the reconstructed anechoic Y signal to create a binaural output. It should be noted that the processing in blocks 213-216 is similar to that of decoder 86 shown in FIG. 6.

For lower bit rates, it is known to use parametric methods to transmit a 5.1 representation using a downmixed 2.1 signal and a set of communication parameters, see for example ETSI Publication TS 103 190-1 V1.2.1 (2015- 06). In such a system, the underlying decoder actually performs the upmix to produce a decoded 5.1 representation. If the embodiment

- 13 042232 shown in FIG. 8b is implemented in such a decoder, the result will be the decoder shown in FIG. 9a.

It should be noted that the base decoder 311 shown in FIG. 9a includes an upmixer 312 for upmixing a 2.1 view into a 5.1 view. The 5.1 presentation is then downmixed to the 2.0 presentation by the downmixer 212 as shown in FIG. 8b.

However, in this context, if the 2.1 representation is already included in the bitstream, the 5.1 upmix is not required and can be omitted to simplify the decoder. Such a simplified decoder is shown in FIG. 9b. Here, the base decoder 411 only decodes the 2.1 representation. This representation is received by a simplified downmixer 412, configured to convert the 2.1 representation to a 2.0 representation according to the equations:

Lo = a*L + b*LFE

Ro = a*R + b*LFE where L, R and LFE denote the full bandwidth left and right channels and the low frequency effects channel of the decoded 2.1 presentation, with a and b being the respective constants affecting the upmix and downmix executed by modules 312 and 212 shown in FIG. 9a.

The process described in FIG. 9a and 9b assumes a downmix of a 2.1 signal and associated communication parameters. A similar approach can be used in a system that uses, for example, a downmixed 3.1 signal and related communication parameters. Alternatively, the system shown in FIG. 8a and 8b can also carry additional ancillary information to allow upmixing of a 5.1 representation to an object-based representation, as discussed in ETSI TS 103 190-1 V1.2.1 (2015-06).

Interpretations

Reference herein to one embodiment, some embodiments, or one of the embodiments means that a particular feature, structure, or characteristic described in connection with a given embodiment is included in at least one embodiment of the present invention. Therefore, the occurrences of the phrase in one embodiment, in some embodiments, or in one of the embodiments at various places in this specification may, but need not, refer to the same embodiment of the invention. In addition, specific features, designs, or characteristics may be combined in one or more embodiments in any suitable manner, as will be apparent from this description to those of ordinary skill in the art.

Within the framework of this description, the use of ordinal numbers first, second, third, etc. for describing an ordinary object indicates solely that reference is made to various examples of similar objects, and are not intended to indicate that the objects described in this way must be in a given sequence in time, space, rank, or in any other way. .

In the following claims and in this description, any of the terms containing, consisting of, or which contains is a non-limiting term, which means the inclusion of at least the following elements/features, but not the exclusion of the rest. Therefore, the term containing when used in the claims should not be interpreted as limiting the means or elements or steps listed after it. For example, the scope of the expression device containing A and B should not be limited to devices containing only elements A and B. If any of the terms used in this description including or which includes also is a non-limiting term that also means the inclusion of at least /distinguishing features following this term, but not excluding the rest. Thus, including is synonymous and means containing.

As used herein, the term exemplary is used in the sense of presenting examples, as opposed to indicating a property. Those. an exemplary embodiment is an embodiment provided as an example, but not necessarily one of the embodiments of an exemplary property.

It should be understood that in the above description of exemplary embodiments of the invention, various features of the invention are sometimes grouped into one embodiment, figure, or description thereof in order to select the optimal way of describing and to provide an understanding of one or more different aspects of the invention. This mode of disclosure, however, should not be interpreted as reflecting the intention that the claimed invention requires more features than the number of features that are expressly listed.

- 14 042232 in each claim. Instead, as the following claims reflect, the features of the invention lie in less than all of the features of the one embodiment described above. Therefore, the claims following the Detailed Description section are thus irrevocably incorporated into this Detailed Description section, with each claim on its own representing a separate embodiment of the present invention.

In addition, although some embodiments of the invention described in this specification include some and not other features included in other embodiments of the invention, combinations of features from various embodiments of the invention are intended to be within the scope of the invention and forming other embodiments of the invention, as will be understood by those skilled in the art. For example, in the following claims, any claimed embodiments of the invention may be used in any combination.

In addition, some embodiments of the invention are described herein as a method or combination of elements of a method that can be implemented by a computer system processor, or other means for performing such a function. Thus, the processor, with the necessary instructions for carrying out said method or method element, forms the means for carrying out the method or method element. In addition, the device embodiment element described herein is an example of the means for carrying out the function performed by the element for carrying out the invention.

In the description provided herein, many specific details are set forth. However, it should be understood that embodiments of the invention may be practiced without these specific details. In other instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this disclosure.

Likewise, it should be noted that the term "linked" when used in the claims should not be interpreted as being limited to direct connections only. The terms bound and connected, along with their derivatives, may be used. It should be understood that these terms are not intended to be synonymous with each other. Thus, the scope of the expression device A associated with device B should not be limited to devices or systems in which the output of device A is directly connected to the input of device B. This means that there is a path between the output of device A and the input of device B, which can be path containing other devices or facilities. Connected can mean that two or more elements are in either direct physical or electrical contact, or that two or more elements are not in direct contact with each other but still cooperate or interact with each other.

Thus, while what are considered to be preferred embodiments of the invention are described herein, it should be understood by those skilled in the art that other and additional modifications may be made to them without departing from the spirit of the invention, and it is understood that all such changes and modifications are claimed to be included within the scope of the invention. For example, any of the above formulas are only examples of procedures that may be used. Functionality may be added to or removed from block diagrams, and operations may be interchanged between functional blocks. Steps may be added to or omitted from the methods described within the scope of the present invention.

Claims (15)

CLAIM 1. A method for encoding an input audio stream containing one or more audio components, characterized in that each audio component is associated with a spatial location, and the method includes the steps of presenting a first representation of the playback stream of the specified input audio stream, and the specified first representation of the playback stream is a set of M1 signals intended to be reproduced by the first sound reproduction system; representing a second playback stream representation of said input audio stream, wherein said second reproduction stream representation is a set of two signals to be reproduced by a second audio playback system, wherein one of the first reproduction stream representation and the second reproduction stream representation is a speaker representation, and wherein one of the first representation of the playback stream and the second representation of the playback stream is an echo or anechoic binaural presentation; determining a set of conversion parameters suitable for converting an intermediate representation of the playback stream, which is a set of M3 signals, to approximately - 15 042232 a second representation of the playback stream, wherein the set of transformation parameters forms a gain matrix of size M3x2, which, when applied directly to the intermediate representation of the playback stream by matrix multiplication, forms the specified approximation of the second representation of the playback stream, and wherein the intermediate representation of the playback stream is one of the first representing the playback stream and the downmix result of the first representation of the playback stream; moreover, the transformation parameters are determined by minimizing the amount of difference between the second representation of the playback stream and the approximation of the second representation of the playback stream formed by applying the transformation parameters to the intermediate representation of the playback stream; and encoding the first representation of the playback stream and the specified set of transformation parameters for transmission to the decoder. 2. Method according to claim 1, characterized in that said conversion parameters are time-varying and/or frequency-dependent. 3. The method according to any of the previous paragraphs, characterized in that M1=2. 4. The method according to any one of claims 1 to 2, wherein M1>2, M3=2, and wherein the method further comprises generating an intermediate playback stream representation by downmixing the first playback stream representation into a two-channel representation. 5. A method for decoding representations of a playback stream from a data stream, the method comprising the steps of receiving and decoding a first presented representation of the playback stream, said first presented representation of the playback stream being a set of M1 signals to be reproduced by a first audio playback system; receiving and decoding a set of transformation parameters suitable for converting an intermediate representation of the playback stream, which is a set of M3 signals, to an approximation of the second presented representation of the playback stream, while said second presented representation of the playback stream is a set of two signals intended to reproduce the second by an audio playback system, wherein the set of transformation parameters form an M3x2 gain matrix which, when applied directly to an intermediate representation of the playback stream by matrix multiplication, produces said approximation of the second representation of the playback stream; wherein one of the first presented representation of the playback stream and the second presented representation of the playback stream is a presentation for speakers, and wherein one of the first presented representation of the playback stream and the second presented representation of the playback stream is an echo or anechoic binaural presentation, and wherein the intermediate representation of the stream the playback is one of the first presented representation of the playback stream and the downmix result of the first presented representation of the playback stream; wherein the transform parameters are determined by the encoder to ensure that the amount of difference between the second represented representation of the playback stream and an approximation of the second represented representation of the playback stream formed by applying the transform parameters to the intermediate representation of the playback stream is minimal; and applying said transformation parameters to said intermediate representation of the playback stream to obtain said approximation of the second represented representation of the playback stream. 6. Method according to claim 5, characterized in that said conversion parameters are time-varying and/or frequency-dependent. 7. Method according to claim 5, characterized in that M1=2. 8. The method according to any one of claims 5, 6, characterized in that M1>2, M3=2, and the method further includes generating an intermediate playback stream representation by downmixing the first presented playback stream representation into a two-channel representation. 9. The method of claim 8, wherein the first presentation of the playback stream is an ambient or immersive presentation, such as a 5.1, 7.1, or 7.1.4 presentation. 10. The method of claim 8, wherein the first representation of the playback stream presented is a 2.1 representation, and wherein said downmixing step comprises converting the 2.1 representation to a 2.0 stereo representation according to the equations Lo = a*L + b*LFE Ro = a*R + b*LFE - 16 042232 where L, R and LFE denote the full bandwidth left and right channels and the low frequency effects channel of the decoded 2.1 presentation, with a and b being the respective constants. 11. The method of claim 5, wherein the approximation of the second represented representation of the playback stream is an anechoic binaural representation, and wherein the method further comprises receiving and decoding one or more additional sets of transform parameters suitable for converting the intermediate representation of the playback stream into one or more inputs to the process of simulating the acoustic environment; applying one or more additional sets of transformation parameters to the intermediate representation of the playback stream to generate one or more inputs to the process of simulating the acoustic environment; applying one or more acoustic simulation process inputs to one or more acoustic environment simulation processes to produce one or more simulated acoustic environment signals; and combining one or more simulated acoustic environment signals to approximate a second represented playback stream representation. 12. The method of claim 5, further comprising receiving and decoding one or more dialog signal evaluation parameter sets suitable for converting an intermediate representation of the playback stream into one or more evaluation dialog signals; applying one or more sets of dialogue evaluation parameters to the intermediate representation of the playback stream to generate one or more evaluation dialogue signals; subtracting one or more estimated dialogue signals from the intermediate representation of the playback stream to obtain one or more signals with suppressed dialogue; applying one or more dialogue-suppressed signals to one or more acoustic simulation processes to obtain one or more simulated acoustic environment signals; and combining one or more simulated acoustic environment signals to approximate the second represented representation of the playback stream. 13. An encoder for encoding an input audio stream containing one or more audio components, each audio component is associated with a spatial location, and the encoder contains a first representation unit for representing a first representation of the playback stream of the specified input audio stream, while the specified first representation of the playback stream is a set of M1 signals to be reproduced by the first sound reproduction system; a second presentation unit for presenting a second reproduction stream representation of said input audio stream, said second reproduction stream representation being a set of two signals to be reproduced by the second audio playback system, wherein one of the first reproduction stream representation and the second reproduction stream representation is a speaker presentation, wherein one of the first playback stream presentation and the second playback stream presentation is an echo or anechoic binaural presentation; a transformation parameter determination unit for determining a set of transformation parameters suitable for converting the intermediate representation of the playback stream, which is a set of M3 signals, into an approximation of the second representation of the playback stream, wherein the set of transformation parameters forms a gain matrix of size M3x2, which, when applied directly to the intermediate representation of the playback stream by matrix multiplication forms the specified approximation of the second representation of the playback stream, and the intermediate representation of the playback stream is one of the first representation of the playback stream and the result of downmixing the first representation of the playback stream, and wherein the transformation parameters are determined by minimizing the amount of difference between the second representation of the playback stream and approximation of the second representation of the playback stream formed by applying the parameter into transformations to an intermediate representation of the playback stream; and an encoding unit for encoding the first representation of the playback stream and the specified set of transformation parameters for transmission to the decoder. 14. A decoder for decoding representations of the playback stream from the data stream, the decoder comprising a base decoder unit configured to receive and decode the first presented representation of the playback stream, when - 17 042232 this first presented representation of the playback stream is a set of M1 signals intended to be reproduced by the first sound reproduction system, and receiving and decoding a set of conversion parameters suitable for converting the intermediate representation of the playback stream, which is a set of M3 signals, into an approximation of the second represented representation of the playback stream, wherein said second represented representation of the playback stream is a set of two signals intended to be reproduced by the second sound reproduction system, the set of transformation parameters forming a gain matrix of size M3x2, which, when applied directly to the intermediate representation of the playback stream matrix multiplication forms the specified approximation of the second representation of the playback stream, while one of the first representation of the stream in and the second presented representation of the playback stream is a speaker view, and wherein one of the first presented representation of the playback stream and the second presented representation of the playback stream is an echo or anechoic binaural representation, and wherein the intermediate representation of the playback stream is one of the first presented representation of the playback stream a playback stream and a downmix result of the first represented representation of the playback stream, wherein the transform parameters are determined by the encoder to ensure that the amount of difference between the second presented representation of the playback stream and an approximation of the second representation of the playback stream formed by applying the transform parameters to the intermediate representation of the playback stream is minimal; and a matrix multiplier for applying said transformation parameters to said intermediate representation of the playback stream to obtain said approximation of the second represented representation of the playback stream. 15. An internal storage medium storing an encoder computer program product comprising computer program code segments which, when executed in the computer processor, cause the computer processor to execute the method of any one of claims 1-12.