ES2708581T3 - Multichannel audio coding using complex prediction and window shape information - Google Patents

Abstract

Audio decoder for decoding a multichannel encoded audio signal (100), the encoded multichannel audio signal (100) comprising a first encoded combination signal (104) generated on the basis of a combination rule to combine a first audio signal channel and a second channel audio signal of a multi-channel audio signal, a residual encoded prediction signal (106) and prediction information (108), comprising: a signal decoder (110) for decoding the first signal of coded combination (104) for obtaining a first decoded combination signal (112) and for decoding the residual encoded prediction signal (106) to obtain a decoded residual signal (114); and a decoder calculator (116) for computing a decoded multichannel audio signal having a first decoded channel signal (117), and a second decoded channel signal (118) using the decoded residual signal (114), the decoding information prediction (108) and the first decoded combination signal (112), such that the first decoded channel signal (117) and the second decoded channel signal (118) are at least approximations of the first channel signal and the second signal channel of the multichannel audio signal, wherein the prediction information (108) comprises an imaginary portion other than zero, wherein the decoder calculator (116) comprises: a predictor (1160) for applying the information of prediction (108) to the first decoded combination signal (112) or to a signal (601) from the first decoded combination signal (112) to obtain a prediction signal (1163); a combination signal calculator (1161) for computing a second combination signal (1165) by combining the decoded residual signal (114) and the prediction signal (1163); and a combiner (1162) for combining the first decoded combination signal (112) and the second combination signal (1165) to obtain the decoded multi-channel audio signal having the first decoded channel signal (117) and the second signal of decoded channel (118), wherein the predictor (1160) comprises a real-to-imaginary converter (1160) to estimate (1160a) an imaginary spectrum of the first decoded combination signal (112) using a real part of the first signal of decoded combination (112) directly in the frequency domain using a two dimensional filtering, the real part of the first decoded combination signal (112) being subjected to a window switching, wherein the predictor (1160) is configured to multiply a imaginary part (601) of the first decoded combination signal (112) by the imaginary part of the prediction information (108) to obtain at least a part of the pre-signal diction (1163), and wherein the predictor (1160) is configured to receive a window shape information (109) and to use different filter coefficients by the imaginary real converter (1160) to calculate the imaginary spectrum of the first decoded combination signal (112), wherein the different filter coefficients depend on different window forms that are indicated by the window shape information (109), wherein the filter coefficients that are used by the predictor (1160) ) depend on a complete window, and where a set of filter coefficients is required for each type of window and for each window transition.

Description

DESCRIPTION

Multichannel audio coding using complex prediction and window shape information

The present invention relates to audio processing and, in particular, to multi-channel audio processing of a multi-channel signal having two or more channel signals.

In the field of stereo or multichannel processing it is known to apply the so-called central / lateral stereo coding. According to this concept, a combination of the left or first audio channel signal and the right or second audio channel signal is formed to obtain a central or mono M signal. Additionally, a difference is formed between the left channel signal or first and the signal of right or second channel to obtain the lateral signal S. This method of central / lateral coding achieves a gain of significant coding, when the left signal and the right signal are very similar to each other, because the lateral signal It will become quite small. In general, a gain of coding of a stage of encoder entropfa / quantifier will be higher, when the range of values to be quantified / encoded by entropfa is smaller. Thus, for a PCM or an entropic arithmetic or Huffman-based encoder, the coding gain increases, when the lateral signal becomes smaller. However, there are certain situations in which the central / lateral encoding will not achieve a coding gain. The situation can occur when the signals in both channels are out of phase with each other, for example, by 90 °. Then, the central signal and the lateral signal may be in a fairly similar range and, therefore, the coding of the central signal and the lateral signal using the entropy encoder will not achieve a gain of coding and may even produce an increase of Bit rate. Therefore a selective central / lateral coding of frequency can be applied in order to deactivate the central / lateral coding in bands, where the lateral signal does not become smaller up to a certain degree with respect to, for example, the signal left original.

Although the lateral signal is zero, when the left and right signals are identical, achieving maximum coding gain due to the elimination of the lateral signal, the situation is again different when the central signal and the lateral signal are identical with respect to the conformation of the waveform, but the only difference between both signals lies in its general amplitudes. In this case, when it is additionally presumed that the lateral signal does not have a lag with respect to the central signal, the lateral signal increases significantly, even though, on the other hand, the central signal does not decrease too much with respect to its range of values. When said situation occurs in a certain frequency band, then the central / lateral encoding is deactivated again due to the lack of coding gain. The central / lateral coding can be applied selectively in frequency or it can be applied, as an alternative, in the time domain.

There are alternative techniques of multichannel coding, which are not based on a type of waveform method such as central / lateral coding, but which are based on parametric processing according to certain binaural signals. Said techniques are known according to the terms "binaural signal coding", "parametric stereo coding" or "coding of the MPEG environment". At this point certain signals are calculated for a plurality of frequency bands. These signals include level differences between channels, coherence measurements between channels, time differences between channels and / or phase differences between channels. These methods start from the premise that a multichannel impression felt by the listener is not necessarily based on the detailed waveforms of the two channels, but is based on the signals supplied in a selective frequency or in the exact information between channels. This means that, in a presentation machine, care must be taken to present multichannel signals that accurately reflect the signals, although the waveforms do not have a significant importance.

This method can be particularly complex in the case where the decoder has to apply a decorrelation processing in order to artificially create stereo signals, which are de-correlated with each other, despite the fact that all of said channels comes from one and the same downmix channel. The de-correctors for this purpose are complex, depending on their implementation and can introduce distortions, particularly in the case of portions of transient signals. Additionally, unlike waveform coding, the parametric coding method is a loss coding method, which inevitably results in a loss of information not only introduced by the typical quantization but also introduced by considering the binaural signals instead of the waveforms in particular. This method produces very low bit rates but may include quality commitments.

There are recent developments in unified speech and audio coding ( USAC) that is illustrated in Figure 7a. A core decoder 700 performs a decoding operation of the stereo signal encoded at the input 701, which may be a center / side encoded signal. The core decoder emits a central signal on line 702 and a side or residual signal on line 703. Both signals are transformed to a QMF domain by filter banks QMF 704 and 705. A decoder is then applied to the environment. MPEG 706 to generate a left signal of channel 707 and a right signal of channel 708. These signals of low band are subsequently entered into a spectral replication decoder bandwidth (spectral band replication, SBR) 709, which produces broadband left and right signals on lines 710 and 711, which are then transformed to a time domain by the synthesis filter banks QMF 712, 713 so that the left and right broadband signals L, R are obtained.

Figure 7b illustrates a situation in which the decoder of the MPEG 706 Environment performs a central / lateral decoding. Alternatively, the decoder block of the MPEG 706 Environment could perform a parametric decoding based on the binaural signal to generate stereo signals from a single mono signal of the core decoder. Naturally, the decoder of the MPEG 706 Environment could also generate a plurality of low band output signals to be input into the SBR 709 decoder block using parametric information such as level differences between channels, coherence measurements between channels or other such Parameters of information between channels.

When the decoder block of the MPEG 706 Environment performs the central / lateral decoding illustrated in Figure 7b, a real gain factor g and DMX / RES can be applied and L / R are down / residual and left / down mix signals / right, respectively, represented in the hybrid complex domain QMF.

The use of a combination of a block 706 and a block 709 only generates a small increase in computational complexity as compared to a stereo decoder used as a base, because the representation of the complex QMF of the signal is already available as part of the decoder of SBR. However, in a non-SBR configuration, the QMF-based stereo coding, proposed in the context of the USAC, results in a significant increase in computational complexity due to the necessary QMF banks, which in this example need banks. of analysis of 64 bands and banks of synthesis of 64 bands. These filter banks should only be added for the purposes of stereo coding.

In the USAC system of the developing MPEG, however, there are also coding modes at high bit rates where, in general, the SBR is not used.

The following documents are examples of multichannel audio (de) coding schemes with which a difference signal is estimated from a descending / adder / mono signal by means of a complex value prediction coefficient:

HEIKO PURNHAGEN ET AL: " Technical description of proposed Unified Stereo Coding in USAC", 90. MPEG MEETING ; 26 - 10 - 2009 - 30 - 10 - 2009; XIAN; (MOTION PICTURE EXPERT GROUP OR ISO / IEC JTC1 / SC29 / WG11 ), No. M16921, October 23, 2009 (October 23, 2009);

MAX NEUENDORF (EDITOR): "WD5 of USAC", 90. MPEG MEETING; 26 - 10 - 2009 - 30 - 10 - 2009; XIAN; (MOTION PICTURE EXPERT GROUP OR ISO / IECJTC1 / SC29 / WG11), No. N11040, December 8, 2009 (08 -12 - 2009), pages 1 - 146;

WO 2009/141775 A1 (KONINKL PHILIPS ELECTRONICS NV [NL]; SCHUIJERS ERIK G P [NL]) November 26, 2009 (26 - 11 - 2009).

In addition, reference is made to WO 2011/124608, published subsequently.

An object of the present invention is to provide an improved audio processing concept, which on the one hand produces a high coding gain and, on the other hand, produces good audio quality and / or reduced computational complexity.

This objective is achieved by an audio decoder according to claim 1, an audio encoder according to claim 13, an audio decoding method according to claim 17, an audio coding method according to the invention. claim 18 and a computer program according to claim 19. The present invention is based on the conclusion that a coding gain of the high-quality waveform coding method can be significantly improved by a prediction of a second one. combining signal using a first combination signal, where both combination signals come from the original channel signals using a combination rule such as the central / lateral combination rule. It has been found that this prediction information is calculated by a predictor in an audio encoder such that an optimization objective is met, only a small overload is incurred, but it achieves a significant decrease in the bit rate necessary for the side signal without losing any audio quality, because the inventive prediction is, however, a coding based on waveform and not a method of multichannel or stereo coding based on parameters. In order to reduce computational complexity, it is preferred to carry out an encoding in the frequency domain, wherein the prediction information comes from input data of the frequency domain in a selective form of bands. The conversion algorithm for converting the representation in the time domain into a spectral representation is preferably a process sampled quantically such as a modified discrete cosine transform ( MDCT) or a modified discrete sine transform (modified discrete sine transform, MDST), which is different from a complex transform due to the fact that only real values or only imaginary values are calculated, while in a complex transformation, real and complex values of a spectrum are calculated, achieving twice the oversampling.

Preferably a transform based on introduction and cancellation of overlap is used. The MDCT, in particular, is a transformation of these characteristics and allows a cross attenuation between subsequent blocks without any overload due to the well-known property of time domain overlapping ( TDCA), which is get through the overlap processing - add on the decoder side.

Preferably, the prediction information calculated in the encoder, transmitted to the decoder and used in the decoder comprises an imaginary part, which can advantageously reflect the phase differences between the two audio channels in arbitrarily selected quantities between 0 ° and 360 °. Computational complexity is reduced significantly when only a real value transform is applied or, in general, a transform that provides, either only a real spectrum or that only provides an imaginary spectrum. In order to use said imaginary prediction information, which indicates a lag between a certain band of the left signal and a corresponding band of the right signal, a converter from real to imaginary or, depending on the implementation of the transform , a complex to real converter is provided in the decoder in order to calculate a residual signal of prediction from the first combination signal, which has the phase rotated with respect to the original combination signal. This residual signal of prediction that has the phase rotated can then be combined with the residual signal of prediction transmitted in the bitstream to regenerate a lateral signal, which can finally be combined with the central signal to obtain the decoded left channel in a certain band and the right channel decoded in this band.

In order to increase the audio quality, the same converter from real to imaginary or the converter from complex to real, which is applied on the decoder side, is also implemented on the encoder side, when the residual signal of prediction is calculated in the encoder.

The present invention provides an advantage because it improves the audio quality and reduces the bit rate as compared to systems that have the same bitrate or have the same audio quality.

Additionally, advantages are obtained with respect to a computational efficiency of unified stereo coding which is useful in the USAC system of MPEG at high bit rates, where, in general, the SBR is not used. Instead of processing the signal in the complex hybrid domain QMF, these methods implement a predictive residual-base stereo encoding in the native MDCT domain of the underlying stereo transform encoder.

According to one aspect of the present invention, the present invention comprises an apparatus or method for generating a stereo signal by complex prediction in the MDCT domain, wherein the complex prediction is carried out in the MDCT domain using a transforming real to complex, wherein said stereo signal may be either a stereo signal encoded on the encoder side or may be, alternatively, a decoded / transmitted stereo signal, when the apparatus or method for generating the stereo signal is applied in the decoder side.

The preferred embodiments of the present invention are described below with respect to the accompanying drawings, in which:

Figure 1 is a diagram of an audio decoder;

Figure 2 is a block diagram of an audio encoder;

Figure 3a illustrates an implementation of the encoder calculator of Figure 2;

Figure 3b illustrates an alternative implementation of the encoder calculator of Figure 2;

Figure 3c illustrates a central / lateral combination rule to apply on the encoder side;

Figure 4a illustrates an implementation of the decoder calculator of Figure 1;

Figure 4b illustrates an alternative implementation of the decoder calculator in the form of a matrix calculator;

Figure 4c illustrates a central / lateral reverse combining rule corresponding to the combination rule illustrated in Figure 3c;

Figure 5a illustrates an embodiment of an audio encoder operating in the frequency domain, which is preferably a real value frequency domain;

Figure 5b illustrates an implementation of an audio decoder operating in the frequency domain; Figure 6a illustrates an alternative implementation of an audio encoder operating in the MDCT domain and using a real-to-imaginary transform according to an embodiment of the present invention;

Figure 6b illustrates an audio decoder operating in the MDCT domain and using a real-to-imaginary transform according to an embodiment of the present invention;

Figure 7a illustrates an audio post-processor using a stereo decoder and a later connected SBR decoder;

Figure 7b illustrates a central / lateral upmix matrix;

Figure 8a illustrates a detailed view in the MDCT block in Figure 6a;

The figure illustrates a detailed view in the MDCT block of Figure 6b;

Figure 9a illustrates an implementation of an optimizer operating at reduced resolution with respect to the MDCT output;

Figure 9b illustrates a representation of a MDCT spectrum and the corresponding bands of lower resolution, in which the prediction information is calculated;

Figure 10a illustrates an implementation of the real-to-imaginary transformer in Figure 6a or Figure 6b;

Y

Figure 10b illustrates a possible implementation of the imaginary spectrum calculator of Figure 10a.

Figure 1 illustrates an audio decoder for decoding a multichannel audio signal encoded and obtained on an input line 100. The encoded multichannel audio signal comprises a first coded combination signal that is generated using a combination rule to combine a first channel signal and a second channel signal representing the multichannel audio signal, a residual signal of coded prediction and prediction information. The encoded multi-channel signal may be a data stream such as a bitstream, which has all three components in a multiplexed form. Additional side information may be included in the multi-channel signal encoded on line 100. The signal is input to an input interface 102. The input interface 102 may be implemented as a data stream demultiplexer, which outputs the first combination signal encoded on line 104, the residual signal encoded on line 106 and the prediction information on line 108. Preferably, the prediction information is a factor that has a real part that is not equal to zero and / or an imaginary part other than zero. The coded combination signal and the coded residual signal are input to a signal decoder 110 for decoding the first combination signal to obtain a first decoded combination signal on the line 112. Additionally, the signal decoder 110 is configured to decode the signal. residual signal encoded to obtain a decoded residual signal on line 114. Depending on the coding processing on one side of the audio encoder, the signal decoder may comprise an entropy decoder such as a Huffman decoder, an arithmetic decoder or any other decoder of entropy and a subsequent dequantization stage connected to perform an dequantization operation that matches an operation of the quantizer in an associated audio encoder. The signals on line 112 and 114 are input to a decoder calculator 115, which outputs the first channel signal on line 117 and a second channel signal on line 118, where these two signals are stereo signals or two. channels of a multichannel audio signal. When, for example, the multichannel audio signal comprises five channels, then the two signals are two channels of the multi-channel signal. In order to fully encode said multi-channel signal having five channels, two decoders can be applied as illustrated in Figure 1, wherein the first decoder processes the left channel and the right channel, the second decoder processes the channel of the left environment and the channel of the right environment, and a third mono decoder is used to carry out a mono coding of the center channel. However, other groupings or combinations of waveform encoders and parametric encoders may also be applied. An alternative way to generalize the prediction scheme to more than two channels is to treat three signals (or more) at the same time, that is, to predict a 3rd combination signal from a 1st and 2nd signal using two prediction coefficients, in very similar to the "two to three" module in the MPEG environment.

The decoder calculator 116 is configured to calculate a decoded multi-channel signal having the first decoded channel signal 117 and the second decoded channel signal 118 using the decoded residual signal 114, the prediction information 108 and the first decoded combination signal 112. In particular, the decoder calculator 116 is configured to operate in such a way that the first decoded channel signal and the second decoded channel signal are at least an approximation of a first channel signal and a second signal channel of the signal multichannel signals introduced in a corresponding encoder, which are combined by the combination rule when the first combination signal is generated and the residual signal of prediction. In particular, the prediction information in line 108 comprises a part of real value other than zero and / or an imaginary part other than zero.

The decoder calculator 116 can be implemented in various ways. A first implementation is illustrated in Figure 4a. This implementation comprises a predictor 1160, a combination signal calculator 1161 and a combiner 1162. The predictor receives the first decoded combination signal 112 and the prediction information 108 and outputs a prediction signal 1163. Specifically, the predictor 1160 is configured to apply the prediction information 108 to the first decoded combination signal 112 or a signal from the first decoded combination signal. The derivation rule for deriving the signal to which the prediction information 108 is applied may be a transform from real to imaginary, or equally, an imaginary to real transform or a weighting operation or, depending on the implementation , a lag operation or a combined weighting / lag operation. The prediction signal 1163 is input together with the decoded residual signal in the combination signal calculator 1161 in order to calculate the second decoded combination signal 1165. Both signals 112 and 1165 are input to the combiner 1162, which combines the first decoded combination signal and the second combination signal for obtaining the decoded multi-channel audio signal having the first decoded channel signal and the second decoded channel signal on the output lines 1166 and 1167, respectively.

As an alternative, the decoder calculator is implemented as a matrix calculator 1168, which receives, as input, the first decoded combination signal or signal M, the decoded residual signal or signal D and the prediction information a 108. The matrix calculator 1168 applies a transform matrix that is illustrated as 1169 to the signals M, D to obtain the output signals L, R, where L is the first decoded channel signal and R is the second decoded channel signal. . The notation in Figure 4b resembles a stereo notation with a left channel L and a right channel R. This notation has been applied in order to provide a simpler explanation although it is clear to those skilled in the art that the signals L, R can be any combination of two channel signals in a multichannel signal having more than two channel signals. The matrix operation 1169 unifies the operations in blocks 1160, 1161 and 1162 of FIG. 4a in a "single trip" matrix type calculation, and the inputs in the circuit of FIG. 4a and the outputs in the circuit of FIG. 4a are identical to the entries in the matrix calculator 1168 or the outputs from the matrix calculator 1168.

Figure 4c illustrates an example for a reverse combination rule applied by the combiner 1162 in Figure 4a. In particular, the combination rule is similar to the combination rule on the decoder side in the well-known central / lateral encoding, where L = MS, and R = M - S. It is understood that the S signal used by the rule The combination of reverse combination in Figure 4c is the signal calculated by the combination signal calculator, ie, the combination of the prediction signal on line 1163 and the decoded residual signal on line 114. It is understood that as used herein. descriptive, the signs on the lines are sometimes designated by the reference numbers for the lines or are sometimes indicated by the reference numbers as such, which have been assigned to the lines. Therefore, the notation is such that a line that has a certain signal is indicating the signal itself. A line can be a physical line in a physically connected implementation. In a computerized implementation, however, a physical line does not exist, but the signal represented by the line is transmitted from one calculation module to the other calculation module.

Figure 2 illustrates an audio encoder for encoding a multichannel audio signal 200 having two or more channel signals, wherein a first channel signal is illustrated at 201 and a second channel is illustrated at 202. Both signals are input to an encoder calculator 203 for calculating a first combination signal 204 and a residual prediction signal 205 using the first channel signal 201 and the second channel signal 202 and the prediction information 206, such that when the signal is combined residual of prediction 205 with a prediction signal from the first combination signal 204 and the prediction information 206 a second combination signal is obtained, wherein the first combination signal and the second combination signal can be obtained from the first signal of channel 201 and the second signal of channel 202 using a combination rule.

The prediction information is generated by an optimizer 207 for calculating the prediction information 206 such that the residual signal of prediction meets an optimization objective 208. The first combination signal 204 and the residual signal 205 are input to an encoder. signals 209 for encoding the first combination signal 204 to obtain a first coded combination signal 210 and to encode the residual signal 205 to obtain a coded residual signal 211. Both coded signals 210, 211 are input to an output interface 212 to combine the first coded combination signal 210 with the residual encoded prediction signal 211 and the prediction information 206 to obtain a coded multichannel signal 213, which is similar to the encoded multi-channel signal 100 input into the audio decoder input interface 102 which is illustrated in figure 1.

Depending on the implementation, the optimizer 207 receives either the first channel signal 201 and the second channel signal 202 or, as illustrated on lines 214 and 215, the first combination signal 214 and the second signal. of combination 215 coming from a combiner 2031 of figure 3a, which will be analyzed in the following.

A preferred optimization target is illustrated in Figure 2, in which the coding gain is maximized, i.e., the bit rate is reduced as much as possible. In this optimization objective, the residual signal D is reduced to a minimum with respect to a. This means, in other words, that the prediction information a is selected in such a way that || S - aM || 2 is reduced to the minimum. This results in a solution for which Figure 2 is illustrated. The signals S, M are provided in block form and are preferably domain spectral signals, wherein the notation || ... || means rule 2 of the argument, and where <...> illustrates the product of points in the usual way. When the first channel signal 201 and the second channel signal 202 are input to the optimizer 207, the optimizer will then have to apply the combination rule, where an exemplary combination rule is illustrated in FIG. 3c. However, when the first combination signal 214 and the second combination signal 215 are entered into the optimizer 207, then the optimizer 207 does not need to implement the combination rule by itself.

Other optimization objectives may be related to perceptual quality. An objective of optimization can contemplate that a maximum perceptual quality is obtained. Then, the optimizer will need additional information from a perceptual model. Other implementations of the optimization objective can refer to the obtaining of a minimum bit rate, or a fixed bit rate. Then, the optimizer 207 will be put into practice to perform an input quantization / coding operation in order to determine the bit rate necessary for certain values of a such that a can be set to meet the requirements such as a minimum bit rate, or alternatively, a fixed bit rate. Other implementations of the optimization objective may be related to a minimum use of the encoder or decoder resources. In the case of an implementation of said optimization objective, the information on the necessary resources is available for a certain optimization in the optimizer 207. Additionally, a combination of these optimization objectives or other optimization objectives can be applied to control the optimizer 207, which calculates the prediction information 206.

The encoder calculator 203 in Figure 2 can be implemented in different ways, wherein a first exemplary implementation is illustrated in Figure 3a, in which an explicit combination rule is carried out in the combiner 2031. An exemplary alternative implementation is illustrated in Figure 3b, where a matrix calculator 2039 is used. The combiner 2031 in Figure 3a can be implemented to carry out the combination rule illustrated in Figure 3c , which is exemplary the well-known central / lateral coding rule, where a weighting factor of 0.5 applies to all branches. However, other weighting factors or no weighting factor can be put into practice depending on the implementation. Additionally, it should be noted that other combination rules may be applied, such as other linear combination rules or combination rules that are not linear, provided that there is a corresponding inverse combination rule, which may be applied in the decoder's combiner 1162. it is illustrated in Figure 4a, which applies a combination rule that is inverse to the combination rule applied by the encoder. Any invertible prediction rule can be used, because the influence on the waveform is "balanced" by the prediction, that is, any error is included in the residual signal transmitted, because the prediction operation carried out by the optimizer 207 in combination with the encoder calculator 203 is a process that preserves the waveform.

The combiner 2031 outputs the first combination signal 204 and a second combination signal 2032. The first combination signal is entered into a predictor 2033, and the second combination signal 2032 is entered into the residual calculator 2034. The predictor 2033 calculates a prediction signal 2035, which is combined with the second combination signal 2032 to finally obtain the residual signal 205. Specifically, the combiner 2031 is configured to combine the two channel signals 201 and 202 of the multi-channel audio signal in two. different ways to obtain the first combination signal 204 and the second combination signal 2032, wherein the two different forms are illustrated in an example in Figure 3c. The predictor 2033 is configured to apply the prediction information to the first combination signal 204 or a signal from the first combination signal to obtain the 2035 prediction signal. The signal from the combination signal may come from any operation not linear or linear, where a transform from real to imaginary / transformed from imaginary to real is preferred, which can be implemented using a linear filter such as an FIR filter that carries out weighted additions of certain values. The residual calculator 2034 in FIG. 3a can perform a subtraction operation in such a way that the prediction signal is subtracted from the second combination signal. However, other operations on the residual calculator are possible. In the same way, the combination signal calculator 1161 in FIG. 4a can perform an addition operation wherein the decoded residual signal 114 and the prediction signal 1163 are aggregated together to obtain the second combination signal 1165.

Figure 5a illustrates an audio encoder. In comparison with the audio encoder illustrated in Figure 3a, the first channel signal 201 is a spectral representation of a first channel signal in the time domain 55a. In the same way, the second channel signal 202 is a spectral representation of a channel signal in the time domain 55b. The conversion from the time domain into the spectral representation is carried out by a time / frequency converter 50 for the first channel signal and a time / frequency converter 51 for the second channel signal. Preferably, but not necessarily, the spectral converters 50, 51 are implemented as real-value converters. The conversion algorithm can be a discrete cosine transform, an FFT transform, where only the real part, an MDCT or any other transform that provides real value spectral values is used. As an alternative, both transforms can be implemented as an imaginary transform, such as a DST, an MDST or an FFT where only the imaginary part is used and the real part is discarded. Any other transform that provides only imaginary values can also be used. One reason why using a pure transform of real value or an imaginary pure transform is computational complexity, because, for each of the spectral values, only a single value such as magnitude or the real part has to be processed or , as an alternative, the phase or the imaginary part. Unlike a totally complex transform such as an FFT, two values have to be processed, that is, the real part and the imaginary part for each of the spectral lines, which is an increase in computational complexity by a factor of minus 2. At this point, another reason to use a real value transform is that such a transform is generally sampled, and thus provides a suitable (and commonly used) domain for signal quantization and entropy coding (the standard paradigm of "perceptual audio coding" that is put into practice in "MP3", AAC, or similar audio coding systems).

Figure 5a further illustrates the residual calculator 2034 as an adder, which receives the side signal as its "plus" input and which receives the output of the prediction signal by the predictor 2033 as its "minus" input. Additionally, Figure 5a illustrates the situation in which the predictor control information is transmitted from the optimizer to the multiplexer 212, which outputs a multiplexed data stream representing the encoded multi-channel audio signal. In particular, the prediction operation is carried out in such a way that the lateral signal is predicted from the central signal as illustrated in the equations to the right of figure 5a. Preferably, the control information of the predictor 206 is a factor as illustrated to the right in Figure 3b. In an example in which the prediction control information only comprises a real portion such as the real part of a complex value a or a magnitude of the complex value a, where this portion corresponds to a non-zero factor, it is it can obtain a significant coding gain when the central signal and the lateral signal are similar to each other due to their waveform structure, although they have different amplitudes. However, when the prediction control information only comprises a second portion, which may be the imaginary part of a complex value factor or the phase information of the complex value factor, where the imaginary part or the phase information is non-zero, the present invention achieves a significant coding gain for the signals, which are out of phase with each other by a value other than 0 ° or 180 °, and which have in addition to the phase shift, similar waveform characteristics and relationships of similar amplitude.

Preferably, a prediction control information is of complex value. Therefore, a significant coding gain can be obtained for the signals that have different amplitude and that are out of phase. In a situation in which the time / frequency transforms provide complex spectra, operation 2034 will be a complex operation in which the real part of the predictor control information is applied to the real part of the complex spectrum M and the imaginary part of the complex prediction information is applied to the imaginary part of the complex spectrum. Then, in adder 2034, the result of this prediction operation is a predicted real spectrum and an imaginary predicted spectrum, and the predicted real spectrum is subtracted from the real spectrum of the side signal S (in the form of a band), and the spectrum The predicted imaginary is subtracted from the imaginary part of the spectrum of S to obtain a complex residual spectrum D.

The signals in the time domain L and R are signals of real value, although the signals in the frequency domain can be of real or complex value. When the signals in the frequency domain are of real value, then the transform is a real value transform. When the signals in the frequency domain are complex, then the transform is a complex value transform. This means that the input to the transforms from time to frequency and the output of the transforms from time to time are of real value, while the signals in the frequency domain could be, for example, signals in the QMF domain. of complex value.

Figure 5b illustrates an audio decoder corresponding to the audio encoder that is illustrated in Figure 5a. Similar elements with respect to the audio decoder of Figure 1 have similar reference numbers.

The output of the bitstream by the multiplexer of the bitstream 212 in FIG. 5a is input to a demultiplexer of the bitstream 102 in FIG. 5b. The bitstream demultiplexer 102 demultiplexes the bit stream in the downmix signal M and the residual signal D. The downmix signal M is input to a dequantizer 110a. The residual signal D is input to a dequantizer 110b. Additionally, the bit stream demultiplexer 102 demultiplexes a control information from the predictor 108 from the bit stream and inputs it into the predictor 1160. The predictor 1160 outputs a predicted lateral signal to ■ M and the combiner 1161 combines the signal residual signal emitted by the dequantizer 110b with the predicted lateral signal in order to finally obtain the reconstructed lateral signal S. The signal is then input to the combiner 1162, which performs, for example, a sum / difference processing, such as illustrated in Figure 4c with respect to central / lateral coding. In particular, block 1162 performs a central / lateral (reverse) decoding to obtain a frequency domain representation of the left channel and a frequency domain representation of the right channel. The representation of the frequency domain is then converted into a time domain representation by the corresponding frequency / time converters 52 and 53.

Depending on the implementation of the system, the frequency / time converters 52, 53 are real frequency / time value converters when the frequency domain representation is a real value representation, or complex value frequency / time converters when the representation of the frequency domain is a complex value representation.

However, to increase the efficiency it is preferred to carry out a real value transformation as illustrated in another implementation in Figure 6a for the encoder according to one embodiment of the present invention and in Figure 6b for the decoder according to an embodiment of the present invention. The real value transforms 50 and 51 are implemented by means of an MDCT. Additionally, the prediction information is calculated as a complex value that has a real part and an imaginary part. Since both spectra M, S are real-value spectra and because, therefore, no imaginary part of the spectrum exists, a real-to-imaginary converter 2070 is provided, which calculates an estimated imaginary spectrum 600 from from the real value spectrum of the M signal. This transformer from real to imaginary 2070 is a part of the optimizer 207, and the imaginary spectrum 600 estimated by the block 2070 is introduced in the optimizer stage of a 2071 together with the real spectrum M in order to calculate the prediction information 206, which now has a real value factor indicated in 2073 and an imaginary factor indicated in 2074. Then, according to the present embodiment, the real value spectrum of the first combination signal M is multiplied by the real part aR 2073 to obtain the prediction signal, which is then subtracted from the real-value side spectrum. Additionally, the imaginary spectrum 600 is multiplied by the imaginary part ai that is illustrated in 2074 to obtain the additional prediction signal, where this prediction signal is then subtracted from the real-value side spectrum as indicated in 2034b. Then, the residual prediction signal D is quantized in quantizer 209b, while the real value spectrum of M is quantized / encoded in block 209a. Additionally, it is preferred to quantize and encode the prediction information a in the quantifier / entropy encoder 2072 to obtain the value of a coded complex, which is transmitted to the bitstream multiplexer 212 of Figure 5a, for example, and which is finally entered into a stream of bits as the prediction information.

With respect to the position of the quantization / coding module (Q / C) 2072 for a, it is emphasized that the multipliers 2073 and 2074 preferably use exactly the same (quantized) that would also be used in the decoder. In this way, 2072 could be moved directly towards the 2071 output, or it could be considered that the quantification of a has already been taken into account in the optimization process in 2071.

Although a complex spectrum could be calculated on the encoder side because all the information is available, it is preferred to perform the transform from real to complex in block 2070 in the encoder in such a way that similar conditions occur with respect to to a decoder that is illustrated in Figure 6b. The decoder receives a coded real-value spectrum of the first combination signal and a real-valued spectral representation of the coded residual signal. Additionally, a complex prediction information encoded at 108 is obtained and an entropy decoding and dequantization is performed in block 65 to obtain the real part aR illustrated in 1160b and the imaginary part ai illustrated in 1160c. The central signals emitted by the weighting elements 1160b and 1160c are added to the residual decoded and unquantized prediction signal. Specifically, the spectral values entered in the weigher 1160c, where the imaginary part of the complex prediction factor is used as the weighting factor, come from the real value spectrum M by the real-to-imaginary converter 1160a, which is set in practice preferably in the same way as block 2070 of figure 6a related to the encoder side. On the decoder side there is no complex value representation of the central signal or the side signal, which implies a difference with respect to the encoder side. The reason is that only the encoded real value spectra have been transmitted from the encoder to the decoder due to the bit rates and for complexity reasons.

The real-to-imaginary transformer 1160a or the corresponding block 2070 of Fig. 6a can be implemented as published in WO 2004/013839 A1 or in WO 2008/014853 A1 or in the US Pat. No. 6,980,933. Alternatively, any other implementation known in the art can be applied, and a preferred implementation is described in the context of Figures 10a, 10b.

In particular, as illustrated in Figure 10a, the imaginary real converter 1160a comprises a spectral frame selector 1000 connected to an imaginary spectrum calculator 1001. The spectral frame selector 1000 receives an indication of a current frame i in the input 1002 and, depending on the implementation, control information in a control input 1003. When, for example, the indication on line 1002 indicates that an imaginary spectrum has to be calculated for a current frame i, and when the control information 1003 indicates that only the current frame has to be used for said calculation, then the spectral frame selector 1000 only selects the current frame i and transmits this information to the imaginary spectrum calculator. Therefore, the imaginary spectrum calculator only uses the spectral lines of the current frame i to perform a weighted combination of the lyses located in the current frame (block 1008), with respect to frequency, next to, or around the current spectral line k, for which an imaginary line has to be calculated as illustrated at 1004 in Figure 10b. However, when the spectral frame selector 1000 receives a control information 1003 which indicates that the previous frame i-1 and the following frame i 1 have to be used also for the calculation of the imaginary spectrum, then the imaginary spectrum calculator additionally receives the values from tables i - 1 and i 1 and performs a weighted combination of the lines in the corresponding tables as illustrated in 1005 for table i - 1 and 1006 for table i 1. The results of the weighting operations are combined by a weighted combination in block 1007 to finally obtain an imaginary line k for the frame fi which is then multiplied by the imaginary part of the prediction information in element 1160c to obtain the prediction signal for this line, which is then added to the corresponding line of the central signal in the adder 1161b for the decoder. The same operation is carried out in the encoder, although a subtraction is carried out in the element 2034b.

It should be noted that the control information 1003 may additionally indicate the use of more frames than the two surrounding frames or, for example, the use of the current frame and exactly one or more previous frames but without using "future" frames in order to reduce the systematic delay.

Additionally, it should be noted that the combination weighted in stages illustrated in figure 10b, in which, in a first operation, the lines coming from a table are combined and subsequently the results of said combination operations are combined by themselves. pictures, can also be carried out in the other order. The other order refers to that, in a first step, the lines for the current frequency k from a number of adjacent frames indicated by the control information 103 are combined by a weighted combination. This weighted combination is carried out for lines k, k - 1, k - 2, k 1, k 2, etc., depending on the number of adjacent lines to be used to estimate the imaginary line. Then, the results of said combinations "in time" are subjected to a weighted combination in the "frequency direction" to finally obtain the imaginary line k for the fi table. The weighted values are preferably set to -1 and 1, and the weights can be implemented in a direct combination of FIR or IIR filter, which performs a linear combination of spectral lines or spectral signals of different frequencies. and different pictures.

As indicated in figures 6a and 6b, the preferred transform algorithm is the MDCT transform algorithm, which is applied in the previous direction in elements 50 and 51 in figure 6a and which is applied in the subsequent direction in the elements 52, 53, following a combination operation in the combiner 1162 operating in the spectral domain.

Figure 8a illustrates a more detailed implementation of block 50 or 51. Specifically, a sequence of audio samples of the time domain is entered into an analysis window system 500, which performs a window operation that it uses a window of analysis and, in particular, it carries out this operation in a frame by means of a frame mode but using a step or overlap of 50%. The result of the analysis window system, ie, a sequence of frames of samples divided into windows, is entered into an MDCT transform block 501, which outputs the sequence of MDCT frames of real value, where said frames are affected. by overlapping. As an example, the analysis window system applies analysis windows that have a length of 2048 samples. Then, the MDCT transform block 501 emits MDCT spectra having 1024 real spectral lines or MDCt values. Preferably, the analysis window system 500 and / or the MDCT transformer 501 can be controlled by a window length or transform length 502 control in such a way that, for example, for the transient portions in the signal, the length window / transform length is reduced in order to obtain better coding results.

Figure 8b illustrates the inverse MDCT operation carried out in blocks 52 and 53. By way of example, block 52 comprises a block 520 for performing a reverse MDCT transform from frame to frame. When, for example, a table of MDCT values has 1024 values, then the output of this MDCT inverse transform has 2048 time samples affected by the overlap. This table is supplied to a system of windows of synthesis 521, which applies a sale of synthesis to this table of 2048 samples. The box divided into windows is then transmitted to an overlap / add processor 522, which, by way of example, applies a 50% overlap between two subsequent frames and, then, performs a sample addition per sample of such So that a block of 2048 samples finally achieves 1024 new samples of the overlap-free output signal. Again it is preferred to apply a window / transform length control using information, which for example is transmitted in the lateral information of the encoded multi-channel signal according to that indicated in 523.

The prediction values of a could be calculated for each of the individual spectral lines of a MDCt spectrum. However, it has been found that this is not necessary and that a significant amount of lateral information can be saved by performing a band calculation of the prediction information. In other words, a spectral converter 50 illustrated in FIG. 9, which is for example an MDCT processor described in the context of FIG. 8a, provides a high frequency resolution spectrum having certain spectral lines that are illustrated. in Figure 9b. This high frequency resolution spectrum is used by a spectral line selector 90 which provides a low frequency resolution spectrum, which comprises certain bands B1, B2, B3, ..., BN. This low frequency resolution spectrum is transmitted to the optimizer 207 to calculate the prediction information such that no prediction information is calculated for each of the spectral lines, but only for each of the bands. For this purpose, the optimizer 207 receives the spectral lines per band and calculates the optimization operation based on the presumption that the same value a is used for all the spectral lines in the band.

Preferably, the bands are configured in a psychoacoustic manner such that the bandwidth of the bands increases from lower frequencies to higher frequencies as illustrated in Figure 9b. As an alternative, although it is not preferred as much as the implementation of increasing bandwidth, frequency bands of equal size could also be used, where each of the frequency bands has at least two or, at general, many more, such as at least 30 lines of frequency. In general, for a spectrum of 1024 spectral lines, less than 30 values of a complexes are calculated, and preferably, more than 5 values of a. For the spectra with less than 1024 spectral lines (e.g., 128 lines), preferably less frequency bands (e.g., 6) are used for a.

To calculate the values of a, the high resolution MDCT spectrum is not necessarily required. Alternatively, you can also use a filter bank that has a frequency resolution similar to the resolution necessary to calculate the values of a. When bands that increase in frequency have to be implemented, in that case said bank of filters must have a variable bandwidth. However, when a constant bandwidth of low to high frequencies is sufficient, then a traditional filter bank with sub-bands of equivalent width can be used.

Depending on the implementation, the sign of the value of a indicated in Figure 3b or 4b can be inverted. However, in order to be coherent, it is necessary that said reversion of the sign be used on the encoder side as well as on the decoder side. Unlike Figure 6a, Figure 5a illustrates a generalized view of the encoder, wherein element 2033 is a predictor that is controlled by the predictor control information 206, which is determined in element 207 and which is incorporated as lateral information in the bit stream. In place of the MDCT used in Figure 6a in blocks 50, 51 a generalized time / frequency transform is used in Figure 5a as described. As indicated above, Figure 6a is the process of the encoder, which corresponds to the decoder process in Figure 6b, where L represents the left signal of the channel, R represents the right signal of the channel, M represents the central signal or the descending mix signal, S represents the lateral signal and D represents the residual signal. Alternatively, L is also known as the first channel signal 201, R is also known as the second channel signal 202, M is also known as the first combination signal 204, and S is also known as the second 2032 combination signal. .

Preferably, the modules 2070 in the encoder and 1160a in the decoder must coincide exactly in order to ensure correct waveform coding. This applies preferably to the case in which said modules use some form of approximation such as truncated filters or when only one or two are used in place of the three MDCT frames, ie the current MDCT frame in line 60, the previous MDCT chart on line 61 and the next MDCT chart on line 62.

Additionally, it is preferred that the module 2070 in the encoder in Figure 6a use the non-quantized MDCT spectrum M as input, although the real-to-imaginary module (R21) 1160a in the decoder has only the quantized MDCT spectrum available as input. Alternatively, an implementation may also be used in which the encoder uses the MDCT coefficients quantified as input in the 2070 module. However, the use of the non-quantized MDCT spectrum as input in the 2070 module is the preferred method from a perceptual point of view.

In the following, various aspects of embodiments of the present invention are described in more detail.

Standard parametric stereo coding is based on the ability of the over-sampled complex QMF (Imbrid) domain to allow perceptually motivated signal processing with variation of time and frequency without introducing overlapping distortions. However, in the case of the down / residual mix encoding (as used for the high bitrates considered in the present context), the resulting unified stereo encoder acts as a waveform encoder. This allows operation in a scientifically sampled domain, such as the MDCT domain, because the waveform coding paradigm ensures that the overlap cancellation property of the MDCT-IMDCT processing chain is kept well enough. .

However, in order to take advantage of the improved coding efficiency that can be achieved in the case of stereo signals with time or phase differences between channels by means of a complex value prediction coefficient a, a value representation is needed. complex in the frequency domain of the DMX downmix signal as input into the complex value upmix array. This can be obtained by using an MDST transform in addition to the MDCT transform for the DMX signal. The MDST spectrum can be computed (in exact form or as an approximation) from the MDCT spectrum.

Also, the parametrization of the upmix matrix can be simplified by transmitting the complex prediction coefficient a instead of the MPS parameters. In this way only two parameters are transmitted (real and imaginary part of a) instead of three (ICC, CLD and IPD). This is possible due to the redundancy in the parameterization of MPS in the case of the down / residual mix coding. The MPS parameterization includes information about the relative amount of decorrelation to be added in the decoder (ie, the energy ratio between the RES and DMX signals), and this information is redundant when the real signals DMX and RES are transmitted.

Due to the same reason, the gain factor g, which is illustrated in the previous upmix matrix, is obsolete in the case of the down / residual mix coding. Thus, the upmix matrix for the down / residual mix coding with complex prediction is as follows:

In comparison with Equation 1169 in Figure 4b, the sign of alpha is inverted in this equation, and DMX = M and RES = D. This is, therefore, an implementation / alternative notation with respect to Figure 4b .

There are two options to calculate the residual signal of prediction in the encoder. One option is to use the quantized MDCT spectral values of the downmix. This will result in the same quantization error distribution as in the M / S coding because the encoder and the decoder use the same values to generate the prediction. The other option is to use the non-quantized MDCT spectral values. This implies that the encoder and the decoder will not use the same data to generate the prediction, which allows the spatial redistribution of the coding error according to the properties of instantaneous masking of the signal at the expense of a somewhat reduced coding gain.

It is preferable to compute the MDST spectrum directly in the frequency domain by means of two-dimensional FIR filtering of three adjacent MDCT frames as analyzed. The latter can be considered as a transform "from real to imaginary" (R2I). The complexity of the frequency-domain computation of the MDST can be reduced in various ways, which means that only one approximation of the MDST spectrum is calculated:

• Limiting the number of FIR filter connections.

• Estimating the MDST only from the actual MDCT chart.

• Estimating the MDST from the previous and current MDCT chart.

Whenever the same approximation is used in the encoder and in the decoder, the waveform coding properties will not be affected. Such approximations of the MDST spectrum, however, can cause a reduction in the coding gain achieved by the complex prediction.

In case the underlying MDCT encoder supports the window-like switching, the coefficients of the FIR two-dimensional filter used to compute the MDST spectrum have to be adapted to the actual window forms. The filter coefficients applied to the MDCT spectrum of the current table depend on the complete window, that is, a set of coefficients is required for each type of window and for each window transition. The filter coefficients applied to the MDCT spectrum of the previous / following table depend only on the overlap of the window half with the current frame, that is, for these a set of coefficients is necessary only for each window type (without additional coefficients for the transitions).

In case the underlying MDCT encoder uses a transform length transition, including the MDCT frame previous and / or next in the approach, this becomes more complicated around the transitions between the different transform lengths. In this case, due to the different number of MDCT coefficients in the current, previous / next frame, two-dimensional filtering is more complicated. In order to avoid the increasing computational and structural complexity, the previous / next frame can be excluded from the filtering in the transform length transitions, at the expense of a lower accuracy of the approximation for the respective frames.

Also, special care must be taken with the lower and higher parts of the MDST spectrum (proximal to CC and fs / 2), where there are fewer surrounding MDCT coefficients for FIR filtering than necessary. In the present case, the filtering process needs to be adapted to compute the MDST spectrum correctly. This can be done either by using a symmetric extension of the MDCT spectrum for the missing coefficients (according to the periodicity of the discrete time signal spectra), or by adapting the filter coefficients accordingly. It is obvious that the treatment of such special cases can be simplified at the expense of a lower accuracy in the vicinity of the limits of the MDSt spectrum.

The computation of the exact MDST spectrum from the MDCT spectra transmitted in the decoder increases the decoder delay by one frame (there are presumed to be 1024 samples).

The additional delay can be avoided by using an approximation of the MDST spectrum that does not require the MDCT spectrum of the following table as input.

The following list of elements summarizes the advantages of unified MDCT-based stereo coding compared to unified stereo coding based on QMF:

• Only a small increase in computational complexity (when SBR is not used).

• Extend the capacity to a perfect reconstruction if the MDCT spectra are not quantified. Note that this does not happen for unified stereo encoding based on QMF.

• Natural extension of the M / S coding and intensity of the stereo coding.

• Cleaning architecture that simplifies the adjustment of the encoder, because the processing of the stereo signal and the quantification / coding can be strongly coupled. It should be noted that in unified stereo coding based on QMF, the frames of the MPEG environment and the MDC ^t frames are not aligned and said scale factor bands do not coincide with the bands of the parameter.

• Optimizes the coding of stereo parameters, because only two parameters have to be transmitted (complex a) instead of three parameters as in the MPEG environment (ICC, CLD, IPD).

• It does not present an additional delay in the decoder if the MDST spectrum is computed as an approximation (without using the following table).

The important properties of an implementation can be summarized as follows:

a) The MDST spectra are computed by means of two-dimensional FIR filtering from the current, previous and following MDCT spectra. Different complexity / quality compensations are enabled for the MDST computation (approximation) by reducing the number of FIR filter connections and / or the number of MDCT frames used. In particular, if an adjacent frame is not available due to frame loss during transmission or transform length switching, that particular frame is excluded from the estimated value of MDST. For the case of the transformation length switching, the exclusion is signaled in the bitstream.

b) Only two parameters, the real and imaginary part of the complex prediction coefficient a, are transmitted instead of ICC, CLD, and IPD. The real and imaginary parts of a are treated independently, limited to the range [-3.0, 3.0] and quantified with a step size of 0.1. If a certain parameter (real or imaginary part of a) is not used in a certain frame, this is signaled in the bit stream, and the irrelevant parameter is not transmitted. Parameters are coded in time differential or differential frequency form and finally Huffman coding is applied using the scale factor code. The prediction coefficients are updated every two scale factor bands, which produces a frequency resolution similar to the frequency resolution of the MPEG Environment. This quantization and coding scheme results in an average bit rate of approximately 2 kb / s for the lateral stereo information within a typical configuration having a bit rate objective of 96 kb / s.

Preferred, additional or alternative details of the implementation include:

c) For each of the two parameters of a, a non-differential (PCM) or differential (DPCM) coding can be chosen per frame or current, indicated by a corresponding bit in the bit stream. For a DPCM coding it is possible, either a differential coding of time or frequency. Again, this can be signaled using a one-bit flag.

d) Instead of reusing a predefined code such as the AAC scale factor code, a specialized non-variable or adaptive signal code may be used to encode the values of the parameter a, or it may be returned to the fixed-length codes (eg example, 4 bits) without sign or two add-ons.

e) The range of the values of the parameter as well as the size of the parameter quantization step can be chosen and optimized in an arbitrary way up to the characteristics of the signals in question.

f) The number and spectral and / or temporal width of the bands of active parameter a can be chosen and optimized in an arbitrary way up to the characteristics of certain signals. In particular, the band configuration can be indicated by frame or by current.

g) In addition to, or in lieu of the mechanisms indicated in a) above, it may be indicated explicitly by means of one bit per frame in the bitstream that only the MDCT spectrum of the current frame is used for compute the approximation of the MDST spectrum, ie, that the adjacent MDCT frames are not taken into account.

The embodiments relate to a system of the invention for unified stereo coding in the MDCT domain. This allows us to use the advantages of unified stereo coding in the USAC system of MPEG even at higher bit rates (where SBR is not used) without the significant increase in computational complexity that occurs with a QMF-based method.

The following two lists summarize the preferred aspects of the configuration that have been described above, which can be used alternatively among themselves or in addition to other aspects:

the) general concept: complex prediction of lateral MDCT from MDCT and central MDST;

lb) calculate / estimate the MDST from MDCT ("R2I") in the frequency domain using 1 or more frames (3 frames cause delay);

lc) truncated filter (even going down to 1 frame 2 connections, that is, [-1 0 1]) to reduce computational complexity;

ld) adequate treatment of CC and fs / 2;

le) suitable treatment of window form switching;

lf) do not use a previous / next frame if it has a different transform size;

lg) prediction based on MDCT coefficients not quantified or quantified in the encoder;

2a) quantifying and encoding the real and imaginary part of the complex prediction coefficient directly (ie, without parametrizing the MPEG environment);

2b) use uniform quantizer for this (step size, for example, 0.1);

2c) use an appropriate frequency resolution for the prediction coefficients (for example, 1 coefficient for each 2 Scales Factor Bands);

2d) economic signaling in the event that all prediction coefficients are real;

2e) bit explicito per frame to force the operation of R2I of 1 frame.

In one embodiment, the encoder further comprises: a spectral converter (50, 51) for converting a time domain representation of the two channel signals into a spectral representation of the two channel signals having sub-band signals for the two channel signals, where the combiner (2031), the predictor (2033) and the residual signal calculator (2034) are configured to process each of the subbands separately so that the first combined signal and the residual signal is obtained for a plurality of subbands, wherein the output interface (212) is configured to combine the first coded combined signal and the coded residual signal for the plurality of subbands.

Although some aspects have been described in the context of an apparatus, it is clear that said aspects only represent a description of the corresponding method, wherein a block or device corresponds to a step of the method or to a characteristic of a step of the method. In an analogous manner, the aspects described in the context of a step of the method also represent a description of a corresponding block or element or feature of a corresponding apparatus.

In one embodiment of the present invention, a suitable window-shaped switching treatment is applied. In figure 10a it is illustrated that a window shape information 109 can be entered in the imaginary spectrum calculator 1001. Specifically, the imaginary spectrum calculator, which performs the real-to-imaginary conversion of the real-value spectrum such as the MDCT spectrum (as, for example, element 2070 in Figure 6a or element 1160a in Figure 6b) can be implemented as an FIR or IIR filter. The FIR or IIR coefficients in this imaginary real 1001 module depend on the window shape of the left half and the right half of the current frame. This window form can be different for a sine window or a Derivative window of Kaiser Bessel (Kaiser Bessel Derived, KBD) and depending on the configuration of a certain window sequence it can be a long window, a start window, a stop window and a stop-start window or a short window. The real-imaginary module can comprise a two-dimensional filter FlR, where one dimension is the time dimension where two subsequent MDCT frames are introduced in the FIR filter and the second dimension is the frequency dimension, where the frequency coefficients of a table.

The following table provides different MDST filter coefficients for a current window sequence for different window forms and different implementations of the left half and the right half of the window.

Table A - MDST Filter Parameters for the Current Window

Additionally, the window shape information 109 provides information of the window shape for the previous window, when the previous window is used to calculate the MDST spectrum from the MDCT spectrum. The corresponding MDST filter coefficients for the previous window are provided in the following table.

Table B - MDST Filter Parameters for the Previous Window

Thus, depending on the information of the window form 109, the imaginary spectrum calculator 1001 in Figure 10a is adapted by applying different sets of filter coefficients.

The information of the window form, which is used on the decoder side, is calculated on the encoder side and transmitted as lateral information together with the output signal of the encoder. On the decoder side, the information of the window form 109 is extracted from the bitstream by the bitstream demultiplexer (for example 102 in FIG. 5b) and is provided to the imaginary spectrum calculator 1001 such as it is illustrated in figure 10a.

When the information of the window form 109 indicates that the previous table has a different transform size, then it is preferred that the above table is not used to calculate the imaginary spectrum from the real value spectrum. The same happens when interpreting the information of the window form 109 it is discovered that the following table has a different transform size. Then, the following table is not used to calculate the imaginary spectrum from the real value spectrum. In such a case when, for example, the previous frame has a different transform size than the current frame and when the next frame again has a different transform size compared to the current frame, then only the current frame, ie the spectral values of the current window are used to estimate the imaginary spectrum.

The prediction in the encoder is based on unquantified or quantized frequency coefficients such as the MDCT coefficients. When, for example, the prediction illustrated by the element 2033 in Figure 3a is based on non-quantized data, then the residual calculator 2034 also preferably operates on non-quantized data and the residual output signal of the calculator is say, the residual signal 205 is quantified before being encoded by entropy and transmitted to a decoder. However, in an alternative embodiment it is preferred that the prediction be based on quantified MDCT coefficients. Then, quantization may occur before combiner 2031 in Figure 3a such that a first quantized channel and a second quantized channel are the basis for calculating the residual signal. Alternatively, the quantization may also occur after the combiner 2031 such that the first combination signal and the second combination signal are calculated in an unquantified form and quantified before the residual signal is calculated. Again, alternatively, the 2033 predictor can operate in the unquantized domain and the 2035 prediction signal is quantized before being entered into the residual calculator. Then, it is useful that the second combination signal 2032, which is also input to the residual calculator 2034, is also quantified before the residual calculator calculates the residual signal 1070 in Figure 6a, which can be implemented within of the 2033 predictor in Figure 3a, operate on the same quantized data that is available on the decoder side. It can then be ensured that the MDST spectrum estimated in the encoder for carrying out the calculation of the residual signal is exactly the same as the MDST spectrum on the decoder side used to carry out the inverse prediction, ie for calculate the lateral signal from the residual signal. For this purpose, the first combination signal such as the signal M in line 204 in figure 6a is quantized before being entered in block 2070. Then, the spectrum of MDST calculated using the quantized MDCT spectrum of the current frame, and depending on the control information, the quantized MDCT spectrum of the previous or following table is entered into the multiplier 2074, and the output of the multiplier 2074 of FIG. 6a again will be an unquantized spectrum. This unquantified spectrum will be subtracted from the spectrum introduced in the adder 2034b and finally quantized in the quantizer 209b.

In one embodiment, the real part and the imaginary part of the complex prediction coefficient per prediction band are directly quantized and coded, ie, for example, without the MPEG environment parameterization.

The quantification can be carried out using a uniform quantizer with a step size, for example, 0.1. This means that no quantum or similar quantum step size is applied, but any linear step size is applied. In an implementation, the range of values for the real part and the imaginary part of the complex prediction coefficient have a range from -3 to 3, which means that 60 or, depending on the details of implementation, are used, 61 quantification steps for the real part and the imaginary part of the complex prediction coefficient.

Preferably, the real part applied in the multiplier 2073 in Figure 6a and the imaginary part 2074 applied in Figure 6a are quantized before being applied such that, again, the same value is used for the prediction on the encoder as well as used on the decoder side. This ensures that the residual prediction signal encompasses - in addition to the quantization error introduced - any errors that might occur when an unquantized prediction coefficient is applied on the encoder side while applying a quantized prediction coefficient on the decoder side. Preferably, the quantization is applied in such a way that - as much as possible - the same situation and the same signals are available on the encoder side and on the decoder side. In this way it is preferred to quantify the input in the calculator from real to imaginary 2070 using the same quantization applied in quantizer 209a. Additionally, it is preferred to quantify the real part and the imaginary part of the prediction coefficient of a to carry out the multiplications in the element 2073 and in the element 2074. The quantization is the same as that applied in the quantizer 2072. Additionally, the signal The lateral one emitted by the block 2031 in FIG. 6a can also be quantized before the adders 2034a and 2034b. However, it is not problematic that the quantifier 209b carries out the quantification after the addition, where the addition by said adders is applied with a non-quantized side signal.

In an additional example, an economic signaling is applied in case all the prediction coefficients are real. It can happen that all the prediction coefficients for a given frame, that is, for the same time portion of the audio signal, are calculated as real. This situation can occur when the complete central signal and the complete lateral signal are not out of phase with each other or are a little out of phase. With the In order to save bits, this is indicated by a single real indicator. Then, it is not necessary to indicate the imaginary part of the prediction coefficient in the bitstream with a code representing a zero value. On the decoder side, the decoder interface of the bit stream, such as a bitstream demultiplexer, will interpret this real flag and then will not search for keywords for an imaginary part but will understand that all bits are in the section. corresponding to the bit stream as bits for real value prediction coefficients. Also, when the predictor 2033 receives an indication that all the imaginary parts of the prediction coefficients in the table are zero, it will not need to calculate a MDST spectrum, or generally an imaginary spectrum from the real value MDCT spectrum. Therefore, element 1160a in the decoder of figure 6b will be deactivated and reverse prediction will only take place using the real value prediction coefficient applied in multiplier 1160b in figure 6b. The same happens for the encoder side where the element 2070 will be deactivated and the prediction will only take place using the multiplier 2073. This lateral information is preferably used as an additional bit per frame and the decoder will read this bit, frame by frame, with in order to decide whether the converter from real to imaginary 1160a will be active for a frame or not. In this way, by providing this information a reduced size of the bit stream is achieved due to the most efficient signaling of all the imaginary parts of the prediction coefficient that are zero for a frame and, additionally, it provides less complexity for the decoder for said table immediately achieving a lower consumption of said processor battery that is put into practice, for example, in a mobile device powered by batena.

The complex stereo prediction according to the preferred embodiments of the present invention is a tool for the efficient coding of channel pairs with level and / or phase differences between the channels. When using a complex value parameter a, the left and right channels are reconstructed by means of the following matrix. dmxim indicates the MDST corresponding to the MDCT of the dmxRe downmix channels.

The above equation is another representation, which is divided with respect to the real part and the imaginary part of a and represents the equation for a combined prediction / combination operation, in which the predicted signal S is not necessarily calculated.

The following data elements are preferably used for this tool:

cplx_pred all 0: Some bands use the L / R encoding, indicated by cplx_pred_used []

1: All bands use complex stereo prediction

cplx_pred_used [g] [sfb] A one bit flag per window group g and scale factor band sfb (after mapping from the prediction bands) indicating that

0: the complex prediction is not being used, the L / R coding is used

1: complex prediction is used

complex_coef 0: aim = 0 for all prediction bands

1: aim is transmitted for all prediction bands

use_prev_frame 0: Use only the current frame to estimate the MDST

1: Use the current and previous chart to estimate the MDST

delta_code_time 0: Frequency differential coding of the prediction coefficients

1: Time differential coding of the prediction coefficients hcod_alpha_q_re Huffman code of aRe

hcod_alpha_q_im Huffman code of aim

These data elements are calculated in an encoder and entered into the lateral information of a multi-channel or stereo audio signal. The elements are extracted from the lateral information on the decoder side by means of a lateral information extractor and are used to control the decoder calculator to carry out a corresponding action.

The complex stereo prediction requires the down-mix MDCT spectrum of the current channel pair and, in the case of complex_coef == 1, an estimated value of the downmix MDST spectrum of the current channel pair, ie the imaginary counterpart of the MDCT spectrum. The estimated value of the downmix MDST is computed from the downmix MDCT of the current table and, in the case of use_prev_frame == 1, from the downmix MDCT of the previous table. The MDCT of downmixing of the previous table of the group of windows g and group of windows b is obtained from the reconstructed left and right spectra of said table.

In the computation of the estimated value of the downmixing MDST, the MDCT transform length of uniform value is used, which depends on window_sequence, as well as filter_coefs and filter_coefs_prev, which are matrices containing the filter cores and the which are obtained according to the previous tables. For all prediction coefficients, the difference of a previous value (in time or frequency) is encoded using a Huffman code. The prediction coefficients are not transmitted for the prediction bands for which cplx_pred_used = 0 was used.

The inverse prediction coefficients quantified alpha_re and alpha_im are provided by

alpha_re = alpha_q_re * 0.1

alpha_im = alpha_q_im * 0.1

It should be noted that the invention not only applies to stereo signals, ie multichannel signals having only two channels, but also applies to two channels of a multi-channel signal having three or more channels such as a 5.1 or 7.1 signal.

The encoded audio signal of the invention may be stored in a digital storage medium or transmitted in a transmission medium such as a wireless transmission medium or a cable transmission medium such as the Internet.

Depending on certain requirements for the implementation, the embodiments of the invention can be implemented in hardware or software. The implementation can be carried out using a digital storage medium, for example a floppy disk, a DVD, a CD, a ROM, a PROM, an EPROM, an EEPROM or a FLASH memory, which has legible control signals. electronically stored in it, which collaborate (or are able to collaborate) with a computer programmable system in such a way that the respective method is carried out.

Some embodiments according to the invention comprise a non-temporary or tangible data carrier having electronically readable control signals, which are capable of cooperating with a programmable computer system, such that one of the methods described herein is carried out.

In general, the embodiments of the present invention can be implemented as a computer program product with a program code which is operative to carry out one of the methods when the computer program product is run on a computer . The program code can be stored, for example, in a computer readable carrier.

Other embodiments comprise the computer program for carrying out one of the methods described herein, stored in a computer readable carrier.

In other terms, one embodiment of the method of the invention is, therefore, a computer program having a program code for carrying out one of the methods described herein, when the computer program is executed on a computer .

A further embodiment of the methods of the invention is, therefore, a data carrier (or a digital storage medium, or a computer readable medium) comprising, recorded therein, the computer program to carry one of the methods described herein.

A further embodiment of the method of the invention is, therefore, a data stream or a signal sequence representing the computer program to carry out one of the methods described herein. The data stream or the signal sequence may be configured, for example, to be transferred via a data communication connection, for example, via the Internet.

Another embodiment comprises a processing means, for example, a computer, or a programmable logic device, configured or adapted to carry out one of the methods described herein.

Another embodiment comprises a computer having the computer program installed in it to carry out one of the methods described herein.

In some embodiments, a programmable logic device (e.g., a field of programmable gate arrays) may be used to carry out some or all of the functionalities of the methods described herein. In some embodiments, a field of programmable gate arrays can collaborate with a microprocessor in order to carry out one of the methods described herein. In general, the methods are preferably carried out by any hardware apparatus.

The embodiments that have been described in the foregoing are merely illustrative of the principles of the present invention. It is to be understood that the modifications and variants of the arrangements and details described herein will be apparent to those skilled in the art. Accordingly, the invention is limited only by the scope of the following claims of the patent and not by the specific details presented by way of description and explanation of the embodiments included herein.

Claims (19)

An audio decoder for decoding a multichannel encoded audio signal (100), the encoded multi-channel audio signal (100) comprising a first encoded combination signal (104) generated on the basis of a combination rule for combining a first signal channel audio and a second channel audio signal of a multichannel audio signal, a residual encoded prediction signal (106) and prediction information (108), comprising: a signal decoder (110) for decoding the first encoded combination signal (104) to obtain a first decoded combination signal (112) and for decoding the residual encoded prediction signal (106) to obtain a decoded residual signal (114) ; Y a decoder calculator (116) for calculating a decoded multichannel audio signal having a first decoded channel signal (117), and a second decoded channel signal (118) using the decoded residual signal (114), the prediction information (108) and the first decoded combination signal (112), such that the first decoded channel signal (117) and the second decoded channel signal (118) are at least approximations of the first channel signal and the second channel signal of the multichannel audio signal, wherein the prediction information (108) comprises an imaginary portion other than zero, wherein the decoder calculator (116) comprises: a predictor (1160) for applying the prediction information (108) to the first decoded combination signal (112) or to a signal (601) from the first decoded combination signal (112) to obtain a prediction signal (1163) ; a combination signal calculator (1161) for computing a second combination signal (1165) by combining the decoded residual signal (114) and the prediction signal (1163); Y a combiner (1162) for combining the first decoded combination signal (112) and the second combination signal (1165) to obtain the decoded multi-channel audio signal having the first decoded channel signal (117) and the second channel signal decoded (118), wherein the predictor (1160) comprises a real-to-imaginary converter (1160) for estimating (1160a) an imaginary spectrum of the first decoded combination signal (112) using a real part of the first decoded combination signal (112) directly in the frequency domain using a two dimensional filtering, the real part of the first decoded combination signal (112) being subjected to a window switching, wherein the predictor (1160) is configured to multiply an imaginary part (601) of the first decoded combination signal (112) by the imaginary part of the prediction information (108) to obtain at least a part of the prediction signal (1163), and wherein the predictor (1160) is configured to receive a window form information (109) and to use different filter coefficients by the imaginary real converter (1160) to calculate the imaginary spectrum of the first decoded combination signal ( 112), wherein the different filter coefficients depend on different window shapes that are indicated by the window shape information (109), wherein the filter coefficients that are used by the predictor (1160) depend on a full window , and where a set of filter coefficients is required for each type of window and for each window transition. 2. Audio decoder according to claim 1, wherein the first coded combination signal (104) and the residual coded prediction signal (106) have been generated using an overlap that generates the spectral-time conversion, wherein the decoder further comprises: a spectral-time converter (52, 53) to generate a first time-domain channel signal and a second time-domain channel signal using a spectral-time conversion algorithm that matches the spectral conversion algorithm - weather; an overlap / aggregate processor (522) to perform an overlap processing - add for the first time signal channel signal and for the second time domain channel signal to obtain a first signal in the domain of time without overlapping and a second signal in the time domain without overlapping. Audio decoder according to one of the preceding claims, wherein the prediction information (108) additionally comprises a non-zero real factor, wherein the predictor (1160) is configured to multiply the first decoded combination signal (112) by the real factor to obtain a first part of the prediction signal (1163), and wherein the combination signal calculator (1161) is configured to linearly combine the decoded residual signal (114) and the first part of the prediction signal (1163) and the at least part of the residual signal of prediction. 4. Audio decoder according to one of the preceding claims, wherein the first encoded combination signal (104) or the first decoded combination signal (112) and the residual encoded prediction signal (106) or the decoded predicted residual signal (114) each comprise a first plurality of sub-band signals, wherein the prediction information (108) comprises a second plurality of prediction information parameters, the second plurality being smaller than the first plurality, wherein the predictor (1160) is configured to apply the same prediction parameter to at least two signals of different subbands of the first decoded combination signal (112), wherein the decoder calculator (116) or the combination signal calculator (1161) or the combiner (1162) are configured to perform subband processing; Y wherein the audio decoder further comprises a bank of synthesis filters (52, 53) for combining signals of sub-bands of the first decoded combination signal (112) and the second decoded combination signal (1165) to obtain a first decoded signal in the time domain and a second signal decoded in the time domain. 5. Audio decoder according to claim 1, wherein the first decoded combination signal (112) comprises a sequence of frames of signals of real value, and wherein the predictor (1160) is configured to estimate (1160a), as the imaginary spectrum of the first decoded combination signal, an imaginary part of the current signal frame using only the current real value signal frame or by using the present actual value signal and, either only one or more preceding real value signal frames or only one or more real value frames following or using the current real value signal frame and one or more real value signal frames precedents and one or more following real value signal tables. Audio decoder according to one of claims 1 to 5, wherein the encoded multi-channel audio signal (100) comprises, as a side information, a real indicator indicating that all the prediction coefficients for a signal box encoded multi-channel audio (100) are of real value, wherein the audio decoder is configured to extract the actual indicator from the encoded multi-channel audio signal (100), and wherein the decoder calculator (116) is configured not to calculate an imaginary signal for a frame, for which the real indicator is indicating only real value prediction coefficients. 7. Audio decoder according to claim 1, wherein the spectrum of a previous frame or the spectrum of a next frame is a MDCT spectrum, wherein the filter coefficients that are applied to the spectrum of the previous frame or that are apply to the spectrum of the following table depend only on the half of window that overlaps a current frame, where only a set of coefficients for each type of window is required. 8. Audio decoder according to claim 1, wherein a window type is either a sine window or a Kaiser Bessel Derived window and, subject to a given window sequence configuration, the window type can be be a long window, a start window, a stop window, a start stop window or a short window. 9. Audio decoder according to claim 1, wherein a window configuration can be a long window window, short window, start window, stop window or start stop window configuration. 10. An audio decoder according to claim 1, wherein the predictor (1160) is configured to calculate an MDST spectrum as the imaginary spectrum using a MDCT spectrum of a current frame as the real part of the first combination signal. decoded (112), wherein the filter coefficients that are used by the predictor (1160) are MDST filter coefficients and depend on the window shape of a left half of the current window and the right half of the current window, where: either the left half is a breast form and the right half is a breast shape, or the left half is a Derived form of Kaiser Bessel and the right half is a Derived form of Kaiser Bessel, or the left half is a breast form and the right half is a Derived form of Kaiser Bessel, or the left half is a Derived form of Kaiser Bessel and the right half is a breast form. 11. Audio decoder according to claim 1, wherein the predictor (1160) is configured to calculate an MDST spectrum as the imaginary spectrum using an MDCT spectrum of a current frame using the following MDST filter coefficients that are select for a left half and a corresponding right half of a current window and a corresponding current window sequence according to the following Table A: Table A - MDST Filter Coefficients for the Current Window

12. Audio decoder according to claim 11, wherein the predictor (1160) is configured to calculate an MDST spectrum as the imaginary spectrum using, in addition, a MDCT spectrum from a previous frame and using the following filter coefficients of MDST that are selected for a left half of a current window and a corresponding current window sequence according to the following Table B: Table B - MDST Filter Coefficients for the Previous Window

13. Audio encoder for encoding a multichannel audio signal having two or more channel signals, comprising: an encoder calculator (203) for calculating a first combination signal (204) and a residual prediction signal (205) using a first channel signal (201) and a second channel signal (202) and a prediction information ( 206), in such a way that a residual signal of prediction (205), when combined with a signal of prediction (2035) from the first signal of combination (204) or a signal from the first signal of combination (204) and the prediction information (206) results in a second combination signal (2032), the first combining signal (204) and the second combining signal (2032) being derived from the first channel signal (201) and the second channel signal (202) using a combination rule; an optimizer (207) for calculating the prediction information (206), such that the residual prediction signal (205) fulfills an optimization objective (208); a signal encoder (209) for encoding the first combination signal (204) and the residual signal of prediction (205) to obtain a first coded combination signal (210) and a residual encoded prediction signal (211); Y an output interface (212) for combining the first coded combination signal (210), the residual encoded prediction signal (211) and the prediction information (206) to obtain a multi-channel encoded audio signal, wherein the encoder calculator (203) comprises: a combiner (2031) for combining the first channel signal (201) and the second channel signal (202) in two different ways to obtain the first combination signal (204) and the second combination signal (2032); a predictor (2033) for applying the prediction information (206) to the first combination signal (204) or a signal (600) from the first combination signal (204) to obtain the prediction signal (2035); Y a residual signal calculator (2034) to calculate the residual signal of prediction (205) by means of the combination of the prediction signal (2035) and the second combination signal (2032), where the predictor (2033) is configured to multiply the first combination signal (204) by a real part (2073) of the prediction information (206) to obtain a first part of the prediction signal (2035); to estimate (2070) an imaginary part (600) of the first combination signal using the first combination signal (204), wherein the predictor (2033) comprises a real-to-imaginary converter (2070) to estimate, directly on the frequency domain, an imaginary spectrum of the first combination signal as the imaginary part (600) of the first combination signal using the first combination signal (204) using two-dimensional filtering, the first combination signal being submitted (204 ) to a window switching; Y to multiply the imaginary part (600) of the first combination signal by an imaginary part (2074) of the prediction information (206) to obtain a second part of the prediction signal (2035); wherein the residual calculator (2034) is configured to linearly combine the first part of the prediction signal (2035) or the second part of the prediction signal (2035) and the second combination signal (2032) to obtain the residual signal of prediction (205), and wherein the predictor (2033) is configured to receive a window shape information (109) and to use different filter coefficients to calculate, using the imaginary real converter (2070), the imaginary spectrum of the first combination signal , wherein the different filter coefficients depend on different window shapes that are indicated by the window shape information (109), wherein the filter coefficients that are used by the predictor (2033) depend on a complete window, and where a set of filter coefficients is required for each type of window and for each window transition. 14. Audio encoder according to claim 13, wherein the predictor (2033) comprises a quantizer for quantizing the first channel signal, the second channel signal, the first combination signal (204), or the second signal of combination (2023) to obtain one or more quantized signals, and wherein the predictor (2033) is configured to calculate the residual prediction signal (205) using quantized signals. 15. Audio encoder according to one of claims 13 to 14, wherein the first channel signal is a spectral representation of a block of samples; wherein the second channel signal is a spectral representation of a block of samples, wherein the spectral representations are either pure real spectral representations or pure imaginary spectral representations, wherein the optimizer (207) is configured to calculate the prediction information (206) as a non-zero real value factor and / or as an imaginary non-zero factor, and wherein the encoder calculator (203) is configured to calculate the first combination signal (204) and the residual prediction signal (205), such that the prediction signal (2035) is derived from the pure real spectral representation or the pure imaginary spectral representation using the real value factor. 16. Audio encoder according to one of claims 13 to 15, wherein the first channel signal is a spectral representation of a block of samples; wherein the second channel signal is a spectral representation of a block of samples, wherein the spectral representations are either pure real spectral representations or pure imaginary spectral representations, wherein the optimizer (207) is configured to calculate the prediction information (206) as a non-zero real value factor and / or as an imaginary non-zero factor, and wherein the predictor (2033) of the encoder calculator (203) comprises the real to imaginary converter (2070) or an imaginary to real transformer for deriving a transformed spectral representation from the first combination signal (204), and wherein the encoder calculator (203) is configured to calculate the first combination signal (204) and the residual signal of prediction (205), such that the residual signal of prediction (205) is derived from the transformed spectral representation using the imaginary factor. Method for decoding a multi-channel encoded audio signal (100), the multi-channel encoded audio signal (100) comprising a first encoded combination signal (104) generated on the basis of a combination rule for combining a first audio signal of channel and a second channel audio signal of a multi-channel audio signal, a residual signal of coded prediction (106) and prediction information (108), comprising: decoding (110) the first coded combination signal (104) to obtain a first decoded combination signal (112), and decoding the coded predicted residual signal (106) to obtain a decoded residual signal (114); Y calculating (116) a decoded multi-channel audio signal having a first decoded channel signal (117), and a second decoded channel signal (118) using the decoded residual signal (114), the prediction information (108) and the first decoded combination signal (112), such that the first decoded channel signal (117) and the second decoded channel signal (118) are at least approximations of the first channel audio signal and the second signal of channel audio of the multichannel audio signal, wherein the prediction information (108) comprises an imaginary portion other than zero, wherein the calculation of the decoded multichannel audio signal (116) comprises: applying the prediction information (108) to the first decoded combination signal (112) or to a signal (601) from the first decoded combination signal (112), to obtain a prediction signal (1163); calculating a second combination signal (1165) by combining the decoded residual signal (114) and the prediction signal (1163); Y combining the first combination signal (112) and the second combination signal (1165) decoded to obtain the decoded multi-channel audio signal having the first decoded channel signal (117) and the second decoded channel signal (118), wherein applying the prediction information (108) comprises estimating (1160a) an imaginary spectrum of the first decoded combination signal (112) using a real part of the first decoded combination signal (112) directly in the frequency domain using a two-dimensional filtering in a real-to-imaginary converter (1160), the real part of the first decoded combination signal (112) being subjected to a window switching, wherein the application of the prediction information (108) comprises multiplying an imaginary part (601) of the first decoded combination signal (112) by the imaginary part of the prediction information (108) to obtain at least a portion of the signal of prediction (1163), and wherein the application of the prediction information (108) comprises receiving a window shape information (109) and using different filter coefficients by the imaginary real converter (1160) to calculate the imaginary spectrum of the first combination signal decoded (112), wherein the different filter coefficients depend on different window shapes that are indicated by the window shape information (109), where the filter coefficients depend on a complete window, and where a set of filter coefficients for each type of window and for each window transition. 18. Coding method of a multi-channel audio signal having two or more channel signals, comprising: calculating (203) a first combination signal (204) and a residual signal of prediction (205) using a first channel signal (201) and a second channel signal (202) and a prediction information (206), such that a residual signal of prediction, when combined with a prediction signal (2035) from the first combination signal ( 204) or a signal from the first combination signal (204) and the prediction information (206) results in a second combination signal (2032), with the first combination signal (204) and the second signal being derived. combination (2032) from the first channel signal (201) and the second channel signal (202) using a combination rule; calculating (207) the prediction information (206), such that the residual prediction signal (205) meets an optimization objective (208); encoding (209) the first combination signal (204) and the residual prediction signal (205) to obtain a first coded combination signal (210) and a residual coded prediction signal (211); and combining (212) the first coded combination signal (210), the residual encoded prediction signal (211) and the prediction information (206) to obtain a multi-channel encoded audio signal, wherein the calculation (203) comprises: combining the first channel signal (201) and the second channel signal (202) in two different ways to obtain the first combination signal (204) and the second combination signal (2032); applying the prediction information (206) to the first combination signal (204) or a signal (600) from the first combination signal (204) to obtain a prediction signal (2035); and calculating the residual signal of prediction (205) by means of the combination of the prediction signal (2035) and the second combination signal (2032), wherein the application of the prediction information (206) comprises: multiplying the first combination signal (204) by a real part (2073) of the prediction information (206) to obtain a first part of the prediction signal (2035); estimating (2070) an imaginary part (600) of the first combination signal using the first combination signal (204) wherein the application of the prediction information (206) comprises estimating, by a real-to-imaginary converter (2070) , directly in the frequency domain, an imaginary spectrum of the first combination signal as the imaginary part (600) of the first combination signal using the first combination signal (204) by means of two-dimensional filtering, the first combination signal (204) to a window switching; Y multiplying the imaginary part (600) of the first combination signal (204) by an imaginary part (2074) of the prediction information (206) to obtain a second part of the prediction signal (2035); wherein the calculation of the residual signal (2034) comprises linearly combining the first part of the prediction signal (2035) or the second part of the prediction signal (2035) and the second combination signal (2023) to obtain the residual signal of prediction (205), and wherein the application of the prediction information (206) comprises receiving a window shape information (109) and using, by the imaginary real converter (2070), different filter coefficients to calculate the imaginary spectrum of the first signal of combination, wherein the different filter coefficients depend on different window shapes that are indicated by the window shape information (109), wherein the filter coefficients depend on a complete window, and where a set of filter coefficients for each type of window and for each window transition. 19. Computer program that is adapted to carry out, when executed in a computer or in a processor, the method of claim 17 or the method of claim 18.