SE527670C2 - Natural fidelity optimized coding with variable frame length - Google Patents

Abstract

A method of encoding multi-channel audio signals comprises generating of a first output signal (x' mono ), being encoding (38) parameters representing a main signal (x mono ). The main signal (x mono ) is a first linear combination (34) of signals (16A,16B) of at least a first and a second channel. The method further comprises generating (30) of a second output signal (p side ), being encoding parameters representing a side signal (x side ). The side signal (x side ) is a second linear combination (36) of signals (16A,16B) of at least the first and the second channel within an encoding frame. The method is characterised in that the generating of the second output signal further comprises scaling of the side signal (x side ) to an energy contour of the main signal (x mono ). A method of decoding is also presented as well as an encoder, a decoder and audio system, all according to the same basic idea.

Description

The signals for the different channels are processed together, rather than separately and individually. The two most commonly used co-stereo coding techniques are known as "Mid / Side" stereo coding (M / S) and intensity stereo coding, which are normally applied to subbands of the stereo or multi-channel signals to be coded.

M / S stereo coding is similar to the procedure described in stereo FM radio, in that it encodes and transmits the sum and difference signals for the subbands of the channels and thereby utilizes redundancy between the subbands of the channels. The structure and operation of an encoder based on M / S stereo coding is described e.g. in U.S. Patent No. 5,285,498 to J.D. Johnston.

Intensity stereo, on the other hand, can handle the lack of stereo connections. It transmits the common intensity of the channels (in the different subbands) together with some position information indicating how the intensity is distributed between the channels. Intensity stereo only provides information about the spectral magnitude of the channels. Phase information is not transported. For this reason and since the time-related interchannel information (more specifically the time difference between the channels) is of great psychoacoustic relevance especially at lower frequencies, intensity stereo can only be used at high frequencies above e.g. 2 kHz. An intensity coding method is described e.g. in the European patent EPO497413 by R. Veldhuis et al.

A recently developed stereo coding method is described e.g. in a conference paper entitled "Binaural cue coding applied to stereo and multi-channel audio compression", 112th AES convention, May 2002, Munich, Germany by C. Faller et al. This method is a parametric audio coding method for multichannels. The basic principle is to combine the input signals from N channels ci, cg, cN to a mono signal m on the coding side. The mono signal is audio coded using any conventional mono audio codec. In parallel with this, parameters are derived from the channel signals, which describe the multi-channel image.

The parameters are encoded and transmitted to the decoder, together with the audio bitstream. The decoder first decodes the mono signal m 'and then regenerates the channel signals ef, c2', ..., cN ', based on the parametric description of the multichannel image.

The principle of the binaural fit coding procedure (BCC, eng.

"Binaural Cue Coding") is that it transmits the coded mono signal and so-called coded level differences between the channels and time differences between the channels for BCC parameters. The BCC parameters include subbands of the original multi-channel input signal. The decoder regenerates the different channel signals by applying level and phase adjustments for each subband of the mono signal, based on the BCC parameters. The advantage over e.g. M / S or intensity stereo is that stereo information that includes time-related interchannel information is transmitted at much lower bit rates. However, this technique requires computationally demanding time-frequency transformers for each of the channels, both at the encoder and at the decoder.

Furthermore, BCC does not handle the fact that much of the stereo information, especially at low frequencies, is diffuse, ie. it does not come from any particular direction. Diffuse sound fields are present in both channels of a stereo recording, but they are largely out of phase in relation to each other. If an algorithm such as BCC is subjected to recordings with a large number of diffuse sound fields, the reproduced stereo image will be confused and jump from left to right because the BCC algorithm can only direct the signal in specific frequency bands to the right or left.

A possible means of coding the stereo signal and ensuring a good reproduction of diffuse sound fields is to use a coding scheme which is very similar to the technology used in broadcasting FM stereo radio, namely to encode the mono signal (left + right) and the difference signal (left-right) separate. 527 670 A technique, described in U.S. Patent No. 5,434,948 to C.E. Holt et al., Use a similar technique as in BCC to encode the mono signal and side information. In this case, the side information consists of predictors or possibly a residual signal. The predictor filters, estimated by a least squares algorithm, when applied to the mono signal, enable the prediction of multi-channel audio signals. With this technology, very low bit rate encoding of multi-channel audio sources can be achieved, however, at the expense of a quality degradation, which is discussed further below.

Finally, for the sake of completeness, a technology used in 3D audio must be mentioned. This technique synthesizes the right and left channel signals by filtering audio source signals with so-called main-related filters.

However, this technology requires that the various audio source signals be separated and thus not generally applicable to stereo or multi-channel coding.

SUMMARY A problem with existing coding schemes based on coding of signal frames, in particular a main signal and one or fl your side signals, is that the division of audio information into frames can introduce unattractive perception artifacts. Dividing the information into frames of relatively long duration generally reduces the average required bit rate.

This can be advantageous, e.g. for music that contains a large amount of diffused sound. For transient-rich music or speech, however, the rapid time variations will be smeared over the duration of the frame, giving rise to spooky sounds or even pre-echo problems. Coding short frames will instead provide a more accurate representation of the sound, which minimizes energy, but requires higher transmission bit rates and greater computational resources. The coding efficiency as such can thus decrease with very short frame lengths. The introduction of more frame boundaries can also introduce discontinuities in coding parameters, which can appear as perception artifacts.

A further problem with schemes based on encoding a main signal and one or more side signals is that they often require relatively large computational resources. Especially when short frames are used, handling discontinuities in parameters from one frame to another is a complicated task. When long frames are used, estimation errors for transient sounds can cause very large side signals, which in turn increases the requirement for transmission speed.

An object of the present invention is therefore to provide an encoding method and an encoding device which improves the Perceptual quality of multi-channel audio signals, in particular to avoid artifacts such as squeaking, ghost-like sounds or frame discontinuity artifacts. A further object of the present invention is to provide a coding method and a coding device which requires less processing power and which has more constant bit rate requirements during transmission.

The above objects are achieved by methods and devices according to the appended claims. In general terms, polyphonic signals are used to create a main signal, typically a mono signal, and a side signal.

The main signal is encoded according to coding principles according to the prior art.

A number of coding schemes for the side signal are provided. Each coding scheme is characterized by a set of subframes of different lengths.

The total length of the subframes corresponds to the length of the coding frame of the coding scheme. The sets of subframes include at least one subframe. The coding scheme to be used on the side signal is selected at least in part depending on the current signal content of the polyphonic signals. In one embodiment, the selection takes place either before the coding, based on signal characteristic analysis. In another embodiment, the side signal is encoded with each of the encoding schemes, and based on measurements of the encoding quality, the best encoding scheme is selected.

In a preferred embodiment, a residual side signal is created such as the difference between the side signal and the main signal scaled by a balance factor.

The balance factor is selected to minimize the residual side signal. The optimized residual side signal and the balance factor are coded and provided as parameters representing the side signal. On the decoder side, the balance factor, the residual side signal and the main signal are used to reset the side signal.

In a further preferred embodiment, the coding of the side signal comprises an energy contour scaling to avoid pre-echo effects. Furthermore, different coding schemes may include different coding procedures in the separate subframes.

The main advantage of the present invention is that the maintenance of the perception of the audio signals is improved. Furthermore, the present invention still allows the transmission of multichannel signals at very low bit rates.

BRIEF DESCRIPTION OF THE DRAWINGS The invention, together with further objects and advantages thereof, may best be understood by reference to the following description of words taken in conjunction with the accompanying drawings, in which: FIG. 1 is a block diagram of a polyphonic signal transmission system; FIG. 2a is a block diagram of an encoder in a transmitter; FIG. 2b is a block diagram of a decoder in a receiver; FIG. 3a is a diagram illustrating coding frames of different lengths; 10 15 20 25 30 527 670 7 FIG. 3b and 3c are block diagrams of embodiments of side signal encoder units according to the present invention; FIG. 4 is a block diagram of an embodiment of an encoder using balance factor coding of the side signal; FIG. 5 is a block diagram of an embodiment of a multisignal system encoder; FIG. Fig. 6 is a block diagram of an embodiment of a decoder suitable for decoding signals from the device of Fig. 5; FIG. 7a and b are diagrams illustrating a pre-eco-artifact; FIG. 8 is a block diagram of an embodiment of a side signal encoder unit according to the present invention, which uses different coding principles in different subframes; FIG. 9 illustrates the use of different coding principles in different frequency subbands; FIG. 10 is a flow chart of the basic steps of an embodiment of a coding method according to the present invention; and FIG. 11 is a fate diagram of the basic steps in an embodiment of a decoding method according to the present invention.

DETAILED DESCRIPTION Fig. 1 illustrates a typical system 1, in which the present invention can be used to advantage. A transmitter 10 includes an antenna 12 which includes associated hardware and software for transmitting radio signals 5 to a receiver 20. The transmitter 10 includes, among other parts, a multi-channel encoder 14 which converts signals for a plurality of input channels 16 into output signals suitable for radio transmission.

Examples of suitable multi-channel encoders 14 are described in detail below.

The signals of the input channels 16 can be provided from e.g. an audio signal storage 18, such as a data cell with digital representation of audio recordings, magnetic tape or vinyl record recordings of sounds, etc. The signals of the input channels 16 can also be provided "live", e.g. from a set of microphones 19. The audio signals, if not already in digital form, are digitized before entering the multi-channel encoder 14.

At the receiver side 20, an antenna 22 with associated hardware and software handles the actual reception of the radio signals 5 representing polyphonic audio signals. Here, typical functionalities are performed, such as e.g. error correction. A decoder 24 decodes the received radio signals 5 and converts the audio data carried therefrom into signals for a number of output channels 26. The output signals can be provided to e.g. speakers 29 for immediate presentation, or can be stored in some kind of audio signal storage 28.

The system 1 may be, for example, a telephone conference system, a system for providing audio service or other audio applications. In some systems, such as e.g. telephone conference systems, the communication must be of the duplex type, while e.g. Distribution of music from a service provider to a subscriber can be of a substantially one-way type.

The transmission of the signals from the transmitter 10 to the receiver 20 can also be performed by any other means, e.g. both different types of electromagnetic waves, cables or fibers and combinations thereof.

Fig. 2a illustrates an embodiment of an encoder according to the present invention. In this embodiment, the polyphonic signal is a stereo signal comprising two channels a and b, received at inputs 16A and 16B, respectively. The signals in channels a and b are provided to a preprocessor 32, where various signal conditioning procedures can be performed. The (perhaps modified) signals from the output of the preprocessor 32 are summed in an addition unit 34. This addition unit 34 also divides the sum by a factor of two. The signal xmono formed in this way is a main signal for the stereo signal, since it basically includes all the data from both channels. Thus, in this embodiment, the main signal represents a pure "mono" signal. The main signal xmono is provided to a main signal encoder unit 38, which encodes the main signal according to any suitable coding principle. Such principles are available from the state of the art and are therefore not discussed further here. The main signal encoder unit 38 provides an output signal pmom which consists of coded parameters representing a main signal.

In a subtraction unit 36, a difference (divided by a factor of two) between the channel signals is provided as a side signal xslae. In this embodiment, the side signal represents the difference between the two channels in the stereo signal. The side signal xside is provided to a side signal encoder unit 30. Preferred embodiments of the side signal encoder unit 30 will be further discussed below. According to a side signal coding procedure, which will be described in more detail below, the side signal xside is transmitted to the coding parameters pad, which represents a side signal xside. In some embodiments, this encoding takes place by also utilizing information from the main signal xmono. Arrow 42 indicates such a provision, where the original uncoded main signal xmono is used. In still other embodiments, the main signal information used in the side signal encoder 30 may be derived from the coding parameters pmm representing the main signal, as indicated by the dashed line 44.

The pmono coding parameters representing the main signal xmono are a first output signal, and the pside coding parameters representing the side signal xside are a second output signal. Typically, these two output signals are multiplexed pmono, psme which together represent the entire stereo sound of a transmission signal 52 in a multiplexing unit 40. In other embodiments, however, the transmission of the first and second output signals pmono, psme may take place separately.

I fi g. 2b, an embodiment of a decoder 24 according to the present invention is illustrated as a block diagram. The received signal 54 which includes coding parameters representing the main and side signal information is provided to a demultiplexing unit 56, which separates a first and a second input signal, respectively. The first input signal corresponding to the coding parameters pmom for a main signal is provided to a main signal decoder 64. In a conventional manner, the coding parameters pmom representing the main signal are used to form a decoded main signal x "m ° n °, which is so similar to the main signal xmon. 2a) in the encoder 14 (Fig. 2a) as possible.

On. similarly, the second input signal corresponding to a side signal is provided to a side signal decoder 60. Here, the pside coding parameters representing the side signal are used to reset a decoded side signal x "sme. In some embodiments, the decoding procedure utilizes information about the main signal x" m. , as indicated by arrow 65.

The decoded main and side signals x fi nonc, x "smc are provided to an addition unit 70, which provides an output signal which is a representation of the original signal for channel a. Similarly, a difference provided by a subtraction unit 68 provides an output signal which is a representation of the original signal for channel b. These channel signals can be post-processed in a post-processing unit 74 according to prior art signal processing procedures.

Finally, the channel signals a and b are provided to the outputs 26A and 26B of the decoder.

As mentioned in the summary, coding is typically performed in one frame at a time. A frame includes sound snippets within a predetermined time period. In the lower part of ñg. 3a illustrates a frame SF2 with a time duration L.

The audio samples within the non-dashed part must be coded together. The previous samples and the subsequent samples are encoded in other frames.

The division of the samples into frames will in any case introduce some discontinuities at the frame boundaries. Changed sounds will result in changed coding parameters, which in principle change at each frame limit.

This will give rise to perceptual errors. One way to compensate for this is to base the coding, not only on the samples to be coded, but also on samples in the absolute environment of the frame, as indicated by the dashed portions. In such a way, there will be a smoother transition between the different frames. As an alternative, or complement, interpolation techniques can sometimes also be used to reduce perception artifacts caused by frame boundaries. However, all such procedures require large additional computational resources, and for some specific coding techniques, it may also be difficult to provide with any resources.

From this point of view, it is advantageous to use as long frames as possible, as the number of frame limits will be small. The coding efficiency will also typically be high and the necessary transmission bit rate will typically be minimized. However, long frames cause problems with pre-eco-artifacts and ghost-like sounds.

By using short frames instead, such as SF1 or even SFO, which have the durations L / 2 and L / 4, respectively, those skilled in the art will recognize that the coding efficiency may decrease, the transmission bit rate may need to be higher and the problems with boundary artifacts will increase. However, shorter frames suffer less from e.g. other perceptual artifacts, such as ghostly sounds and echoes. In order to minimize the coding error as much as possible, you should use as short a frame length as possible.

According to the present invention, the sound perception will be improved by using a frame length for encoding the side signal which is dependent on the actual signal content. Since the influence of different frame lengths on the sound perception will differ depending on the nature of the sound to be coded, an improvement can be obtained by allowing the nature of the signal itself to affect the frame length used. The main signal coding is not the object of the present invention and is therefore not described in detail. However, the frame lengths used for the main signal may or may not be equal to the frame lengths used for the side signal. 10 15 20 25 30 527 670 12 Due to small time variations, it can e.g. in some cases it may be advantageous to encode the side signal using relatively long frames. This may be the case with recordings with a large amount of diffused sound field such as concert recordings. In other cases, such as stereo calls, short frames may be preferable. Deciding which frame length is preferable can be done in two basic ways.

An embodiment of a side signal encoder unit 30 according to the present invention is illustrated in Figs. 3b, in which a decision in a closed circuit is used.

A basic coding frame of length L is used here. A number of coding schemes 81, characterized by a separate set 80 of subframes 90, are created. Each set 80 of subframes 90 includes one or more subframes 90 of equal or different lengths. However, the total length of the set 80 of subframes 90 is always equal to the basic coding frame length L. Referring to fi g. 3b, the top coding scheme is characterized by a set of subframes comprising only one subframe of length L. The next set of subframes comprises two frames of length L / 2. The third set comprises two frames of length L / 4, accompanied by an L / 2 -frame.

The Xside signal provided to the side signal encoder 30 is encoded by all encoding schemes 81. In the top encoding scheme, the entire basic encoding frame is encoded in one piece. However, in the other coding schemes, the signal xsmc in each subframe is coded separately from each other.

The result of each coding scheme is provided to a selector 85. A natural fidelity measurement means 83 determines a natural fidelity measure for each of the coded signals. The natural fidelity measure is an objective quality value, preferably a signal-to-noise measure or a weighted signal-to-noise ratio. The natural fidelity measures associated with each coding scheme are compared and the result controls a switching means 87 for selecting the coding parameters representing the side signal 10 from the coding scheme giving the best natural fidelity measure as the output signal pad, from the side signal encoder 30.

Preferably, all possible combinations of frame lengths and the set of subframes that provide the best objective quality are tested, e.g. signal-to-noise ratio is selected.

In the present embodiment, the lengths of the subframes used are selected according to: 1. where 1, are the lengths of the subframes, I, is the length of the coding frame and n is an integer. In the present embodiment, n is chosen between 0 and 3. However, any frame lengths may be possible to use, as long as the total length of the set is kept constant.

I fi g. 3c illustrates another embodiment of a side signal encoder unit 30 according to the present invention. Here, the frame length decision is an open circuit decision, based on the signal statistics. In other words, the spectral characteristics of the side signal will be used as a basis for deciding which coding scheme to use. As before, different coding schemes characterized by different sets of subframes are available. In this embodiment, however, the selector 85 is located before the actual coding. The input side signal xsme enters the selector 85 and a signal analysis unit 84. The result of the analysis is the input signal to a switch 86, in which only one of the coding schemes 81 is used.

The output signal from that encoding scheme will also be the output signal pad, from the side signal encoder unit 30.

The advantage of a decision in an open circuit is that only an actual coding needs to be performed. The disadvantage, however, is that the analysis of the signal characteristics can be really very complicated and it can be difficult to predict possible behaviors in advance in order to be able to make a learning choice in the switch 86. A large amount of statistical analysis of sound must performed and included in the signal analysis unit 84. Any small change in the coding schemes can turn the statistical behavior upside down.

By using closed circuit selection (Fig. 3b), coding schemes can be replaced without making any changes to the rest of the unit. On the other hand, if many coding schemes have to be examined, the computational requirements will be high.

The advantage of such a variable frame length coding for the side signal is that one can choose between on the one hand a fine time resolution and coarse frequency resolution and on the other hand a coarse time resolution and fine frequency resolution. The above embodiments will preserve the stereo image in the best possible way.

There are also some requirements for the actual codings used in the various coding schemes. Particularly when the closed circuit selection is used, the computational resources to perform a number of more or less simultaneous encodings must be large. The more complicated the coding process, the more computational credit is needed. Furthermore, a low bit rate during transmission is preferred.

The method presented in US 5,434,948 uses a filtered version of the mono signal (the main signal) to mimic the side or difference signal. The filter parameters are optimized and allowed to vary over time.

The filter parameters are then transmitted, which represent an encoding of the side signal. In one embodiment, a residual side signal is also transmitted. In many cases, such an approach would be possible to use as a side signal coding method within the scope of the present invention. However, this approach has some disadvantages. The quantization of the filter coefficients and any residual side signal often requires relatively high bit rates for transmission, since the filter order must be high to provide an accurate side signal estimate. The filter estimation itself can be problematic, especially in the case of transient-rich music. Estimation errors will give a modified side signal which is sometimes larger in magnitude than the unmodified signal. This will lead to higher bit rate requirements.

Furthermore, if a new set of koltercoefficients is calculated every N sample, the koltercoefficients need to be interpolated to achieve a smooth transition from one set of koltercoefficients to another, as discussed above.

Interpolation of ko ltercoefficients is a complicated task and errors in the interpolation will manifest themselves in larger side error signals, which leads to higher bit rates being needed for the difference error signal encoder.

One means of avoiding the need for interpolation is to update the ko lter coefficients on a sample-by-sample basis and to rely on backward adaptive analysis. For this to work well, it is required that the bit rate of the residual encoder is quite high. This is therefore not a good option for low bit rate stereo coding.

There are cases, e.g. quite common with music, where the mono and difference signals are almost uncorrelated. The filter estimation then becomes very worrying with the added risk of simply making things worse for the difference error signal encoder.

The solution according to US 5,434,948 can work quite well in cases where the filter coefficients vary very slowly over time, e.g. conference telephony system. In the case of music signals, this approach does not work so well because the filters need to be changed very quickly to follow the stereo image.

This means that subframe lengths of very different sizes must be used, which means that the number of combinations to test is rapidly increasing. This in turn means that the requirements for calculating all possible coding chips become impractically high. Therefore, in a preferred embodiment, the coding of the side signal is based on the idea of reducing the redundancy between the mono and side signals by balancing factor instead of a bit rate consuming predictor filter. The rest of this operation is coded using a simple complex sedan. The size of such a residue is relatively small and does not give rise to very high bit rate requirements for transmission. This idea is really very suitable to combine with the variable frame set approach described earlier, as the computational complexity is low.

The use of a balance factor combined with the variable frame length approach removes the need for complex interpolation and the associated problems that interpolation can cause. Furthermore, the use of a simple balance factor instead of a complex filter gives fewer problems with estimates because possible estimation errors for the balance factor have less impact. The preferred solution will be able to recreate both panning signals and diffused sound fields with good quality and with limited bit rate requirements and computational resources.

Fig. 4 illustrates a preferred embodiment of a stereo encoder according to the present invention. This embodiment is very similar to that shown in Fig. 2a, however with the details of the side signal encoder 30 revealed.

The encoder 14 in this embodiment has no preprocessor and the input signals are provided directly to the addition and subtraction units 34, 36. The mono signal xmono is multiplied by a certain balance factor gm in a multiplier 33. In a subtraction unit 35 the multiplied mono signal is subtracted from the side signal x. . essentially the difference between the two channels, to give a residual signal. The balance factor gsm is determined based on the content of the mono and side signals by the optimizer 37 to minimize the side residual signal according to a quality criterion. The quality criterion is preferably a least squares criterion. The residual signal is encoded in a residual encoder 39 according to any coding procedures. Preferably, the page residual encoder 39 is a low bit rate transform encoder or a CELP (Codebook Excited Linear Prediction). The coding parameters psmc representing the side signal then include the coding parameters psid., Fesidum representing the side residual signal and the optimized balance factor 49.

In the execution form in fi g. 4, the mono signal 42 used to synthesize the side signals is the target signal xmom of the monocoder 38. As mentioned above (in connection with fi g. 2a), the local synthesizing signal of the monocoder 38 can also be used. In the latter case, the total encoder delay may increase and the computational complexity of the side signal may increase. On the other hand, the quality can be better because then it is possible to repair coding errors in the monocoder.

On. in a more mathematical way, the basic coding scheme can be described as follows. Denote the two channel signals such as a and b, which may be the left and right channels in a stereo pair. The channel signals are combined into a mono signal by addition and into a side signal through a subtraction. In equation form, the operations are described as: xmán) = 0-5 (a (f =) + 1201)) x,. »L., (Fl) = 0-5 (<1 (n) -b (fl)) - It is advantageous to scale down the xmono and xside signals by a factor of two.

It is suggested here that there are other ways to create xmøno and Xside. For example, you can use: XMÅH) = w (n) + (1 - r) b (~) nu. (n) = w (n) - (1 - r) b (n) O S 7 S 1.0.

A modified or residual side signal is calculated on blocks of the input signals according to: xsideruiduul (n) = xxide (n) _ f (xmono.xxide) xmmm (n) Y 10 15 20 25 527 670 18 where fl xmuno, Xsme) is a balance factor function based on the block of N samples, i.e. a subframe, from the side and mono signals strives to remove so. as much as possible from the side signal. In other words, the balance factor is used to minimize the residual side signal. In the special case where it is minimized in a least squares sense, this is equivalent to minimizing the xsme fesidum energy of the residual side signal.

In the above-mentioned special case, f (xmmpx fl de) is described as: R f (xmono 9 xsídu) = RM MM frameend RM ", 1 í z x fi l fl llß (n) xm07 | 0 rwframesrarr frame / nd Rsm = [E xxide (ràxmono (n 2 n = fi nmesinrt where xsidc is the side signal and Xmono is the mono signal, note that the function is based on a block that starts at "frame start" and ends at "frame end".

It is possible to add weighting in the frequency domain to the calculation of the balance factor. This is done by folding the xside and xmono signals with the impulse response of a weighting filter. It is then possible to extend the estimation error to a frequency range where they are less easy to hear. This is called perception weighting.

A quantized version of the balance factor value given by the function f (xmo ,,,, x ,,. D,) is transmitted to the decoder. It is preferable to take quantization into account already when the modified side signal is created. The expression below is then obtained: 10 15 20 25 527 670 19 xside rexidual (n) = xxide 01) - g Q x mono (n) _ R ,,, gg = Q-'lßßírll ~ Qg (..) is a quantization function which applied to the balance factor given by the function f (xmww, x _ ,, de). The balance factor is transmitted on the transmission channel. In normal left-right panning signals, the balance factor is limited to the interval [-l.O 1.01. On the other hand, if the channels are out of phase with each other, the balance factor can extend beyond these limits.

As an additional means of stabilizing the stereo image, one can limit the balance factor if the normalized cross-correlation between the mono and side signal is small as given by the equation below: gg = Q§'lQ-ll§llš;: ll * where R _ .VTI. gsm V Rs: Rmm fi- ameend Run = [E xside (ràxmrmo (n 'n-fmme: um These situations occur quite often with eg classical music or studio music with a large proportion of diffused sound, where in some cases a- and the b-channels could largely extinguish each other at certain times when a mono signal is created.The effect on the balance factor is that it can jump fast, causing a confused stereo image.The fixation above alleviates this problem.10 15 20 25 30 527 670 20 The alter-based approach in US 5,434,948 has similar problems, but in that case the solution is not so simple.

If E: is the coding function (eg a transform encoder] for the residual side signal and En, is the coding function of the mono signal, then the decoded a "and b" signals at the decoder end can be described as (it is assumed here that 7 = 0.5): a ' (n) = + gQ) x; nno (n) + xzide (n) bin) = (1 - g.,) »;., .. (~> - (n) xçíde = Ex. 1 (xxide rexidual xfïi fl ïlfl = (XUNOPIO An important advantage of calculating the balance factor for each frame is that the use of interpolation is avoided. Instead, as described above, the frame machining is normally performed with overlapping frames.

The coding principle that uses balance factors works particularly well in the case of music signals, where rapid changes are typically needed to follow the stereo image.

Recently, multi-channel coding has become popular. An example is 5.1 channel surround sound in DVD movies. The channels are arranged there as: front left, front center, front right, rear left, rear right and subwoofer. Fig. 5 shows an embodiment of an encoder which encodes the three front channels in such an arrangement using interchannel redundancies according to the present invention.

Three channel signals L, C, R are provided at three inputs 16A-C, and the mono signal xmom is created by a sum of all three signals. A center signal encoder unit 130 is added, which receives the center signal xcentfe.

The mono signal 42 in this embodiment is the coded and decoded mono signal x "m ° n °, and multiplied by a certain balance factor gQ in a multiplier 133. In a subtraction unit 135, the multiplied mono signal is subtracted from the center signal Xcenue for The balance factor gQ is determined based on the content of the mono and center signals by an optimizer 137 to minimize the center residual signal according to the quality criterion The center residual signal is encoded in a center residual encoder 139 according to some encoder procedure. The encoding parameters pwme representing the center signal then include the encoding parameters pcenue fesiauai representing the center residual signal and the optimized balance factor 149. The center residual signal and the scaled mono signal are added in a component of the addition signal el.

The side signal xside, i.e. the difference between the left L and right R channels is provided to the side signal encoder unit 30 as in previous embodiments. Here, however, the optimizer 37 depends on the modified center signal 142 provided by the center signal encoder unit 130.

The side residual signal will therefore be created as an optimal linear combination of the mono signal 42, the modified center signal 142 and the side signal in the subtraction unit 35.

The variable frame length concept described above can be applied to either of the side and center signals, or to both.

Fig. 6 illustrates a decoder unit suitable for receiving coded audio signals from the encoder unit in fi g. 5. The received signal 54 is divided into coding parameters pmom representing the main signal, coding parameters Damm.- representing the center signal and coding parameters pad representing the side signal. The decoder 64 uses the coding parameters pmm, which represent the main signal, to create a main signal x "m0n0. coding parameters pside representing the side signal, which creates a side signal xäde, based on main signal x fi nono and center signal xkentfe.

The procedure can be expressed mathematically as follows: The input signals Xiefi, X fi ght and xwm are combined into a mono channel according to: xmonn (n) = æçlq / l (n) + ßxrighl (n) + xx center (n) 'a,, ß and g are i the remaining section set to 1.0 for simplicity, but they can be set to arbitrary values. the a, ß and x values can either be constant or depend on the signal contents to emphasize one or two channels to obtain an optimal quality.

The normalized cross-correlation between the mono and center signal is calculated as: where frameend _ Rcc = í E xcenlrv (n lxcenlre (n) n = frames1ari frameend _ Rmrn = I: E xmono (n) xmono (n) n = fi nmestarl frameend _ Rem = í 2 xcemne (roxmnnr: (n) 'n = fi1 zme negligence 10 15 20 25 527 670 23 xcem is the center signal and xmm is the mono signal.

The center residual signal to be coded is: xcenlre rcsidual (n) = xcenlre (n) _ g Q 'x mono (n) _ R gß = Q! (24%)) - Qg is a quantization function to be applied to the balance factor.

The balance factor is transmitted on the transmission channel.

If EC is the coding function (eg a transform encoder) for the center residual signal and Em is the coding function for the mono signal, then the decoded xjem, signal in the decoder can be described as: n _: rv: xcentre (n) '_ g Q xmono ( n) + x center residual (n) I ___ -l xcenlrefesidual _ E: c (xcenlre residual xrçiana = E: (Em (xmonn The residual signal to be coded is: x side residual (n) = (xleji (n) _ xrighr ( n _ gQrm xrzrnno (n) _ gQsc xzenlre (n) 1 where gQm and ggn are quantized values of the parameters gm and gu that minimize the expression: frame end (xlc fl (n) _ xrighl _ gsmxgnna (n) _ gxrxzenlre 'n = framesrarr 10 15 20 25 30 527 670 24 17 can, for example, be equal to 2 for a least squares minimization of the error, the gm and gR parameters can be quantized together or separately.

If E, is the coding function of the residual signal, then the decoded xfe fl and x fl gh, channel signals are given as: xlï fi (n) = xrfrono (n) _ xgenlrc (n) + xxïde (n) xííghl (n) = 'xr'; 1ono (n ) _ 'xzenlre (n) _' xš-'ide (n) xšde (n) = 'xçlkierexídual + gQsm xrçronf: (n) + gQscxgenlre (n) n _ -l xsideresidual _ E: .r (xslderexizlunl' En av the most disturbing perceptual artifacts are the pre-echo effect.In fi g. 7a-b diagram illustrates such an artifact, assume a signal component having the time evolution shown by curve 100. In the beginning, starting from tO, the signal component is not present in At a time t between t1 and t2, the signal component suddenly appears.When the signal component is encoded using a frame length of t2-t1, the occurrence of the signal component will be "smeared" over the entire frame, as indicated in curve 101. If a decoding of curve 101 takes place, the signal component appears a time föret before the intended behavior of the signal component, and a "pre-echo" is experienced.

The pre-eco-artifacts become more emphasized if long coding frames are used.

By using shorter frames, the artifact is somewhat suppressed. Another way of dealing with the pre-echo problems described above is to take advantage of the fact that the mono signal is available at both the encoder and decoder ends.

This makes it possible to scale the side signal according to the energy contour of the mono signal. In the decoder spirit, the inverse scaling is performed and some of the pre-echo problems can thus be mitigated.

An energy contour of the mono signal is calculated over the frame such as: 10 15 20 25 527 670 25 m + L Ec (m) = [zw (n) xf, w, m (n): \, frame start S m S frame end, n = mL where w (n) is a window-forming function. The simplest window-forming function is a rectangular window, but other window types such as a Hamming window may be more desirable.

The page residual signal is then scaled as: page resultual (n) x ¿_ ,, de, e _ ,,,, uu, (n) = W, frame start S n S frame end.

In a more general form, the above equation can be written as: gc __ \ _ ,. de, e _ ,, d ,, a, (n) = frame start S n S frame end, where f is a monotonic continuous function. In the decoder, the energy contour of the decoded mono signal is calculated and applied to the decoded side signal as: than * (n) = xfw (n) f (E, (n)), frame start S n S frame end.

Since this energy contour scaling is in some way an alternative to the use of shorter frame lengths, this concept is particularly well suited to be combined with the concept of variable frame length, described further above.

By having certain coding schemes that apply energy contour scaling, some that do not, and some that apply energy contour scaling only under certain subframes, a more flexible set of coding schemes can be provided. I ñg. 8, an embodiment of a signal encoder unit 30 according to the present invention is illustrated. Here, the various coding schemes 81 include dashed subframes 91, which represent coding that apply the energy contour scaling, and non-dashed subframes 92, which represent coding procedures that do not apply the energy contour scaling. In this way, combinations of not only subframes of different lengths but also of subframes with varying coding principles are available.

In the present explanatory example, the application of energy contour scaling differs between different coding schemes. In a more general case, any coding principles can be combined with the variable length concept in an analogous manner.

The set of coding schemes in fi g. 8 includes schedules that handle e.g. for-eco-artifacts in different ways. In some schemes, longer subframes with pre-eco-minimization are used according to the energy contour principle. In other schemes, shorter subframes without energy contour scaling are used.

Depending on the signal content, one of the options may be more advantageous.

For very severe pre-eco-cases, coding schemes utilizing short frames with energy contour scaling may be necessary.

The proposed solution can be used in the whole frequency band or in one or more distinct subbands. The use of subbands can be applied either to both the main and side signals, or to one of them separately. A preferred embodiment comprises a division of the side signal into several frequency bands.

The reason is simply that it is easier to remove the possible redundancy in an isolated frequency band than in the entire frequency band. This is especially important when encoding music signals with rich spectral content.

One possible use is to encode the frequency band below a predetermined threshold with the above method. The predetermined threshold may preferably be 2 kHz or even more preferably 1 kHz. For the remaining part of the frequency range of interest, one can either encode another additional frequency band with the above method or use a completely different method. A motive for using the above method, preferably for low frequencies, is that diffuse sound fields generally have a low energy content at high frequencies. The natural reason is that sound absorption typically increases with frequency. In addition, the diffuse sound field components seem to play a less important role for the human sound system at higher frequencies. Therefore, it is advantageous to use this solution at low frequencies (below 1 or 2 kHz) and rely on other, even more bit-efficient coding schemes at higher frequencies. The fact that the scheme is only applied at low frequencies provides a large saving in bit rate because the required bit rate with the proposed method is proportional to the required bandwidth. In most cases, the monocoder may encode the entire frequency band, while the proposed side signal coding is proposed to be performed only in the lower part of the frequency band, as schematically illustrated by ñg. Reference numeral 301 refers to a coding scheme according to the present invention for the side signal, reference numeral 302 refers to other coding schemes for the side signal, and reference numeral 303 refers to a coding scheme for the side signal.

There is also the possibility of using the proposed method for several distinct frequency bands.

I fi g. 10, the main steps of an embodiment of an encoding method according to the present invention are illustrated as a fate diagram.

The procedure begins in step 200. In step 210, a main signal derived from the polyphonic signals is encoded. In step 212, coding schemes are provided, which include subframes of different lengths and / or order. A side signal derived in step 214 from the polyphonic signals is encoded by a coding scheme which depends at least in part on the actual signal content of the present polyphonic signals. The procedure ends in step 299.

I fl g. 11, the main steps of an embodiment of a decoding method according to the present invention are illustrated as a fate diagram. The procedure begins in step 200. In step 220, a received coded master signal is decoded. In step 222, coding schemes are provided, which include subframes of different lengths and / or order. A received side signal is decoded in step 224 by a selected coding scheme. In step 226, the decoded main and side signals are combined into a polyphonic signal.

The procedure closed step 299.

The embodiments described above are to be understood as some illustrative examples of the present invention. It will be appreciated by those skilled in the art that various modifications, combinations and changes may be made to the embodiments without departing from the scope of the present invention. In particular, different sub-solutions in the different embodiments can be combined in other arrangements, where this is technically possible. However, the scope of the present invention is defined by the appended claims.

REFERENCES European Patent 0497413 U.S. Patent 5,285,498 U.S. Patent 5,434,948 "Binaural cue coding applied to stereo and multi-channel audio compression", 112th AES Convention, May 2002, Munich, Germany by C.

Faller et al.

Claims (26)

10 15 20 25 30 527 670 29 PATENT REQUIREMENTS A method of encoding polyphonic signals, comprising the steps of: generating (210) a first output signal (pmono), which eel 'encoding parameters representing a main signal based on signals for at least a first and a second channel (a, b). ; L, R); and generating (214) a second output signal (psidc), which are coding parameters representing a side signal based on signals for at least the first and second channels (a, b; L, R) within a coding frame (80), characterized of the further step: providing (212) at least two coding schemes (81), each of the at least two coding schemes being characterized by a respective set of subframes (90) which together form the coding frame (80), the sum of the subframes (90) lengths in each coding scheme are equal to the length of the coding frame (80); each set of subframes (90) comprises at least one subframe (90): the step of generating (214) the second output signal (psme) comprises the step of selecting an encoding scheme (81) at least in part depending on the present side signal (xsme) signal content; wherein the second output signal (psfde) is encoded in each of the subframes (90) in the selected set of subframes (90) separately. A method according to claim 1, characterized in that the step of generating (214) the second output signal (psme) in turn comprises the steps of: generating coding parameters representing a side signal (xside), which is a first linear combination of signals from at least the first and second channels (a, b; L, R), within all subframes (90) for each of the at least two sets of subframes (90) separately; calculating a total measure of natural fidelity for each of the at least two coding schemes (81); and selecting the coding signal from the coding scheme (81) which has the best measure of fidelity to nature such as the coding parameters (psiae) representing the side signal. A method according to claim 2, characterized in that the measure of fidelity to nature is based on a signal-to-noise measure. A method according to any one of claims 1 to 3, characterized in that the subframes (90) have lengths Isf according to: 1. where lf is the length of the coding frame (80) and n is an integer. A method according to claim 4, characterized in that n is less than a predetermined value. A method according to claim 5, characterized in that the at least two coding schemes (81) comprise all permutations of subframe lengths (90). A method according to any one of claims 1 to 6, characterized in that the step of generating (210) coding parameters (pmom) representing the main signal in turn comprises the steps of: creating a main signal (Xmono) as a second linear grain combination of signals from at least the first and second channels (a, b; L, R); and encoding the main signal into coding parameters (pmom) representing the main signal, the step of encoding the side signal in turn comprising the steps of: creating a side residual signal (xsme fesiauai) as a difference between the side signal and the main signal (xmono) scaled by a balance factor (gsm); Which balance factor (gsm) is determined as the factor which minimizes the side residual signal according to a quality criterion; coding of the side residual signal and the balance factor (gsm) as the coding parameters (psme) representing the side signal. A method according to claim 7, characterized in that the quality criterion is based on a minimum square dimension. A method according to any one of claims 1 to 8, characterized in that the step of encoding the side signal further comprises the step of: scaling the side signal (Xsidc) to an energy contour of the main signal (xmm). A method according to claim 9, characterized in that the scaling of the side signal (xsme) is a division by a factor which is a monotonic continuous function of the energy contour of the main signal (xmono). A method according to claim 10, characterized in that the monotonic continuous function is a square root function. A method according to claim 10 or 11, characterized in that the main signal, xmm,, energy contour, EC, is calculated over a subframe according to: m + L EC (m) = í 2w (n) x, fm ,,,,, front start S m S frame end n = mL where L is an arbitrary factor, n is a summation index, m is the sample within the subframe and w (n) is a window creating function. A method according to claim 12, characterized in that the window-creating function is a rectangular window-creating function. A method according to claim 12, characterized in that the window creating function is a Hamming window function. 10 15 20 25 30 527 670 32 A method according to any one of claims 1 to 14, characterized in that the at least two coding chips (81) comprise different coding principles for the side signal (xside). A method according to claim 15, characterized in that at least a first coding scheme of the at least two coding schemes (81) comprises a first coding principle of the side signal (xside) for all subframes (90) and at least a second coding scheme of the at least two coding schemes (81). ) includes a second coding principle for the side signal (Xsme) for all subframes (90). A method according to claim 15 or 16, characterized in that at least one coding scheme of the at least two coding schemes (81) comprises the first coding principle of the side signal (xside) for one subframe and the second coding principle of the side signal (Xsme) for another subframe. A method according to claim 1, characterized in that the step of generating (214) the second output signal (pause) in turn comprises the steps of: analyzing spectral characteristics of a side signal (xside), which is a first linear combination of signals from at least the first and second channels (a, b; L, R); selecting a set of subframes (90) based on the analyzed spectral characteristics; and encoding the side signal (xside) within all subframes (90) in the selected set of subframes (90) separately. A method according to any one of claims 1 to 18, characterized in that the step generation (214) of a second output signal (psme) is applied in a limited frequency band. A method according to claim 19, characterized in that the step generation (214) of a second output signal (pside) is applied only for frequencies below 2 kHz. 10 15 20 25 30 527 670 33 A method according to claim 20, characterized in that the step of generating (214) a second output signal (pside) is applied only for frequencies below 1 kl-1z. A method according to any one of claims 1 to 21, characterized in that the polyphonic signals represent music signals. A method of decoding polyphonic signals, comprising the steps of: decoding (220) encoding parameters (pmono) representing a main signal; decoding (224) encoding parameters (psme) representing a side signal within an encoding frame (80); and combining (226) the at least the decoded main signal (x "m0n <>) and the decoded side signal (xkide) into signals for at least a first and a second channel (a, b; L, R), characterized by the step: providing ( 222) of at least two coding schemes (81), each of the at least two coding schemes (81) being characterized by a set of subframes (90) which together form the coding frame (80), the sum of the lengths of the subframes (90) in each coding scheme ( 81) is equal to the length of the coding frame (80), each set of subframes (90) comprising at least one subframe (90), the step of decoding (224) the coding parameters (psme) representing the side signal in turn comprising the step of decoding the coding parameters (pside) representing the side signal separately in the subframes (90) of one of the at least two coding schemes (81). An encoding apparatus (14), comprising: input means (16; 16A-C) for polyphonic signals (a, b; L, R, C) comprising at least a first and a second channel (a, b; L, R), 527 670 34 means (38) for generating a first output signal (pmono) which are coding parameters representing a main signal based on signals from at least the first and the second channel (a, b; L, R) ; and means (38) for generating a second output signal (pside) which are coding parameters representing a side signal based on signals from at least the first and second channels (a, b; L, R) within a coding frame (80); and output means (52); characteristics! means for providing at least two coding schemes (81), each of the at least two coding schemes (81) being characterized by a respective set of subframes (90) which together form the coding frame (80), the sum of the lengths of the subframes (90) in each coding scheme (81) is equal to the length of the coding frame (80); each set of subframes (90) comprises at least one subframe (90): the means (30) for generating the second output signal (pside) in turn comprises means (86, 87) for selecting a coding scheme (81) at least to a portion depending on the signal content of the present side signal (xsme); means for encoding the side signal (Xsme) in each of the subframes (90) of the selected encoding scheme separately. A decoder apparatus (24), comprising: input means (54) for coding parameters (pmom) representing a main signal and coding parameters (pside) representing a side signal; means (64) for decoding the coding parameters (pmono) representing the main signal; means (60) for decoding the encoding parameters (pslde) representing the side signal within an encoding frame (80); and means (68, 70) for combining at least the decoded main signal (x "m0n °) and the decoded side signal (xide) into signals for at least a first and a second channel (a, b; L, R); and output means (26; 26A-C), characterize! in that the means (60) for decoding the coding parameters (pi-ide) representing the side signal in turn comprises: means for providing at least two. coding schemes (81), each of the at least two coding schemes (81) being characterized by a respective set of subframes (90) which together form the coding frame (80), the sum of the lengths of the subframes (90) in each coding scheme being equal to the coding frame ( 80) length; each set of subframes (90) comprising at least one subframe (90); and means for decoding the coding parameters (pside) representing the side signal separately in the subframes (90) of one of the at least two coding schemes (81). A sound system (1) comprising at least one of: a coding apparatus (14) according to claim 24, and a decoding apparatus (24) according to claim 25.