EP1230827A2

EP1230827A2 - Method and device for processing a stereo audio signal

Info

Publication number: EP1230827A2
Application number: EP00985148A
Authority: EP
Inventors: Bodo Teichmann; Oliver Kunz; Jürgen HERRE; Klaus Peichl; Michael Beer
Original assignee: Fraunhofer Gesellschaft zur Forderung der Angewandten Forschung eV
Current assignee: Fraunhofer Gesellschaft zur Forderung der Angewandten Forschung eV
Priority date: 1999-12-08
Filing date: 2000-12-07
Publication date: 2002-08-14
Anticipated expiration: 2020-12-07
Also published as: JP2007316658A; US20030091194A1; WO2001043503A2; JP4579273B2; JP4000261B2; DE19959156C2; EP1230827B1; ATE251376T1; DE50003945D1; JP2003516555A; US7260225B2; DE19959156A1; WO2001043503A3

Abstract

In a device for processing a stereo audio signal having a first channel and a second channel the stereo signal is at first analyzed to obtain a measure for a quantity of bits required by a coder to code the stereo audio signal using a coding algorithm. The first channel and the second channel are then modified when the measure for the quantity of bits is larger than a predetermined value, the modification being performed in such a way that the energy of a sum signal of the first and the second modified channel is in a predetermined relation to the energy of a sum signal of the first and the second channel and that a difference signal of the first and the second modified channel is attenuated in contrast to the difference signal of the first and the second channel. Especially for audio coders requiring a constant output bit rate the side channel is attenuated in the case of stereo audio signals, the coding of which cannot meet the output bit rate of the coder, by which a stereo channel separation is abandoned for the benefit of an increased audio bandwidth or a reduction of quantizing disturbances, respectively.

Description

Method and device for processing a

Stereo audio signal

description

The present invention relates generally to the encoding of audio signals, and more particularly to processing of stereo signals.

A stereo signal comprises at least two channels, i.e. H. a left channel and a right channel. In addition, stereo signals can still have a left and right surround channel. There is also the possibility that a stereo signal has five different channels, i. H. a front left channel, a front center channel and a front right channel, and a left rear and a rear right channel.

For data-reducing coding of stereo signals, there is the possibility that similarities of at least two channels can be exploited in order to reduce the amount of bits that are required to code a stereo signal with at least two channels.

A known method for processing stereo signals in order to achieve more efficient coding is referred to as the center / side method (M / S method). In the M / S process, the first and second channels are combined to create a center channel and a side channel. For reasons of clarity, the following will no longer speak of a first and a second channel, but of a left channel (L channel) and a right channel (R channel). It is known that the center channel is equal to the sum of the left channel L and the right channel R multiplied by a factor of 0.5, while the side channel is the difference between the left channel L and the right channel R. , multiplied by a factor from Z. B. 0.5 (other factors are also possible). Expressed equally, this means:

M = 0.5 • (L + R) S = 0.5 • (- R)

If the left channel L and the right channel R are relatively similar to one another, M / S processing brings a significant saving in the amount of bits required for coding, since the side channel will have relatively less energy than R or L. In the borderline case, where the left channel L and the right channel R are identical, the center channel will be the same as the left channel L or the right channel R, while the side channel is 0. It can be seen that, due to the fact that the side channel is 0, a theoretically maximum bit rate saving of 50% is achieved when coding, since only the middle channel has to be coded, while no bits are used for the side channel Need to become.

There is thus a general rule that the more similar the right and left channels are, the smaller, i. H. the side channel will be less energy intensive, and the fewer bits will be needed to encode the side channel.

A listener will perceive the similarity of the left and right channels by perceiving, in the case of identical channels, a speaker or an orchestra exactly in the middle between the two speakers. On the other hand, a listener will perceive dissimilar channels in that he has a pronounced stereo effect, ie that a speaker, an orchestra or individual instruments of an orchestra can be located exactly to the left and / or right. Consider the case that the left channel has a lot of energy and that the right channel has little energy, that is, the case where z. B. only a single instrument is arranged on the far left in the recording room, and is only audible in the left channel, while on the right channel reads If there is noise, the center channel will be approximately the same as the left channel after M / S processing. In addition, the side channel will also be approximately the same as the left channel. In this case, both the center channel and the side channel have almost the same amount of energy and both have to be encoded with a relatively large number of bits. Compared to the original case, the amount of bits required for coding did not decrease due to the M / S coding in this signal constellation, but even doubled in the limit case if it is assumed that the left channel L comprises a certain amount of energy, while the right channel R is 0. Here it would have been much cheaper not to perform M / S processing, but only L / R processing. The effects on the number of bits required to encode a stereo signal thus range in one extreme case from a saving of 50% to the other extreme case, which results in a doubling of the bits required for encoding. When using an M / S process, care must therefore be taken to determine whether the piece is suitable for M / S processing or not. In the case in which a stereo signal (e.g. a test section of 20 ms, which is also referred to as a frame) is not suitable for M / S processing, M / S processing is omitted for reasons of bit efficiency, and Both the left and the right channel are coded individually. This "normal" case is also called L / R processing.

Common audio coding methods, such as those used to code audio signals that are decoded according to one of the MPEG standards, are generally divided into several steps. First, an audio signal, for example in the form of PCM samples, as z. B. outputs a CD player, converted into a spectral representation by means of a time-frequency transformation or a filter bank. Typically, a block with a certain number of samples, also called a "frame" is used to generate a block of complex spectral values that form a short-term spectrum of the frame of audio samples ("samples"). Block formation is achieved using transformation windows which are, for example, 1024 samples long. If, for example, overlapping windows are used for the transformation, the degree of overlap of which is 50%, 1024 spectral values are formed from 1024 samples. These spectral values are then quantized by means of a known iteration process, after which the quantized spectral values of an entropy coding z. B. using a plurality of fixed Huffmann code tables, in order to finally obtain a bit stream which on the one hand contains the coded quantized spectral values and on the other hand also has side information relating to the windows, to scale factors which are calculated during quantization, and refer to other information needed to decode the bit stream.

Mid / side processing can either be done before transforming to the spectral range, i. H. using the digital time discrete samples. Alternatively, mid / side processing can also be done after the transformation, i.e. H. with the complex spectral values. The latter alternative also has the advantage that center / side processing cannot be used for the entire spectrum, as in the time domain, but also for certain frequency bands when certain spectral values are subjected to center / side processing and others cannot ,

Audio coders are usually designed in such a way that they deliver a constant bit rate, ie a certain number of bits per second. Another constraint is that the quantization noise introduced by the quantization is, if possible, chosen such that its energy is below the psychoacoustic masking threshold or Monitoring threshold of the audio signal is. The basic method to adjust the quantization noise in the frequency domain is to "shape" the noise using the scale factors. For this purpose, as is known, the spectrum is divided into several groups of spectral coefficients called scale factor bands, to which a single scale factor is assigned. A scale factor represents a multiplication value that is used to change the amplitude of all spectral coefficients in this scale factor band. This mechanism is used to adjust the spectral quantization noise mapping generated by the quantizer so that in each scale factor band the quantization noise energy is below the psychoacoustic masking threshold in that scale factor band. It can be seen that neither quantization nor entropy coding are processes that favor a constant bit rate. It should be noted that - on the contrary - both methods favor a variable bit rate. For transmission applications, however, it is often required that the encoder have a constant bit rate at the output. In order to provide a constant bit rate, a so-called bit reservoir is usually used. If the audio signal is such that briefly fewer bits are required than specified by the external bit rate at the output of the encoder, bits are assigned to the bit reservoir in order to be able to give more bits in the case of an audio signal section which requires more bits for coding , which empties the bit reservoir again.

It should be noted that, as mentioned, one constraint of such an encoder is the constant output bit rate and the other constraint is that the quantization noise is less than or equal to the psychoacoustic masking threshold in order to be masked or masked by the audio signal.

In the following, possibilities are discussed, what about action is taken if the "inner bit rate" of the encoder deviates from the outer constant output bit rate. If the internal bit rate is so low that, for example, the bit reservoir is filled to its maximum value, there is of course no problem, since the quantizer can then be controlled in such a way that it now quantizes even more finely than necessary, which means that more bits are required for quantization. This is carried out until the "outer" constant bit rate is reached.

However, the case in which the "inner bit rate" of the encoder is higher than the constant bit rate required on the output side is more critical. This case will occur when the audio signal is difficult to encode, i. H. when the encoder has to spend many bits to encode the audio signal, which can also be descriptively referred to as the "high load" of the encoder. The transformation encoding is characterized by the fact that it can encode tonal pieces relatively efficiently, but that noisy signals which have relatively high energies and which also have a relatively complicated spectrum, such as speech or drum or drum music, are compressed relatively little can be. Also signals that are transient, i. H. which have an irregular time behavior can only be coded with relatively great effort if no coding artifacts are to be generated. In the case of transient signals, a switch is made from long windows to shorter windows during the windowing in order to achieve a better temporal resolution or to ensure that the quantization noise is “smeared” only over a smaller number of audio samples. In the case of short windows, significantly more page information is generated.

An encoder that determines that the output bit rate is not sufficient and that has already "run dry" the bit reservoir now has several options for "violently" reducing its internal bit rate to meet the criterion of the constant output bit rate. One way is to avoid switching to short windows. However, this leads to audible coding artifacts.

Another possibility is to deliberately violate the psychoacoustic masking threshold during quantization in order to quantize more coarsely than is actually necessary in order to achieve a lower bit rate. This also leads to audible interference.

Another possibility is to reduce the audio bandwidth, i. H. no longer to encode the full audio bandwidth, but to set the overlying spectral values to 0 from a certain cutoff frequency dependent on the output bit rate, in order to reduce the output bit rate. This method does not cause audible quantization disturbances, but leads to a loss of highs in the audio signal. However, this loss is often perceived less strongly than an audible quantization noise.

A particular problem with the coding of stereo signals is the effect called "stereo unmasking", which is briefly explained below. If normal L / R coding is used, both the left channel and the right channel are transformed, quantized and encoded for themselves, so that the quantization noise introduced in the left channel and right channel for data reduction is independent of the other channel. That is, the quantization noise in the left channel and the quantization noise in the right channel are not correlated. If the case is considered that the left and right channels are relatively similar, this means that after decoding a listener will perceive this signal so that, for example, a speaker is in the middle. The "stereo unmasking" effect now consists in the fact that due to the fact that the quantization noise in the two channels is not correlated, the quantization noise of the left channel on the left and the quantization noise of the right channel on the right is perceived. A high concealment of the rough However, schens only takes place in the middle, where the useful signal is, but not left and right.

In addition to its data rate reducing effect on special signals, M / S coding also has the advantage that the quantization noise in both the left and right channels is correlated with the quantization noise of the other channel, so that the quantization noise in the Center takes place and is covered there completely or substantially better by the useful signal than in the uncorrelated case. The situation is different in which the left and right channels are relatively dissimilar. If M / S coding is used here, the useful signal will be either left or right due to the stereo effect, while due to the M / S coding the quantization noise is correlated and lies more in the middle. Stereo unmasking also takes place here, so to speak.

Recently, more and more scalable audio encoders have been investigated. Scalable audio coders are arranged such that their bit stream on the output side has at least a first and a second scaling layer. A decoder, which is of simple design, will only take the first scaling layer from the scaled bit stream, which for example has an encoded audio signal with reduced bandwidth or is an audio signal encoded with a simple encoding algorithm. Another decoder, fully designed, will take both the first scaling layer and the second scaling layer from the bitstream to decode the first scaling layer with a first decoder, and then to decode the second scaling layer either alone or together with the decoded first scaling layer provides a full bandwidth audio signal.

Scalable encoders are particularly desirable in the field of stereo signals, since here a first scaling layer Mono signal, ie the middle channel, can be used, while z. B. the side channel can be taken. A simple decoder or a decoder, which is designed for fast operation, will only deliver the mono signal, while a better decoder or a decoder, in which the speed of the transmission is not the most decisive criterion, in addition to the mono or The middle layer will also take the side layer to produce a full stereo signal at the decoder output.

There are various options for the construction of the scaling layers. The first scaling layer can differ from the second scaling layer or from any number of further scaling layers in the audio coding method itself, in the audio bandwidth, in the audio quality, with respect to mono / stereo and or a combination of the quality criteria mentioned or other conceivable criteria. For high coding efficiency, the aim is for the second scaling layer to have the smallest possible number of bits, or for a decoder which decodes the second scaling layer to also use the first scaling layer as extensively as possible. If a scalable encoder for stereo signals is considered, which as the first scaling layer the center signal, i. H. provides the mono signal, and which supplies the side channel as a second layer, it can be seen that the more often the M / S coding is used, the better its overall efficiency. However, this requirement conflicts with bit efficiency for certain stereo signals, namely for stereo signals that have a high stereo channel separation. On the other hand, M / S processing provides a certain "natural" scalability and leads to a correlation of the quantization noise in the left channel and in the right channel.

The problems mentioned with regard to M / S coding apply all the more the more an audio signal to be coded suddenly changes its properties with regard to M / S coding. If an audio signal to be encoded suddenly no longer has the property that the left channel is similar to the right one, the M / S coding gain is eliminated. One consequence will therefore usually be an increase in the quantization disturbance possibly beyond the psychoacoustic hearing threshold and / or a reduction in the audio bandwidth depending on the specific implementation of the encoder.

This problem is particularly noticeable, but not only, in scalable audio coding, and particularly where the so-called mono-stereo scalability is used, as was explained above.

The object of the present invention is to provide an apparatus and a method for processing a stereo audio signal which leads to less audible interference.

This object is achieved by a device for processing a stereo audio signal according to claim 1 and by a method for processing a stereo audio signal according to claim 18.

The present invention is based on the finding that it is often cheaper for stereo audio signals to dispense with high stereo channel separation in order to achieve a higher audio bandwidth and / or less audible interference in comparison with the case in which the stereo channel separation is maintained, while the audio bandwidth is reduced, or noise introduced by quantization becomes audible.

Experience has shown that a listener will find audible quantization disturbances to be more uncomfortable than less stereo channel separation. Audible quantization disturbances are generally a foreign body in an audio signal, while a listener of a stereo signal processed according to the invention does not necessarily know how the stereo channel separation of the Output signal was and thus a lower stereo channel separation will not be perceived as coding artifact.

A reduction in the stereo channel separation is thus used in order to generally reduce the bit rate of the encoder on the output side or to reduce it to a predetermined value.

An apparatus according to the invention for processing a stereo signal having a first channel and a second channel comprises means for analyzing the stereo audio signal in order to obtain a measure of the amount of bits required by an encoder to use the stereo audio signal Encoding algorithm and means for modifying the first and second channels to obtain modified first and modified second channels, the means for modifying being responsive to the means for analyzing to be effective if the measure of the Amount of bits exceeds a predetermined amount, and wherein the means for modifying is configured such that a sum signal from the first and second modified channels is substantially equal to the characteristic at least according to a characteristic of the signal that changes similar to the energy of the signal a sum signal s is from the first and second channels, and that a difference signal from the first and second modified channels is attenuated compared to the difference signal from the first and second channels.

It should be noted that the characteristic, which is similar to energy, can be the energy itself, but also e.g. B. is the sum of squared samples in a certain time period, the sum of squared spectral values in a certain frequency range, the sum of sample amounts in a certain time period or the sum of squared spectral values in a certain frequency range or a combination between two or more of the above characteristics. For the sake of simplicity, however, the following is an example of energy as a characteristic that is similar to energy.

Modifying the stereo audio signal, i.e. H. reducing the channel separation is carried out provided that the volume of the signal does not fluctuate. A reduced channel separation itself will not lead to annoying artifacts in the decoded signal, but a fluctuation in the volume will. Therefore, the first and second channels, e.g. B. the left channel and the right channel, modified so that the volume, d. H. the sum signal, compared to the unmodified first and second channels, remains essentially the same, at least in terms of energy and preferably even in terms of signal, while the difference signal is damped.

The preprocessing of the stereo signal according to the invention will always start when it is determined that the amount of bits required to encode the stereo audio signal becomes too high. The measure of the amount of bits needed to encode the stereo audio signal can be derived from the stereo audio signal by analyzing it in various ways.

First, the center and side channels of the stereo audio signal can be viewed to determine how many bits are needed based on an energy ratio or a difference in the logarithms of the energies thereof. Without having to determine the exact number of bits, it can be concluded that in the case of a small energy ratio between the center and side channels, ie in the case of channels of approximately the same size, a large number of bits will be necessary. The lower the energy ratio between the center and the side channel, the more attenuation of the side channel will be necessary to achieve a certain output bit rate. A little energy The relationship between the center and side channels is when the original audio signal has a high stereo channel separation, for example when the left channel has a lot of energy while the right channel has essentially noise. However, there is also a small energy ratio if the language of one speaker is in the left channel and if the language of another speaker is in the right channel, which means that the left channel and the right channel may have the same amount of energy that however, both channels are uh-correlated. In this case too, there is a high stereo signal separation, and the center channel and the side channel will have a relatively small difference in the logarithms of the energy.

However, another way of determining the measure for a set of bits, regardless of the nature of the center channel and the side channel, is to look at the encoder itself. A measure of the number of bits required by an encoder is the so-called perceptual entropy (PE), which is equal to the energy ratio between the useful audio signal and the psychoacoustic monitoring threshold calculated for the useful audio signal. If the PE is large, it can be concluded that the audio signal has a relatively low masking ability. However, if the PE is small, ie if the energy of the useful signal is only slightly above the psychoacoustic listening threshold, the useful signal only has to be quantized relatively roughly, and the quantization noise is still "hidden" below the psychoacoustic listening threshold. If it is found that the sum of the PE of the left channel, preferably averaged over a certain time, and the PE, also preferably averaged over a certain time, for the right channel is above a predetermined value, the side channel is damped according to the invention to reduce the number of bits required. This alternative aspect of the present invention is therefore not concerned with the individual appearance of the center and side channels, but with the stereo audio signal itself, which is not related to its M / S-Co- ability, but its general audio coding ability, ie the difficulty in coding the same in order to achieve a certain target bit rate.

A generalization of the second aspect is to use some other size as a measure of the amount of bits that indicates the "load" of the encoder. Such a size can also be, for example, a signal which, on the basis of transient properties of the audio signal, indicates that an audio encoder must use short windows for windows, since it is a fact that short windows require a higher bit rate, not least because of the increased number of page information. Thus, for the purposes of the present invention, the full range of control quantities of an audio encoder can be used to find a measure of how much the side channel must be attenuated in order to reduce the output bit rate of the encoder.

Preferred embodiments of the present invention perform an increasing or decreasing attenuation of the side channel in order to prevent a listener from perceiving the decreasing stereo channel separation immediately, but rather that the reduction in the stereo channel separation gradually occurs or the enlargement of the stereo channel separation gradually and increases to disguise the manipulation of the stereo audio signal as well as possible.

It should be noted that in order to maintain a non-fluctuating volume due to the modification, the sum signal of the modified left and right channels need not necessarily be identical to the sum signal of the unmodified left and right channels, but that it is sufficient that only the energies of the two sum signals are substantially the same or are in a predetermined relationship to one another. A listener does not know what the volume of the unmodified stereo audio signal was and will therefore not perceive it as a disturbance if the volume has been changed in the direction of higher volume or lower volume by the preprocessing. However, because of the simplicity of implementation, it is preferred that this ratio be 1.

Preferred embodiments of the present invention are explained in detail below with reference to the accompanying drawings. Show it:

1 shows a basic block diagram of the device according to the invention for processing a stereo audio signal;

2 shows a more detailed illustration of a preferred embodiment of the device for modifying; and

3 shows a block diagram of a device according to the invention as a preprocessing stage for a scalable encoder with mono / stereo scalability.

1 shows a block diagram of the device according to the invention for processing a stereo audio signal, which is fed into the device at an input 10 and has a first channel L and a second channel R. The stereo audio signal in the form of the first channel L and the second channel R is fed on the one hand into a device 12 for analyzing the stereo audio signal, and on the other hand is also fed into a device 14 for modifying the first and second channels in order to produce a modified first at an output 16 To obtain channel L 'and a modified second channel R'. In general, the modified first channel L 'and the modified second channel R' at the output 16 will differ from the unmodified first channel L and from the unmodified second channel R at the input 10 in that the modified stereo audio signal present at the output 16 is less Channel separation as will have the unmodified stereo audio signal at input 10.

The means 12 for analyzing the stereo audio signal determines a measure of an amount of bits which is required by an encoder (not shown in FIG. 1) in order to encode the stereo audio signal using an encoding algorithm specified by the encoder. The measure for the bit quantity is supplied by the device 12 for analysis via a signal path 18 to the device 14 for modification. If the measure of the bit quantity supplied via the signal path 18 exceeds a predetermined measure, the means 14 for modifying becomes effective in order to modify the first channel L and the second channel R. According to the invention, the modification of the first and second channels is carried out in such a way that the energy of the sum of the modified stereo audio signal at the output 16 is in a predetermined ratio and preferably substantially equal to the energy of the unmodified stereo audio signal at the input 10, but the difference signal, that apart from the factor of e.g. B. 0.5 corresponds to the side channel, is attenuated in the modified stereo audio signal at the output 16 compared to the unmodified stereo audio signal at the input 10.

1 shows two possibilities for feeding the device 12 for analysis, which can be used individually or in combination. The first possibility is represented by a left arrow 15a, which in a sense represents a forward coupling, ie the device for analyzing the stereo audio signal is fed with the unmodified signal L, R. The other possibility is to feed the device 12 for analysis with the modified signal L ', R'. Especially in cases where the attenuation of the side signal is slow, it is irrelevant whether the attenuation depends on the current unmodified signal or on one of the last processing blocks of the modified signal is controlled to a certain extent in terms of feedback. It is therefore irrelevant whether the stereo audio signal itself is analyzed directly or indirectly using a previously modified signal.

Various configurations of the device 12 for analyzing the unmodified stereo audio signal at the input 10 are discussed below. One possibility is that the means 12 for analyzing form both the center and the side channel of the stereo audio signal and then consider the ratio of the energies of the center and the side channel. The energy ratio between the center and the side channel is preferably averaged over a certain time, which can be, for example, in the order of 10 audio frames, which corresponds to a value of 200 ms if an MPEG-2-AAC is used as the audio encoder. Encoder is used, which can have a frame length of about 20 ms. With regard to the MPEG-2 AAC encoder, reference is made to the ISO / IEC 13818-7 standard, in which the individual function blocks of an audio encoder and an audio decoder and their interaction are described in detail.

If it is found that the energy ratio or the difference in the logarithms is smaller than a certain value to be determined empirically depending on the application, which can be chosen, for example, to 6 dB, the means 14 for modifying is activated in order to attenuate the side channel achieve, as will be explained in more detail with reference to FIG. 2. According to the first aspect of the present invention set out above, the device 12 for analyzing the stereo audio signal thus operates on the basis of a direct examination of the MS coding capability of the stereo audio signal. In an implementation of this first aspect of the present invention, the inventive device for processing the stereo audio signal will only attenuate the side channel if the signal is no longer as good as MS coding. ability because, for example, both channels are either different in terms of energy and / or signal. According to this aspect, stereo channel separation is therefore reduced whenever maintaining the original stereo channel separation would lead to a too high output bit rate and if the stereo channel separation was high at all.

According to a further aspect of the present invention, the attenuation of the side channel is used to reduce the output-side encoder bit rate, regardless of whether the stereo audio signal has a certain MS coding capability or not. This second aspect according to the invention assumes that even in the case of a small stereo channel separation, further attenuation of the side channel can still be achieved in order not to exceed a predetermined output bit rate of the audio encoder. For this purpose, the number of bits required to encode the audio signal is estimated regardless of the MS coding capability of the audio signal.

As is known in the art, modern audio encoders, and for example also an MPEG-2-AAC audio encoder, use a psychoacoustic model which is used to calculate the frequency-dependent psychoacoustic masking threshold from an audio signal to be encoded. Roughly speaking, the psychoacoustic model provides an energy value for each scale factor band as a psychoacoustic masking threshold. If the quantization noise introduced by the quantizer is below the energy value or if the noise introduced by the quantization disturbances is equal to the energy value, then the introduced quantization noise will be essentially inaudible according to the psychoacoustic theory.

The energy ratio or the difference between the logarithms of the audio signal itself and its psychoacoustic masking threshold, which is also referred to as perceptual entropy (PE). is thus provides a measure of how many bits are required to encode the audio signal. If the PE is high, many bits are required because the masking ability of the audio signal is relatively poor and must therefore be finely quantized. On the other hand, if the PE is small, relatively few bits are required, since the audio signal is masked relatively well, and therefore only a relatively rough quantization is required.

According to a preferred embodiment, in the second aspect of the present invention the measure for the amount of bits is determined as follows. The PE values for the individual scale factor bands are integrated over the frequency, i. H. summed up. This is done for both the left and right channels. The PE sum for the left channel is then summed up to the PE sum for the right channel. This sum PE value from the left and right channel represents the bit requirement for a frame. This sum channel PE value is then preferably over a certain number of frames, such as. B. 10, averaged to obtain an average PE value for the stereo audio signal. If this averaged PE value is greater than or equal to a typically empirically determined predetermined value, the means for multiplying is activated to dampen the side channel.

In general, any other controlled variable can therefore be used as a measure of the amount of bits that an encoder will need, which is a measure of the "load" of the encoder, such as e.g. B. a control signal of the encoder, which signals the use of short windows when windows. Windows with short windows per se lead to a higher number of bits, since shorter windows cannot be coded as bit-saving as longer windows.

With regard to the damping amount of the side channel, there are several options which differ here in terms of their effort. The easiest way is in agreeing a predetermined damping value as the target value, which can be determined empirically, for example. Another possibility, however, is to adaptively determine the damping value, ie to dampen the side channel by a predetermined increment amount, and then to see again whether the number of bits has already decreased sufficiently or not. A new iteration loop with a further increment damping amount can then be entered in order to again determine whether the number of bits is already sufficiently small. This process can be repeated until the number of bits required by the encoder lies in a target corridor. However, it can be seen that the computing time and implementation effort in the case of adaptive damping adaptation is considerably higher than in the case of a predetermined damping. On the other hand, adaptive damping adjustment delivers the best and most accurate results.

2, in which a detailed illustration of the device 14 for modifying according to a preferred exemplary embodiment of the present invention is shown. The means 14 for modifying can be understood to have a first input 20a for the first channel L and a second input 20b for the second channel R. The device 14 comprises a first multiplier 22a for multiplying the first channel L by a certain factor x, a second multiplier 22b for multiplying the first channel L by a factor y, a third multiplier for multiplying the second channel R by the factor x and finally a fourth multiplier 22d for multiplying the second channel R by the factor y. Furthermore, the means 14 for modifying comprises a first summer 24a for summing the output signal of the first multiplier 22a with the output signal of the fourth multiplier 22d, and a second summer 24b for summing the output signal of the second multiplier 22b with the output signal of the third multiplier 22c. At the exit 26a of the Finally, the modified first channel L 'is applied to the first summer 24a, and the modified second channel R' is applied to the output 26b of the second summer 24b.

In the following, the determination of the two multiplication factors x, y is dealt with in order to achieve a damped side channel, while the center channel at the output 26a, 26b is equal to the center channel at the input 20a, 20b of the device 14, which is shown in FIG 2 is shown. The following matrix applies to the signal processing carried out by the modification device 14:

L '= xL + yR (1)

R '= yL + xR (2)

The task now is to determine x and y, so that:

L '+ R' = L + R = M = M ', (3)

and that applies:

L'- R '= S' = damping * S = damping * (L-R) (4)

The result is:

M = 0.5 (x + y) (L + R) (5)

Since M should not be modified by the processing, the following equation also applies:

x + y = 1 (6)

For the side channel:

S = 0.5 (x - y) (L - R) (7) From equation (7) it follows that S is reduced by the factor xy or, expressed logarithmically, is damped by 10 -log10 (xy) dB = att, att stands for the damping, and it holds that att is less than Is 0 dB.

For attenuation in dB steps:

att (in dB) = 20 * logl0 (x-y) (8)

The following expression then results from equation (8):

exp (0.05 att) = x - y (9)

Equation (6) and equation (9) result in equation (10) for x and equation (11) for y.

x = 0.5 * (1 + exp (0.05 att)) (10) y = 0.5 * (1 - exp (0.05 att)) (11)

The attenuation "att" (in dB) is determined depending on one of the control variables described. Equations (9) and (10) thus result in the factors x and y for the damping matrix represented by FIG. 2, which is reflected in equations in Equations (1) and (2). In order to save implementation effort and computational effort, a completely adaptive adaptation of the attenuation att does not have to be carried out, but a certain attenuation value att, which has been determined empirically, can be used if the measure for the amount of bits exceeds a predetermined limit value.

According to the invention, the attenuation is not increased suddenly, since a reduction in the channel separation that takes place suddenly could possibly lead to an audible disturbance or to the listener being amazed, for example if a speaker was initially placed on the left and is suddenly noticed in the middle. Therefore, in the event that it is determined that the side channel is to be damped, a gradual attenuation of the side channel, for example using a predetermined increment value, is undertaken such that, vividly speaking, the news anchor slowly "wanders" from the left side to the center. If, in the opposite case, it is found that the measure for the amount of bits is again smaller than the predetermined value, the attenuation is not abruptly canceled, but slowly returned to 0, such that, in order to remain in the example, the speaker slowly moves away from the Center to the side "wanders". This gradual damping or gradual cancellation of the damping should take place as slowly as possible so that the damping of the side channel is practically not noticed. However, the attenuation must be reduced so quickly that, due to the high bit rate at the output, the encoder does not begin to violate the psychoacoustic masking threshold or to remove audio bandwidth. According to the invention, this bit reservoir is therefore used in encoders which have a bit reservoir mechanism in order to slowly increase the attenuation until the target value is reached at which the attenuation is so high that the predetermined bit rate at the output of the encoder can be maintained. If the damping is then released again, the bit reservoir can be emptied again.

In the implementation shown in FIG. 2, a boundary condition for determining x and y was such that the sum signal, which corresponds to the center channel up to a factor of 0.5, was not changed. However, signals are conceivable in which the left and right channels are similar but have a phase shift in the range of 180 degrees to one another. It should be noted that such signals are not particularly common since they cannot be presented well with mono playback devices. Nevertheless, such signals are conceivable. In this case the center channel M would be small and the side channel would be large. If S were to be attenuated so much that S becomes smaller than M, the overall volume would also be greatly affected. In contrast to however, a reduction in stereo channel separation is intolerable for a listener if the volume fluctuates greatly, regardless of the audio signal itself. A listener will find such a disturbance annoying.

To avoid this problem, it is preferred to additionally determine in the analyzing device 12 whether the phase shift of L and R is close to 180 degrees. If this is determined, the sign of R can simply be reversed. The originally desired spatial stereo effect is then lost, but the effect of the reduced volume is avoided, which will disturb a listener less.

As an alternative to the sign reversal, the M-channel could also be amplified to a certain value in the device for modifying or in a downstream encoder stage such that the energy of the modified M-channel is in a predetermined ratio to the energy of the M-channel of the unmodified stereo audio signal. A value of 1 is preferred for the energy ratio, although a certain amplification or attenuation can also be carried out by the modifier device, but the ratio to the unmodified stereo audio signal should always be essentially maintained so that a listener does not experience any significant volume fluctuations due to the preprocessing will perceive. Of course, small volume fluctuations are not as problematic and sometimes even imperceptible. Large volume fluctuations will be annoying for a test listener.

At this point, it should be pointed out that it is immaterial whether time-discrete sample values are present at input 10 of the device according to the invention for processing a stereo audio signal, or whether spectral values are present. All operations for analyzing the stereo audio signal can be carried out both with discrete-time samples and with spectral values. In addition, all operational tion in the device for modifying both discrete-time samples and spectral values. The device according to the invention for processing a stereo audio signal could thus also be arranged after the time-frequency transformation stage of a time / frequency transformation-based encoder, such as, for. B. an MPEG audio encoder. This concept even gives the additional possibility that the stereo preprocessing can be carried out in a frequency-selective manner, that is to say, for example, that a different attenuation of the signal S can be carried out depending on the frequency. This is particularly useful since the ability to locate the human ear is not equally sensitive to all frequencies. If the processing according to the invention is thus carried out spectrally, the spectral values of the side channel can be attenuated the more the less the human ear hears directionally in this frequency range, while spectral values that are in the frequency ranges in which the human hearing provides good directional tracking.

It should be noted that in modern audio encoders it is anyway determined in frequency using the so-called M / S mask where M / S coding is to be carried out and where L / R coding is better. In this case, the processing according to the invention would only be applied to the frequency ranges in which MS coding is present, i. H. in which the MS mask is set. Alternatively, the MS mask could also be set in more bands, i. H. MS coding can be carried out, the side channel being attenuated in these additional MS bands compared to the known method in order to comply with bit rate requirements.

In the following, reference is made to FIG. 3, in which a device for processing a stereo audio signal is shown which, in addition to the radio signals shown in FIG. tion blocks also includes an MS encoder 30 and a scalable encoder 32 which outputs a scaled bit stream BS on the output side. The MS encoder 30, as is known in the art, includes a summer 30a for summing the modified left channel L 'and the modified right channel R' to, after multiplication by a multiplier 30b which has a factor of e.g. B. 0.5 is assigned to generate the multiplied center channel. In addition, the MS encoder 30 comprises a subtractor 30c and a further multiplier 30d in order to generate the modified side channel S ′, which is attenuated with respect to a side signal which is formed from the unmodified stereo audio signal at the input 10. The center channel M 'and the side channel S' are both fed into the scalable encoder 32, which preferably has a mono-stereo scalability. The first scaling layer will represent the mono signal M 'and the second scaling layer will comprise the modified side channel S'. Other scaling options, such as B. that the modified or unmodified mono channel M 'is additionally band-limited, and that the upper mono band is also included in the second scaling layer in addition to the modified side channel.

The scalability effect in the mono-stereo encoder 32 is particularly favorable if an LR coding is used instead of an MS coding. The stereo signal processing according to the invention by the devices 12 and 14 is therefore particularly advantageous particularly in connection with the scalable encoder 32. In order to obtain mono-stereo scalability, MS coding can also be used if it is actually no longer preferable to LR coding. This is achieved in that the side channel at the input of the scalable encoder 32 is damped compared to the unmodified case.

3 is a dashed signal path 36 from scalable encoder 32 for device 12 for analysis. This dashed signal path 36 is intended to symbolize that certain measures to derive a measure of the amount of bits that the scalable encoder will need to encode the stereo audio signal at the input 10 do not have to be calculated directly in the device 12, but from the device scalable encoder can be output into the device 12, such as. B. Perceptual Entropy PE, the reference to the use of short windows, etc. This means that these function blocks do not have to be present both in the device 12 for analysis and in the scalable encoder 32, but that their implementation is only sufficient in the scalable encoder 32 ,

In this case, the means for modifying 14 in order to determine the measure 18 for the bit quantity would initially not carry out any modification. The device shown in FIG. 3 would then be, so to speak, in a "pre-run mode" where no bit stream is written, but where only the required degree of attenuation for the side channel is determined. In the then subsequent coding mode, in which the bit stream BS is then written by the scalable encoder, the means 14 for modifying will work with correspondingly defined factors x, y.

If the device shown in FIG. 3 is operated with spectral values for the first channel L and the second channel R, and if the scalable encoder is a time / frequency transformation encoder, then the stage of the scalable encoder 32, which is the time frequency -Transformation performs, upstream of the input 10. The devices 12, 14 and 30 would then be embedded in the scalable encoder 32.

The signal paths 36a, 36b illustrate that the modified channels can also be routed to the scalable encoder without M / S coding, so that the latter can then determine whether M / S or L / R coding is cheaper.

Claims

claims

1. Device for processing a stereo audio signal, which has a first channel (L) and a second channel (R), with the following features:

means (12) for analyzing the stereo audio signal or a signal derived from the stereo audio signal to obtain a measure of an amount of bits required by an encoder (32) to encode the stereo audio signal using an encoding algorithm; and

means (14) for modifying the first and second channels (L, R) to obtain a modified first and a modified second channel (L ', R'),

said means (14) for modifying being responsive to said means (12) for analyzing to take effect when the measure (18) of the amount of bits exceeds a predetermined measure, and

the means (12) for modifying being designed such that a characteristic of a sum signal from the first and second modified channels (L ', R'), which is similar to the energy of the sum signal, is in a predetermined ratio to the characteristic of a sum signal is from the first and the second channel (L, R), and that a difference signal from the first and the second modified channel (L ', R') is damped compared to a difference signal from the first and the second channel (L, R) ,

2. Apparatus according to claim 1, wherein the device

(14) has the following features for analysis:

a device for determining the characteristic of the Sum of the first and second channels over a predetermined period of time;

means for determining the characteristic of the difference from the first and second channels over a predetermined period of time; and

means for forming the ratio of the characteristic of the sum of the first and second channels and the characteristic of the difference of the first and second channels, the ratio of the characteristics being the measure (18) for the amount of bits.

i. Device according to Claim 1, in which the device (12) for analyzing has the following features:

first means for determining a first characteristic ratio between the first channel and the psychoacoustic masking threshold of the first channel over a predetermined time;

second means for determining a second characteristic ratio between the second channel and the psychoacoustic masking threshold of the second channel over a predetermined time; and

means for summing the first and second characteristic ratios, the sum of the first and second characteristic ratios indicating the measure (18) for the amount of bits.

I. Apparatus according to claim 1, in which the encoder (32) is arranged to use long or short windows for converting a temporal stereo audio signal into a spectral stereo audio signal in response to the temporal structure of the stereo audio signal, and in which the means (12) for analyzing is arranged to detect whether short or long windows in the Encoder (32) are used, the measure of the amount of bits is that short windows are used.

Apparatus according to any one of the preceding claims, in which the means (14) are arranged to be operative so that the difference signal from the first and second channels is gradually attenuated from no attenuation to a certain attenuation, and so on to be effective that the attenuation is gradually reduced from the particular attenuation to no attenuation.

Apparatus according to claim 5, wherein the speed of attenuation is selected as slow as possible, but still so fast that a bit reserve mechanism of the encoder (32) is used so that the encoder (32) neither reduces the audio bandwidth nor a psychoacoustic masking threshold at one Quantization violated.

Device according to one of the preceding claims, in which the device (14) is arranged for modification in order to adaptively attenuate the difference signal depending on the determined measure.

Apparatus according to claim 2, wherein the means (14) for modifying is arranged to attenuate the difference signal depending on a characteristic ratio generated by the means for forming the characteristic ratio, so that the attenuation of the difference signal is high when that Characteristic ratio is small, and that the attenuation of the difference signal is low when the characteristic ratio is high.

Apparatus according to Claim 7 or Claim 8, in which the means (14) for modifying are designed such that they adaptively adapt the difference signal dampens that the characteristic ratio of the difference signal to the sum signal is substantially equal to a predetermined value.

10. Device according to one of the preceding claims, in which the device (14) for modifying has the following features:

a first multiplier (22a) for multiplying the first channel (L) by a first factor (x);

a second multiplier (22b) for multiplying the first channel (L) by a second factor (y);

a third multiplier (22c) for multiplying the second channel by the first factor (x);

a fourth multiplier (22d) for multiplying the second channel (R ') by the second factor (y);

a first summer (24a) for summing the output of the first multiplier (22a) and the output of the fourth multiplier (22d) to produce the modified first channel (L '); and

a second summer (24b) for summing the output of the third multiplier (22c) and the output of the second multiplier (22b) to produce the modified second channel (R ');

the first and second factors (x, y) being chosen such that the sum signal of the first and second channels and the sum signal of the modified first and second channels are substantially the same, and the difference signal is attenuated by a certain factor.

11. Device according to one of the preceding claims, in which the device (12) for analyzing further has the following feature:

means for determining whether a phase angle between the first and second channels (L, R) has a value in the vicinity of 180 °; and

the modification (18) device further comprising:

means for reversing the sign of a channel (L, R) if the phase angle is near 180 °.

12. Device according to one of the preceding claims, in which the first and the second channel (L, R) of the stereo signal are given by spectral values which have been generated from a temporal stereo signal by conversion into the spectral range, the device for modifying (14 ) is arranged to perform a frequency-selective attenuation of the difference signal.

13. The apparatus of claim 12, wherein the means for modifying is arranged to attenuate more in a frequency range in which the directional order of the human ear is reduced than in a frequency range in which the directional location of the human ear is not reduced.

14. Device according to one of the preceding claims, further comprising the following features:

center / side means (30) for creating a center channel (M ') equal to half the sum of the modified left (L') and modified right channel (R '), side means (30) for generating a side channel equal to half the difference of the modified first channel (L ') and the modified second channel (R'); and

a scalable encoder (32) arranged to encode and write to the center channel (M ') and write to a bit stream (BS) as the first scaling layer, and further arranged to encode the side channel (S') to encode and write to the bitstream (BS) as a second scaling layer.

15. The apparatus of claim 14, wherein the scalable encoder (32) is arranged to use a bit reservoir device in the event that the measure of the amount of bits exceeds a predetermined value, so that the audio bandwidth is not reduced and / or the psychoacoustic masking threshold is violated.

16. The method according to any one of the preceding claims, wherein the characteristic which is similar to energy, the energy itself, the sum of squared samples in a certain period of time, the sum of squared spectral values in a certain frequency range, the sum of sample amounts in one certain period of time and / or the sum of squared spectral values in a certain frequency range.

17. The method according to any one of the preceding claims, in which the stereo audio signal is processed in blocks, and in which the signal used in the analysis derived from the stereo audio signal is the modified signal of a preceding processing block.

18. A method of processing a stereo audio signal having a first channel (L) and a second channel (R), comprising the following steps: Analyzing (12) the stereo audio signal or a signal derived from the stereo audio signal to obtain a measure of an amount of bits required by an encoding algorithm to encode the stereo audio signal; and

Modifying (14) the first and second channels (L, R) to obtain a modified first and a modified second channel (L ', R') if a measure (18) determines the amount of bits in the analyzing step which exceeds a predetermined amount, the modification being carried out in such a way that a characteristic of a sum signal from the first and second modified channels (L ', R'), which is similar to the energy of the sum signal, in a predetermined ratio to one Characteristic of a sum signal from the first and the second channel (L, R), and that a difference signal from the first and the second modified channel (L ', R') compared to a difference signal from the first and the second channel (L, R ) is damped.