EP1145227B1 - Method and device for error concealment in an encoded audio-signal and method and device for decoding an encoded audio signal - Google Patents

Abstract

In a method for concealing an error in an encoded audio signal a set of spectral coefficients is subdivided into at least two sub-bands ( 14 ), whereupon the sub-bands are subjected to a re-verse transform ( 16 ). A specific prediction is performed ( 18 ) for each quasi time signal of a sub-band to obtain an estimated temporal representation for a sub-band of a set of spectral coefficients following the current set. A forward transform ( 20 ) of the time signal of each sub-band provides estimated spectral coefficients which can be used ( 28 ) instead of erroneous spectral coefficients of a following set of spectral coefficients, e.g. in order to conceal transmission errors. Transforming at the sub-band level provides independence from transform characteristics such as block length, window type and MDCT algorithm while at the same time preserving spectral processing for error concealment. Thus the spectral characteristics of audio signals can also be taken into account during error concealment.

Description

The present invention relates to coding or Decoding audio signals and in particular on obfuscation of errors ("Error Concealment") in digital encoded audio signals.

With the increasing spread of modern audio encoders and corresponding Audio decoder that complies with one of the MPEG standards work, has the transmission of coded audio signals via radio networks or via wired networks, such as z. B. the Internet has already achieved great importance. When transmitting coded audio signals using digital broadcasting but also when transmitting audio signals A non-ideal one lies over wired networks Transmission channel before that can lead to coded Audio signals are disturbed during transmission. Therefore the task on the decoder side arises, as with transmission errors to be dealt with or how transmission errors should be "veiled". The error concealment is used to prevent transmission errors in any way and Way to manipulate the subjective auditory impression of a to improve such erroneous decoded audio signals.

Several error concealment methods are already known. The simplest way to conceal errors is by Muting method, also known as "muting" becomes. A decoder detects that data is missing or incorrect the playback is switched off. The Missing data are thus replaced by a zero signal. This avoids that due to a transmission error too loud or unpleasant noises from a decoder become. This is due to psychoacoustic effects sudden drop and rise in signal energy when the decoder outputs error-free data again, but as felt uncomfortable.

Another well-known procedure that involves the sudden drop and avoids re-increase in signal energy is the procedure data repetition. For example, a block falls or more blocks of audio data, part of the last sent data repeated in a loop until error-free again, d. H. sound data is intact. This However, the process leads to disruptive artifacts. Become Repeats only short parts of the audio signal, that's what it sounds like repeated signal independent of the original signal with a fundamental frequency at the repetition frequency. Become longer parts repeated, certain echo effects arise, which are also perceived as disturbing.

For block-oriented transformation encoders / decoders, where a spectral representation of a temporal audio signal would also be possible a spectral-value prediction in the case of incorrect ones Perform audio data. It is found that Spectral values in a block are incorrect, so can these spectral values based on the spectral values of a previous blocks or several previous blocks predicted, d. H. be predicted or estimated. The predicted spectral values correspond within certain limits the incorrect spectral values when the audio signal is relative is stationary, d. H. if the audio signal is not like that undergoes rapid changes in the signal envelope. If, for example, a according to the MPEG-AAC standard (ISO / IEC 13818-7 MPEG-2 Advanced Audio Coding) working method is considered to have a normal block of coded Audio data 1024 spectral values. In the method of spectral value Prediction will therefore be 1024 in parallel working predictors in the decoder needed to e.g. B. in In the event of a complete block failure ("frame loss") to be able to predict all spectral values.

A disadvantage of this method is the relatively high computing effort, which is currently receiving real-time decoding of a Multimedia or audio data signal impossible.

Another major disadvantage of this method is through the transformation algorithm used, the modified discrete cosine transformation (MDCT), conditional. It it is well known that the MDCT algorithm is not an ideal one Fourier spectrum delivers, but rather a "spectrum" that extends different from an ideal Fourier spectrum. investigations have shown that z. B. a sine time function, which has a Fourier spectrum that has a single spectral line at the frequency of the sine function has an MDCT "spectrum" has that at the frequency of the sine function has a dominant spectral coefficient, however additional spectral coefficients at other frequency values having. In addition, the height of an MDCT "spectrum" a sine function not from block to block same, but the same varies from block to block. A another fact is that the MDCT transform is not is strictly energy-saving. So it can be stated that the MDCT transformation, together with an inverse MDCT transformation works exactly, however, the MDCT spectrum significant differences to a Fourier spectrum Has. A spectral-value prediction of MDCT spectral coefficients turned out to be insufficient when high quality demands are made.

Another disadvantage of the spectral-value prediction in particular in connection with modern audio coding methods is that modern audio coding methods are different Use window lengths or window shapes. In order to avoid, that there are rapid changes in the code to be encoded Audio signal, d. H. in the case of transients or attacks, the introduced by the quantization of the MDCT spectral coefficients Quantization noise "smeared" over a long block, d. H. use so-called pre-echoes modern transformation encoders for transient audio signals, d. H. Audio signals with stops, short windows, about temporal resolution at the expense of frequency resolution to increase. However, this leads to the fact that with a spectral value Prediction constantly both window length as well Window shape (there are also transition windows to a Windowing from short to long blocks and vice versa) must be taken into account, which too a complication of the spectral-value prediction contributes and significantly affect computing efficiency would.

DE 40 34 017 A1 relates to a method for recognition errors in the transmission of frequency-coded digital signals. Here, the frequency coefficient previous and possibly future blocks an error function formed on the basis of which the occurrence of an error is determined. An incorrect frequency coefficient will no longer used for the evaluation of subsequent blocks.

DE 197 35 675 A1 discloses a method for concealing errors in an audio data stream. For this, the spectral energy of a subset of intact audio data calculated. After creating a template for replacement data based on those calculated for the subset of the intact audio data spectral energy will be substitute data for faulty or non-existent audio data belonging to the subgroup correspond, generated based on the template.

The document "Error Concealment in Spectrally Coded Audio signals ", Jürgen Herre, dissertation of the university Erlangen-Nuremberg, 1995, discloses a prediction configuration for concealing errors in spectrally coded Audio signals.

The object of the present invention is a precise and flexible error concealment for audio signals to create that implemented with limited computational effort can be.

This task is accomplished by a method of disguising a Fault according to claim 1 and a device for Disguising an error according to claim 12 solved.

Another object of the present invention is a robust and flexible decoding of audio signals to accomplish.

This task is accomplished by a method for decoding a encoded audio signal according to claim 10 and by a Device for decoding an encoded audio signal according to Claim 13 solved.

The present invention is based on the finding that that the disadvantages of the spectral-value prediction, which in the dependence on the transformation algorithm used and depending on the window shape and block length exist, can be avoided that for Error concealment a prediction is used, which in "Quasi" time range works. For this, a set of spectral values, preferably a long block or one Corresponds to the number of short blocks, divided into subbands. A subband of the current set of spectral coefficients can then undergo a backward transformation to obtain a time signal that corresponds to the spectral coefficient of the subband. For generation of estimates for a subsequent set of spectral coefficients is a prediction based on the Time signal of this subband performed.

It should be noted that this prediction is in the quasi-time range takes place because the temporal signal on which Basis of which the prediction is made, only that Time signal of a subband of the encoded audio signal and not the time signal of the entire spectrum of the audio signal is. The time signal generated by prediction becomes one Subjected to forward transformation to estimated, i.e. H. predicted Spectral coefficients for the subband of the following Obtain set of spectral coefficients. Now it is found that in the following set of spectral coefficients one or more incorrect spectral coefficients , the incorrect spectral coefficients can be caused by the estimated, d. H. predicted, spectral coefficients be replaced.

In contrast to pure spectral value-based prediction required the inventive method for concealing Errors require less computing effort because of the Grouping of spectral coefficient predictions only for each subband and no longer for each spectral coefficient must be carried out. In addition, the invention provides Processes high flexibility because of the properties of the signals to be processed are taken into account can be.

The noise substitution according to the present invention works especially good for tonal signals. However, it was found that tonal signal components tend to be in the lower frequency range Range of the spectrum of an audio signal occur while the higher-frequency signal components tend to be non-stationary, d. H. are intoxicated. "Noisy signal components" are signal components in the sense of the present description that are not very stationary. These noisy signal components do not necessarily have to have noise in the classic Represent meaning, but only rapidly changing Useful signals.

It enables to further reduce the computing effort the present invention, therefore, only low frequency To subject signal components to a prediction, while higher-frequency ones Signal components not processed at all become. In other words, it is possible only the lower subband or bands of a backward transformation, a prediction and a forward transformation undergo.

This feature of the present invention compares to a complete transformation of the whole Audio signal in the time domain and a prediction of the whole temporal audio signal from block to block below Using a so-called "long-term" predictor one represents significant advantage, since the advantages according to the invention the prediction in the time domain with the advantages of the spectral Decomposition can be combined. Spectral decomposition only allows properties of the audio signal that are dependent on the frequency. The Number of subbands that are used when dividing the set of Spectral coefficients are generated, can be selected as desired. If only two subbands are selected, the result is already the advantage of considering the tonality in the lower Frequency range of the audio signal. In contrast, there will be many Subbands selected, so the predictor is in the quasi-time range have a relatively short length such that its Delay doesn't get too big. Because the individual subbands preferably processed in parallel, would be at an embodiment of the present invention using of a hardwired integrated circuit many parallel predictor circuits necessary.

Will the present invention in connection with a Transformation encoder used, the different Block lengths used, there is the advantage that the Predictor itself of block length ("frame length") and window shape ("Window Shape") is independent. Besides, will through the back transformation the dependence on that used transformation algorithm itself, the above regarding the MDCT has been eliminated. Farther provides the concept for concealing errors estimated spectral coefficients due to the backward transformation, the prediction in the time domain and the forward transformation is in phase, i.e. H. it there are no phase jumps in the time signal due to a predicted spectral coefficients versus a time signal of a previous intact set of spectral coefficients on. This makes tonal signals so good substituted for incorrect or missing signal components become a common listener in the vast majority of cases doesn't even notice that an error has occurred.

Finally, the method according to the invention is particularly suitable good for a combination with an error concealment technique, which is described in DE 197 35 675 A1, those for the substitution of noisy signal components suitable is. If tonal signal components of a missing Blocks obscured by the inventive method, and become noisy signal components through the just mentioned known methods that on an energy similarity between builds substituted data and intact data, combines, so completely failed blocks can almost be inaudibly obscured by a normal listener.

Fig. 1

einen Decodierer, der eine erfindungsgemäße Fehlerverschleierungseinrichtung aufweist;

Fig. 2

ein detaillierteres Blockschaltbild der Fehlerverschleierungseinrichtung von Fig. 1;

Fig. 3

ein detaillierteres Blockschaltbild der Fehlerverschleierungseinrichtung von Fig. 1, die zudem eine Rauschersetzung aufweist und basierend auf dem Prädiktionsgewinn arbeitet;

Fig. 4

ein Flußdiagramm für das erfindungsgemäße Verfahren zur Fehlerverschleierung;

Fig. 5

ein detailliertes Blockschaltbild eines bevorzugten Ausführungsbeispiels der Fehlerverschleierungseinrichtung für einen MPEG-2 AAC-Decodierer;

Fig. 6

ein detailliertes Blockschaltbild des Prädiktors von Fig. 5; und

Fig. 7

eine schematische Darstellung der Blockstruktur nach dem AAC-Standard.

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

Fig. 1

a decoder having an error concealment device according to the invention;

Fig. 2

a more detailed block diagram of the error concealment device of Fig. 1;

Fig. 3

a detailed block diagram of the error concealment device of Figure 1, which also has a noise replacement and works based on the prediction gain.

Fig. 4

a flow diagram for the inventive method for error concealment;

Fig. 5

a detailed block diagram of a preferred embodiment of the error concealer for an MPEG-2 AAC decoder;

Fig. 6

a detailed block diagram of the predictor of Fig. 5; and

Fig. 7

is a schematic representation of the block structure according to the AAC standard.

Fig. 1 shows a block diagram of a decoder according to one preferred embodiment of the present invention. The decoder block diagram shown in Fig. 1 corresponds basically the MPEG-2 AAC decoder, as in Standard MPEG-2 AAC 13818-7 is set. The encoded audio signal first gets into a bitstream demultiplexer 100 to separate spectral data and side information. The Huffman encoded spectral coefficients are then in a Huffman decoder 200 is input to from the Huffman code words to obtain quantized spectral values. The quantized spectral values are then converted into an inverse Quantizer 300 fed in and then by scale factor band multiplied by corresponding scale factors. The encoder according to the invention can follow the inverse quantizer 300 several other functionalities have, such as B. a middle / side step, a predictor step, a TNS level etc. as specified in the standard.

According to a preferred embodiment of the present Invention comprises the decoder immediately before a synthesis filter bank 400 an error concealer 500, which works according to the invention and ensures that the Effects of transmission errors in the encoded audio signal, which is fed into the bitstream demultiplexer 100 will be alleviated or made completely inaudible can. In other words, the error concealer does 500 that transmission errors disguised become, d. H. that they are in a temporal audio signal on Synthesis filter bank output not or only weakly are audible.

2 shows a general block diagram of the error concealment device 500. It includes a reverse transformation device 502, a device 504 for generation of estimates and a device 506 for forward transformation. Both the reverse transformation device 502 as well as the forward transformation device 506 depend on the block type that is currently Controllable via a block type line 508. The error encryption facility 500 also includes a parallel branch, around the input spectral coefficients below Bypassing the reverse transformer 502, the Means for generating estimates 504 and the forward transform means 506 directly from the entrance to To direct output. This parallel branch includes one Time delay stage 510 to ensure that behind forward transformer 506 estimated spectral coefficients for a subsequent block at the same time with "real", possibly faulty Spectral coefficients for the following block on one Error selection facility 512 may be pending incorrect spectral coefficients in the real spectral coefficients for the following block by estimated spectral coefficients to be able to replace for the following block. This spectral value replacement is by a Switch symbol 512 shown in Fig. 2. It was on it indicated that the error replacement device 512 either spectral value or block or block can work. Depending on the requirement, the same can also work sub-band. At the exit of the error replacement device 512 then lies the following set of spectral coefficients in which may have been originally incorrect Spectral coefficients through estimated spectral coefficients have been replaced, d. H. in which errors are obscured.

At this point it should be noted that the block diagram, that is shown in Fig. 2, only a part of the error concealer 500. This However, presentation was chosen for reasons of clarity. As shown in Fig. 5 based on a preferred embodiment of the present invention explained 2, the circuit shown in FIG. 2 is a device upstream for subdivision into subbands. Analogous for this purpose the error replacement device 512 is a device to undo the division into subbands downstream, such that the filter bank 400 (Fig. 1) one "normal" set of spectral coefficients obtained without anything from the previous concealment of errors notice. The error concealment device 500 (FIG. 1) thus includes a plurality of those described with reference to FIG. 2 Circuits, one circuit for each Subband. The parallel circuits are on the input side connected by the partitioning device and on the output side through the undo facility Subdivision as detailed later.

It was pointed out earlier that modern Transformation encoder to increase the temporal resolution in the case of transients in an audio signal to be encoded use short windows. It is common that the number of temporal samples or the number the spectral coefficients in a long window or Block an integer multiple of the number of temporal Samples or spectral coefficients in a short window or block. An advantageous effect of the present Invention is that device 504 to generate estimates regardless of the one used Transformation, from the block length used or can work from the type of window used. Therefore both the reverse transformer 502 and forward transformer 506 block type dependent controlled by means 504 for generating estimates always supply the same number of time samples or dissipate from the same.

To further illustrate this property, see below with reference to Fig. 7 to the situation for MPEG-2 AAC. 7 includes a time axis 700, the extent of a long block 702 is shown is. A long block contains 2048 samples, from which 1024 spectral coefficients result if a 50% Window overlap is used as is known. Background information on the modified discrete cosine transformation used (MDCT) and the window overlap itself in the standard already cited. In Fig. 7 are also eight short blocks 704 are drawn, each 256 Samples back to due to the 50% overlap 128 spectral coefficients. For reasons of clarity 7 the overlap of the short Blocks as well as the overlap of the long block with one preceding long block or with a preceding one or a subsequent start or stop window located. In any case, it can be seen from FIG. 7 that the number of spectral coefficients of a long block equal to eight times the number of spectral coefficients of a short block. In other words, includes long block the same duration of the audio signal as eight short blocks.

As shown in Fig. 2, the backward transformation means 502 via block type line 508 such controlled to have eight consecutive backward transformations of the spectral coefficients in corresponding Sub-bands of short blocks and the the quasi-time signals obtained simply line up in series, around means 504 for generating estimated values to be supplied with a time signal of a certain length. Analogously, the forward transformation device 506 again eight consecutive forward transformations perform, one after the other, with the values given by the means 504 for generating estimated values serially be issued. Thus this "duty cycle" requires that in the case of short blocks, the same number of spectral coefficients is spent, as in the case of long Blocks. The spectral coefficients generated by the error concealer 500 spent in a "duty cycle" are in the sense of the present invention as Denoted set of estimated spectral coefficients. Out For practical reasons, the number of spectral coefficients corresponds in a set the number of spectral coefficients in a long block and the number of spectral coefficients of eight short blocks. Of course can be any other relationship between long and short block, for example 2, 4 or 16. Usually the situation will be such that the number the spectral coefficients of a long block through the Number of spectral coefficients of a short block divisible is. However, if for some reason this is not the case Case would be the number of a set of spectral coefficients the smallest common multiple of correspond to long and short block, such that independence of the block type at the predictor level, d. H. in the Device 504 for generating estimated values has been reached becomes.

In the following, Fig. 3 is discussed, which is a preferred one Further development of the error concealment device of Fig. 2. In particular, the error concealment device a noise generator 514 extended depending on a prediction gain signal 516 instead of with the forward transformation device 506 the error replacement device via a noise substitution switch 518 can be connected. The Noise Substitution Device 514 works according to that in DE 197 35 675 A1 described method to get noisy signal components in the Approach audio signal. Since it is noisy spectral components is no longer the phase of the spectral coefficients considered, but only the energy several spectral coefficients in a subgroup. The Noise generator 514 generates depending on the Energy in a subset of the most recent ones intact audio data a corresponding subset of Spectral coefficients, the energy in the subgroup the generated spectral coefficients of the energy of the corresponding Subset of the previous spectral coefficients corresponds to or is derived from the same. The phases of spectral coefficients generated in the noise replacement however set at random.

Noise substitution switch 518 is provided by a prediction gain signal 516 controlled. In general, the prediction gain relates on the ratio of the output signal of the Means 504 for generating estimates of the input signal. Is it found that the Output signal differs relatively little from the input signal, it can be assumed that the audio signal relatively stationary in this subband, i. H. tonal, is. In contrast, the output signal of the predictor differs very strongly from the input signal, so it can be assumed become that the signal is transient, d. H. atonal or intoxicating. In this case there is a noise replacement deliver better results than a prediction, because noisy Signals per se cannot be reliably predicted can. For example, the noise substitution switch 518 can be controlled such that it is the forward transformation device 506 with the error replacement facility 512 connects when the prediction gain a certain one Threshold exceeds, or that the noise replacement device 514 connected to the error replacement device 512 if the prediction gain falls below this threshold, to optimize both substitution processes combine.

In the following, referring to FIG. 4, the method the noise substitution according to the invention was discussed in more detail. First, a current set of spectral coefficients received (10). 4 is shown assumed for reasons of clarity that the current Set of spectral coefficients only intact Has spectral coefficients or already an error concealment method subjected to Fig. 2 or 3 has been. The current set of spectral coefficients will be processed on the one hand by the filter bank 400 (FIG. 1) and output to a loudspeaker, for example (12). on the other hand becomes the current set of spectral coefficients used a following set of spectral coefficients predict, d. H. to estimate or to predict. For this purpose, according to the invention, a division of the current Set of spectral coefficients performed in subbands (14). In the case of a long block, the subdivision takes place in subbands in such a way that only one subband per set is generated with a corresponding frequency range. In the case of short blocks, the current set of spectral coefficients a plurality of successive ones complete spectra. Then in the crotch 14 corresponding subbands for each complete spectrum generated, d. H. several per set of spectral coefficients Sub-bands.

After subdivision into subbands there is a reverse transformation per subband (16). In the case of long blocks d. H. the number of spectral coefficients of a block corresponds to the number of spectral coefficients in a set, a single inverse transformation is carried out per subband, before moving on to prediction 18. In the event of from short blocks, several back-transformations become corresponding of the subbands of each "short" spectrum, before making a prediction for all subbands together 18 is carried out.

The prediction 18 takes place in the quasi-time range, i. H. for each subband "time" signal to an estimated subband time signal to get for the following sentence. This esteemed Quasi-time signal then becomes one again Undergone forward transformation 20, the forward transformation for a long block again only once is executed or for short blocks N times, where N is the Relationship between the number of spectral coefficients long blocks to the number of spectral coefficients of a short block.

After step 20 there are estimated for each subband Spectral coefficients. In a step 22 the in the subdivision introduced in step 14 is reversed again made such that after step 22 a following sentence of spectral coefficients is present.

In step 24, the decoder does the following Set of spectral coefficients received. This sentence will error detection 26 to determine if one spectral coefficient, several spectral coefficients or even all spectral coefficients of the following set are incorrect are. The fault detection takes place for experts known way takes place, for example, the CRC checksum (CRC = Cyclic Redundancy Code) over one frame is checked. It is found that a checksum, the is calculated based on the transmitted data, using a the checksum transferred is different the estimated spectral coefficients by the step 22 have been generated instead of the spectral coefficients of the faulty blocks are used. The faulty Spectral coefficients are thus against the estimated Spectral coefficients exchanged (28). Finally be the error-concealed spectral coefficients of the following Set processed to output the temporal samples to be able to (30).

The flow chart of Fig. 4 represents a snapshot, so to speak processing a set of spectral coefficients to a next set of spectral coefficients If the flowchart of FIG. 4 is implemented, for example, of course, only one Filter bank 400 (FIG. 1) used to complete steps 12 and 30 perform. Of course, only one will be the same only device for receiving the current set of Spectral coefficients or to receive the following Set of spectral coefficients may be required to obtain the Implement steps 10 and 24. The temporal synchronicity for steps 10 and 24 in a device, which implements the method according to the invention, through which Time delay stage 510 in the parallel branch (Fig. 2) ensured.

Fig. 5 shows a more detailed representation of the general 2 using the example of an MPEG-2 AAC transform encoder, the fault concealment device according to the invention 500 has. As referring to it already 2 has been included, the Error concealment device 500 (FIG. 1) 520 for dividing the blocks of spectral coefficients in preferably 32 subbands. In the case of long blocks each subband has 32 spectral coefficients. Because the subbands the short blocks cover the same frequency ranges, in the case of short blocks each subband has 4 Spectral coefficients. A division of an entire spectrum into equal sized subbands for simplicity preferred, but a division into unequal Subbands would also be possible, for example based on to the psychoacoustic frequency groups. Every subband will then an inverse modified discrete cosine transform subjected. In the case of long blocks the IMDCT runs once and receives 32 input values. in the In the case of short blocks, eight are consecutive IMDCTs performed, each with 4 of the spectral coefficients, such that there are again 32 quasi-time samples at the output result. These then become the predictor 504 supplied, which in turn estimated 32 quasi-time samples generated, which are transformed using the MDCT 506. in the In the case of long blocks, a single MDCT with 32 temporal values performed while in the case of short Blocks eight consecutive MDCTS with each 4 samples can be executed. Although only one in FIG Branch for the zero subband is shown noted that for each subband an identical branch exists if the subbands are all the same length. If the subbands have different lengths, they are The regulations of the IMDCT and MDCT have been adapted accordingly. For one practical implementation offers a parallel Processing. Of course, there is also one serial processing of the subbands possible in succession, if appropriate storage capacities are provided. The output values of the MDCT 506 for each subband are shown in means 522 for undoing the subdivision, d. H. fed into an inverse subdivision device, um in the case of the preferred embodiment at the AAC-MDCT level an estimated set of spectral values issue.

6 shows a further detailed illustration of the predictor 504. The heart of the predictor 504 is the preferred one Embodiment a so-called LMSL predictor 504a, which has a length n = 32. Details about the LMSL predictor are in the book "Adaptive Signal Processing", Bernard Widrow, Samuel Stearns, Prentice-Hall, 1995, pp. 99 ff., to find. The LMSL predictor 504a is a time delay stage 504b upstream. Predictor 504 includes on the input side also a parallel-serial converter 504c and on the output side a serial-to-parallel converter 504d. It also has a prediction gain calculator 504e, which is the output signal of the predictor 504a compared to the input signal to determine whether a stationary signal or a transient signal has been processed. The prediction profit calculator 504e supplies the prediction gain signal on the output side 516, which is used to control the switch 518 (FIG. 3) is used to either predict spectral coefficients or spectral coefficients obtained by noise substitution to be used for concealing errors. The predictor 504 also includes two in its implementation as an LMSL predictor Switches 504f and 504g, which have two switch positions. The switch position "1" affects the case that spectral coefficients of the following block are correct while the switch position "2" relates to the case that spectral coefficients of the following sentence are incorrect. In Fig. 6 shows the case where the spectral coefficients are faulty. In this case, the switch is held at 504g of the input signal is a reference signal with a value of 0 fed into the predictor. In the case of error-free spectral coefficients (Switch position "1" of switch 504g) the output values of the parallel-serial converter fed into the LMSL predictor from below.

In the case of using the error concealment method according to the invention in connection with an AAC encoder it is preferred for all forward and backward transformations the corresponding transformation algorithms (MDCT or IMDCT) to use. For the concealment of errors however, it is not necessary to move backwards or the same transformation method for the forward transformation is used in the coding of the Audio signal was used to measure the spectral coefficient to build.

Due to the division of the spectrum into subbands and based on the individual transformations for each subband are frequency-time domain transformations for each subband with lower order than the frequency resolution used accordingly. Thus, special estimates for tonal signal components in the intermediate level using the Predictor generated. As a forward transformation / synthesis Lower order time-frequency domain transformations than the original frequency resolution used with the same order as for the frequency-time domain transformation used. Consequently provides the error concealment according to the invention on the one hand Flexibility using previous knowledge spectral Properties of audio signals and on the other hand independence of the transformation method used in the encoder by generating the estimates in the quasi-time signal, not at the spectral coefficient level. If the Prediction in the quasi-time domain to replace tonal signal components is used and if the noise replacement is used for noisy spectral components, so Errors for a large class of audio signals themselves complete block loss are obscured such that almost no audible interference occur. Have attempts demonstrated that normal test signals are not critical Listener, d. H. untrained test listeners, even with complete ones Block loss only in 1 out of 10 irregularities heard in the audio signal.

Claims (13)

1. A method for concealing an error in an encoded audio signal, where the encoded audio signal has successive sets of spectral coefficients, where a set of spectral coefficients is a spectral representation for a set of audio sampled values, comprising the following steps:

   subdividing (14) a current set of spectral coefficients into at least two sub-bands with different frequency ranges, where one sub-band of the at least two sub-bands has at least two spectral coefficients;

   reverse transforming (16) the spectral coefficients of the one sub-band to obtain a temporal representation of the at least two spectral coefficients of the one sub-band;

   performing (18) a prediction using the temporal representation of the at least two spectral coefficients of the one sub-band to obtain an estimated temporal representation for a sub-band of a set following the current set, where the sub-band of the following set has the same frequency range as the sub-band of the current set;

   forward transforming (20) the estimated temporal representation to obtain at least two estimated spectral coefficients for the sub-band of the following set;

   determining (26) whether a spectral coefficient of the sub-band of the following set is erroneous; and

   as reaction to the step of determining, if there is an erroneous spectral coefficient, using (28) an estimated spectral coefficient instead of an erroneous spectral coefficient of the following set so as to conceal the erroneous spectral coefficient of the following set.
2. A method according to claim 1, wherein the one sub-band that is processed in the step of reverse transforming (16) has low-frequency spectral coefficients and the other of the at least two sub-bands has higher-frequency spectral coefficients.
3. A method according to claim 1 or 2, wherein the number of spectral coefficients in a set of spectral coefficients is equal to the number of spectral coefficients in a block (702) of the first length and is N times the number of spectral coefficients in a block (704) of the second length, and wherein N blocks (704) of the second length follow each other, where
   the step of subdividing (14) is performed in such a way that the sub-bands of the blocks of the first length have the same frequency ranges as the sub-bands of the blocks of the second length, so that the number of spectral coefficients of a sub-band of the block of the first length is equal to N times the number of spectral coefficients of the corresponding sub-band of the block of the second length;
   the step of reverse transforming (16) is performed in succession for each corresponding sub-band of the N blocks of the second length to obtain a temporal representation of the spectral coefficients of the corresponding sub-bands of the N blocks of the second length;
   the step of performing (18) a prediction is effected with the temporal representation of all the corresponding sub-bands of the N blocks of the second length; and
   the step of forward transforming (20) is performed successively for each corresponding sub-band of the N blocks of the second length.
4. A method according to one of the preceding claims, wherein a plurality of sub-bands is generated in the step of subdividing (14) such that all the sub-bands together form the spectral representation of the encoded audio signal in a set of spectral coefficients.
5. A method according to one of the preceding claims, wherein the following step is performed after the step of determining (26) whether a spectral coefficient of a sub-band is erroneous:

   determining (504e) whether the spectral coefficient represents a tonal portion of the uncoded audio signal by comparing the spectral coefficient with the corresponding estimated spectral coefficient;

   if the spectral coefficient is found to be tonal, using the estimated spectral coefficient, and, if the spectral coefficient is found to be non-tonal, performing a noise substitution (514) for an erroneous spectral coefficient of the following set.
6. A method according to one of the claims 3 to 5, wherein the spectral coefficients are MDCT coefficients, the length of a set corresponds to the length of a long block and has 1024 MDCT coefficients, while a set of spectral coefficients comprises eight short-length blocks, each with 128 MDCT coefficients, and wherein 32 sub-bands, each with 32 MDCT coefficients for a long block or each with 4 MDCT coefficients for a short block, are formed in the step of subdividing.
7. A method according to one of the preceding claims, wherein an adaptive back-coupled predictor (504a), preferably an LMSL predictor, is used in the step of performing (18) the prediction.
8. A method according to one of the preceding claims, wherein the transform algorithm which forms the basis of the encoded audio signal is the same transform algorithm that is used in the step of reverse transforming (16) and in the step of forward transforming (20).
9. A method according to one of the preceding claims, wherein the transform algorithm which is used in the step of reverse transforming (16) is the exact inverse of the transform algorithm that is used in the step of forward transforming (20).
10. A method for decoding an encoded audio signal which comprises successive sets of spectral coefficients, wherein a set of spectral coefficients is a spectral representation for a set of audio sampled values:

    receiving (10) a current set of spectral coefficients;

    subdividing (14) a current set of spectral coefficients into at least two sub-bands with different frequency ranges, where one sub-band of the at least two sub-bands has at least two spectral coefficients;

    reverse transforming (16) the spectral coefficients of the one sub-band to obtain a temporal representation of the at least two spectral coefficients of the one sub-band;

    performing (18) a prediction using the temporal representation of the at least two spectral coefficients of the one sub-band to obtain an estimated temporal representation for a sub-band of a set following the current set, where the sub-band of the following set has the same frequency range as the sub-band of the current set;

    forward transforming (20) the estimated temporal representation to obtain at least two estimated spectral coefficients for the sub-band of the following set;

    receiving (24) a following set of spectral coefficients and subdividing the following set into sub-bands which cover the same frequency range as the sub-bands of the current set;

    determining (26) whether a spectral coefficient of the sub-band of the following set is erroneous;

    as reaction to the step of determining, if there is an erroneous spectral coefficient, using (28) an estimated spectral coefficient instead of an erroneous spectral coefficient of the following set so as to conceal the erroneous spectral coefficient of the following set; and

    processing (30) the following set using the estimated spectral coefficient used in the step of using (28) to obtain the following set of audio sampled values.
11. A method according to claim 10, wherein the spectral coefficients of the encoded audio signal are entropy-coded and quantized, which includes the following steps before the step of receiving (10) the current set or the following set:

    cancelling (200) the entropy coding to obtain quantized spectral coefficients;

    requantizing (300) the quantized spectral coefficients to obtain requantized spectral coefficients;

    and wherein the step of processing includes the following step:

    reverse transforming (400) the following set using a transform algorithm which is inverse to the transform algorithm used for transforming to obtain the spectral coefficients of the encoded audio signal.
12. A device for concealing an error in an encoded audio signal, where the encoded audio signal has successive sets of spectral coefficients, where a set of spectral coefficients is a spectral representation for a set of audio sampled values, with the following features:

    a unit (520) for subdividing (14) a current set of spectral coefficients into at least two sub-bands with different frequency ranges, where one sub-band of the at least two sub-bands has at least two spectral coefficients;

    a unit (502) for reverse transforming (16) the spectral coefficients of the one sub-band to obtain a temporal representation of the at least two spectral coefficients of the one sub-band;

    a unit (504) for performing (18) a prediction using the temporal representation of the at least two spectral coefficients of the one sub-band to obtain an estimated temporal representation for a sub-band of a set following the current set, where the sub-band of the following set has the same frequency range as the sub-band of the current set;

    a unit (506) for forward transforming (20) the estimated temporal representation to obtain at least two estimated spectral coefficients for the sub-band of the following set;

    a unit for determining (26) whether a spectral coefficient of the sub-band of the following set is erroneous; and

    a unit (512) for using (28) an estimated spectral coefficient instead of an erroneous spectral coefficient of the following set so as to conceal the erroneous spectral coefficient of the following set.
13. A device for decoding an encoded audio signal which comprises successive sets of spectral coefficients, where a set of spectral coefficients is a spectral representation for a set of audio sampled values, comprising:

    a unit (100) for receiving (10) a current set of spectral coefficients;

    a unit (520) for subdividing (14) a current set of spectral coefficients into at least two sub-bands with different frequency ranges, where one sub-band of the at least two sub-bands has at least two spectral coefficients;

    a unit (502) for reverse transforming (16) the spectral coefficients of the one sub-band to obtain a temporal representation of the at least two spectral coefficients of the one sub-band;

    a unit (504) for performing (18) a prediction using the temporal representation of the at least two spectral coefficients of the one sub-band to obtain an estimated temporal representation for a sub-band of a set following the current set, where the sub-band of the following set has the same frequency range as the sub-band of the current set;

    a unit (506) for forward transforming (20) the estimated temporal representation to obtain at least two estimated spectral coefficients for the sub-band of the following set;

    a unit (502, 510) for receiving (24) a following set of spectral coefficients and for subdividing the following set into sub-bands which cover the same frequency range as the sub-bands of the current set;

    a unit for determining (26) whether a spectral coefficient of the sub-band of the following set is erroneous;

    a unit (512) for using (28) an estimated spectral coefficient instead of an erroneous spectral coefficient of the following set so as to conceal the erroneous spectral coefficient of the following set; and

    a unit for processing (30) the following set using the estimated spectral coefficient to obtain the following set of audio sampled values.

Images