EP1247275B1 - Device and method for determining a coding block raster of a decoded signal - Google Patents

Abstract

In determining a coding block raster on which a decoded signal is based, a segment of the decoded signal is picked out first (11), said segment beginning at a certain output sampling value of the decoded signal. Said segment is then converted into a spectral representation (12), whereupon said spectral representation is then evaluated in relation to a predetermined criterion (13) in order to obtain an evaluation result for the segment. This procedure is repeated for a plurality of different segments beginning at different output sampling values each, in order to obtain a plurality of evaluation results. Finally, the plurality of the evaluation results is searched (14) in order to establish the evaluation result that has an extreme value as compared to the other evaluation results, in such a way that it can be assumed that the segment to which this evaluation result is allocated matches the coding block raster on which the decoded signal is based. According to the invention, this method can be used to determine the coding block raster for any decoded signal that has no explicit information about its coding block raster.

Description

The present invention relates generally to Analysis of coded and decoded in some way Signals and especially on analyzing a decoded signal using an encoding algorithm has been processed on a spectral Representation of the original signal builds up.

It is well known to include audio and / or video signals Use a particular encoding method to encode to get a coded version of the original signal being the encoded version of the original signal basically in that of the original signal should distinguish that the amount of data encoded Signal less than the amount of data in the original signal is. In such a case, the coding algorithm, to get the encoded signal from the original signal, and also the decoding algorithm, which is essentially is a reversal of the coding algorithm as a data reducing Coding algorithm called.

Various exist for data reduction of audio signals Coding algorithms, which are the subject of a number of international Standards are such. B. MPEG-1, MPEG-2, MPEG-4 or also MPEG-2 AAC (AAC = Advanced Audio Coding), whereby the latter coding algorithm, for example, in the international standard ISO / IEC 13818-7 described in detail is.

In the following, reference is made to FIG. 7, which is a Block diagram of an MPEG audio coding method shows. On such an audio encoder typically includes an audio input 70 on which a stream of discrete-time samples which, for example, PCM samples are fed in which are, for example, 16-bit wide. In an analysis filter bank 71 becomes the stream of discrete-time audio samples in coding blocks or frames of samples divided, using an appropriate window function windowed and then into a spectral representation for example by a filter bank or by a Fourier transformation or a variant of the Fourier transform, such as B. a modified discrete cosine transform (MDCT). At the exit of the analysis filter bank 71 are thus successive coding blocks or frames of spectral coefficients before, one block of spectral coefficients the spectrum of a coding block of audio samples. Often there is a 50% overlap successive coding blocks used, so that a window of, for example, 2048 audio samples per block is considered, and through this processing 1024 new spectral coefficients are generated.

The discrete-time audio signal at input 70 is also shown in a psychoacoustic model 72 fed in for data reduction to achieve such that the masking threshold is known of the audio signal depending on the frequency is calculated to in a block 73 which is with quantization and coding, quantization of the spectral coefficients perform which of the masking threshold depends.

In other words, the quantization of the spectral coefficients carried out so roughly that the hereby introduced Quantization noise still below the psychoacoustic Masking threshold caused by the psychoacoustic Model 72 is calculated so that the quantization noise is ideally inaudible. This procedure causes that typically a certain number of spectral coefficients, those at the output of the analysis filter bank 71 are still not equal to 0, set to 0 after quantization because the psychoacoustic model found 72 that they are masked by neighboring spectral coefficients and are therefore inaudible.

Also independent of a psychoacoustic or psychooptical A model exists for each quantizer Quantization step size, with spectral values being smaller than the step size, by quantizing to zero be set. Depending on the quantizer, there are also those Possibility that only values that are significantly smaller than the step size are set to zero, and values, that are just below the increment are rounded up. In the vast majority of cases, every quantizer sets at least some values to zero, which already results in data reduction is achieved.

After quantization there is a spectral representation of the Coding blocks of discrete-time samples before in the the quantization noise below the psychoacoustic if possible Masking threshold is. This data-reducing quantized spectral values can then be dependent from the encoder used using entropy coding, which z. B. be a Huffman coding can be encoded without loss. This will create a stream of Codewords obtained in a bit stream multiplexer 74 Page information still required by a decoder be added, such as B. Information regarding the Analysis filter bank, information regarding quantization, such as B. scale factors, or page information regarding further function blocks. Such others Function blocks in MPEG-2-AAC are, for example TNS processing, the intensity stereo processing, the Middle / side stereo processing, or a prediction from spectrum to spectrum.

At an output 75 of the encoder, which is also a bit stream output is then referred to, that is according to that shown in Fig. 7 Coding algorithm pre-coded signal block by block.

In the case of the decoder, the coded signal is output 75 of the encoder shown in FIG. 7 into a bit stream input 80 fed to a decoder shown in FIG. 8, which is initially in a block 81, which acts as a bitstream demultiplexer is a bit stream demultiplexing operation performs the spectral data from the page information to separate. At the exit of block 81 there are again Codewords before which the individual spectral coefficients represent. Using an appropriate table the code words are decoded to quantized spectral values to obtain. These quantized spectral values will be then in a block 82 which is labeled "inverse quantization" is processed by the data in block 73 (FIG. 7) Recalculate introduced quantization. At the exit of the Blocks 82 are then again dequantized spectral coefficients which is now using a synthesis filter bank 83, which works inversely to the analysis filter bank 71 (FIG. 7), in the time range will be transferred to an audio output 84 to receive the decoded signal.

When considering the coding / decoding concept, which in the 7 and 8, it is clear that it is this is a block-oriented procedure, whereby the Block generation by the analysis filter bank block 71 of Fig. 7 is effected, and wherein the block formation only on Audio output 84 of the decoder shown in FIG. 8 again will be annulled.

It also becomes clear that this is a lossy one Encoder concept acts because that at the audio output 84th present decoded signal generally less information includes as the original present at audio input 70 Signal. Because of the psychoacoustic model 72 controlled quantizer 73 information from the am Audio input 70 present original signal removed which are no longer added in the decoder, but which is waived. This waiver is purely subjective on information, however, due to the psychoacoustic Model 72, based on human hearing is adjusted, ideally no loss of quality led, but only to a desired data compression.

At this point it should be pointed out that in Fig. 7 8 shows the encoder concept described using the example of an audio signal accordingly also on image or video signals is applied, instead of the temporal audio signal a video signal is present, the spectral representation here is not a sound spectrum, but a spatial spectrum. Otherwise, there is also a video signal compression Analysis filter bank, a psycho-optical model, one through it controlled quantization and redundancy coding instead, where also the whole coding / decoding concept block by block expires.

The decoded signal (using the example of FIG. 8 the decoded Audio signal at audio output 84) is typically again a stream of discrete-time samples that have an encoding block grid is based on that in the decoded signal however, is generally not visible unless special Precautions are taken.

During the process of decoding the normal case in the Application, namely the transfer and storage of Audio and / or image signals, is, there are cases, however those of interest, a given decoded signal "to be translated back" into a bitstream representation. This is of particular interest in the following cases, if only the decoded signal is available.

Furthermore, there is often a need to use coding systems of the signals they have coded and decoded again investigate, for example, to find out why an encoder, that is still unknown sounds so good.

There is also a need in the area of copyright protection, to prove beyond any doubt that a piece of music or an image with a particular encoder originally has been encoded.

Finally, in the area of transmission, for example over several networks with different bandwidth the need to re-encode a decoded signal to it for example, to implement a different bandwidth. In In this case, the encoder / decoder concept shown in Fig. 7 and Fig. 8 becomes successively to an original one Audio signal exercised. There are problems here in that so-called tandem coding distortions follow Codec levels are introduced if the following Codec levels based on a different coding block grid work as the previous codec stages. It it is clear that the use of a different coding block raster audible in a subsequent codec stage Introduces distortion into the audio signal when coding block formation has not been carried out exactly like in the first codec stage because the concept is based on education based on short-term spectra and especially the psychoacoustic Masking threshold of a coding block from discrete-time samples from the coding block grid depends.

In the specialist publication "NMR Measurements on Multiple Generations Audio Coding ", Michael Keyhl, Jürgen Herre, Christian Schmidmer, 96th AES meeting, February 26 to 1 March 1994, Amsterdam, Preprint 3803, is proposed to Overcoming the tandem coding distortions an identification mark into a decoded signal the subsequent encoder stages can access to based on this identification mark their coding block division the decoded to be re-encoded Perform signals such that all codec stages in one Chain of codec stages the same coding block grid use.

Although this method the tandem coding distortion is significant has reduced, it is disadvantageous in that that the identification mark by a decoder must be introduced and by a subsequent encoder must be extracted and interpreted again. So there are Changes to both a decoder and one Encoder required. This concept is also a matter of course only for tandem encoding of decoded Applicable signals that this identification mark for have the coding block grid. For signals this A codec level cannot have identification mark naturally not in a chain of codec stages access an identification tag.

Similar problems or limitations of flexibility arise also the MOLE concept, which is in "ISO / MPEG Layer 2 - Optimal re-encoding of decoded audio using a MOLE signal ", John Fletcher, 104th AES Convention, May 16-19 1998, Preprint No. 4706. Generally speaking additional data is introduced into the decoded audio signal, which describe in detail how this is present decoded audio signal encoded and decoded has been. This data is called the MOLE signal. If the decoded audio signal needs to be encoded again, a specially designed encoder of this MOLE signal extract from the signal to be encoded and on the basis perform the individual encoder steps of this signal.

Similar to the concept of identification marking a disadvantage here, too, in that the decoder, the one encoded original signal decoded for the first time, must introduce the signal into the decoded audio signal. On such a decoder thus differs from the usual ones Standard decoders. Furthermore, an encoder that a decoded signal again, extract the determination signal to work accordingly. This, so to speak second encoder must also be modified, such that he can read and interpret the determination signal can. Finally, this concept is also disadvantageous effective only for decoded signals that have such Determination signal, but not for signals that do not have such a determination signal.

Both the identification mark and the MOLE determination signal give information about which coding block grid the decoded signal to which the identification mark or assigned the MOLE determination signal is underlying. However, these signals must be explicit be brought in what is described above Brings flexibility disadvantages with it.

The publication "Extracting coding parameters from pre-coded MPEG-2 video, "Chen Y. et al., IEEE Computer Soc., October 1998, Conference volume 5, pages 360 to 364, XP 002166941, discloses concepts for Extract encoding parameters from pre-encoded MPEG-2 video data. To the problem of deterioration in quality addressing multiple cascaded coding processes is detected whether an image is intra-coded originally. coding parameters are extracted if the picture is intra-coded. An intra-coded image has a discrete histogram of DCT coefficients based on an original picture has a relatively continuous spectrum.

The object of the present invention is a Device and method for determining an encoding block raster, that underlies a decoded signal is to create a decoded signal that does not has an explicit reference to a coding block grid.

This object is achieved by a device for determining a Coding block grid according to claim 1 or by a method for determining an encoding block grid solved according to claim 11.

The present invention is based on the finding that that the coding block grid, which by a block oriented Encoder is set virtually randomly decisive influence on the spectral representation of the signal Has. Even minimal deviations or coding block grid offsets cause the spectral representation of the decoded signal a completely different appearance has actually from a spectral representation of the decoded signal would be expected if the same the same coding block grid is used, which is the basis of the decoded signal. With data-reducing Encoding algorithms using a psychoacoustic model or a psychooptical one Working model is known from the start that due to quantization using a psycho-optical or psychoacoustic masking threshold a certain Number of spectral coefficients is 0.

It should be noted that regardless of one Quantization by a psychoacoustic or psychooptical Model is controlled, usually always certain Values are set to zero, namely the values that are significantly smaller than the quantization step size.

If, however, the coding block rasterization to generate a spectral representation of the decoded signal not with the coding block grid division that the decoded signal per se is the same, so this property occurs in the spectral representation of the decoded signal no longer. But also with coding concepts, that are not necessarily data-reducing, or which, although they would be data-reducing, are due to of the input signal no decisive data reduction effect have an encoding block raster offset already that the spectrum of the decoded signal, the is based on a different coding block grid division than the coding block raster division that the decoded Underlying signal. This results in a changed spectral structure that is a heavily "smeared" Appearance, which is particularly evident in the fact that the individual spectral components are no longer well separated can be.

This characteristic of the spectrum can be used as a criterion to find out if there is a coding block grid offset is present. For a spectrum with grid offset is the fluctuation of z. B. logarithmic amplitude of the spectral coefficient is slower or less abrupt than for a spectrum without grid offset, in which a fast or strongly abrupt fluctuation in the amplitude of the spectral coefficients is noticeable.

Generally speaking, a short-term spectrum of the decoded Signal using a coding block division is generated which of the coding block grid division corresponds to the basis of the decoded signal a certain appearance, for example with regard to the Separation of the spectral lines, based on the number of Spectral lines that are equal to 0 or that are very small, Etc.

The invention is therefore used to determine a coding block grid picked out a section of the decoded signal, whereupon the selected section into a spectral representation of the same is implemented. Subsequently becomes the spectral representation of the selected section with respect to at least one predetermined criterion examined to give an evaluation result for the section to obtain. This concept is different Sections performed, always a different coding block grid is taken as a basis, so that different Evaluation results for different coding block grid divisions and thus coding block raster offsets result. An encoding block raster offset that matches the best meets predetermined criteria, d. H. the one Evaluation result has that with the other evaluation results is extreme, then among the evaluation results, by evaluating the spectral representations of the differently selected sections are determined and output. This is the coding block grid division, the basis of a decoded signal lies, without using an explicitly in the decoded signal contained auxiliary signal can be clearly reconstructed.

This concept basically allows each one to be decoded Signal the coding block raster underlying the same to determine and thus provides considerable flexibility in that all decoded signals are processed can be, and not just decoded signals already an identification mark or a MOLE identification signal to have. This allows almost any decoding Signals are analyzed for distortion free Perform tandem coding for more information regarding the basis of the decoded signal To get encoder algorithm, or to prove at all, with which encoder the decoded signal originally has been encoded.

The coding block grid determined according to the invention can preferably on which the decoded signal is based, in the decoded signal itself to be entered in order any decoded signals for existing codec stages adjust which on the identification mark or build up the MOLE determination signal.

Furthermore, the concept according to the invention allows indexing almost all coding parameters, especially starting from knowledge of the coding block grid and using corresponding iteration algorithms practically all Encoder functionalities to a certain extent "recalculated" can be. However, the prerequisite for this is the provision of the coding block grid itself, since the coding block grid all subsequent parameters of a coding algorithm influenced on the spectral representation of a signal to be encoded. The determination of the Coding block grid is thus, so to speak, the "entrance gate", completely by a decoded signal to analyze which coding / decoding concept is the same underlying.

Fig. 1

ein Blockschaltbild einer erfindungsgemäßen Vorrichtung zum Bestimmen eines Codierungs-Blockrasters;

Fig. 2

ein Flußdiagramm eines erfindungsgemäßen Verfahrens zum Bestimmen eines Codierungs-Blockrasters;

Fig. 3

eine Prinzipdarstellung eines decodierten Signals zur Veranschaulichung verschiedener Codierungs-Blockraster-Versätze;

Fig. 4

eine spektrale Darstellung eines Abschnitts des decodierten Signals mit einem Rasterversatz von einem Abtastwert nach links;

Fig. 5

eine spektrale Darstellung eines Abschnitts des decodierten Signals ohne Rasterversatz;

Fig. 6

eine spektrale Darstellung eines Abschnitts des decodierten Signals mit einem Rasterversatz von einem Abtastwert nach rechts;

Fig. 7

ein Blockschaltbild eines bekannten Codierers, der auf der Basis einer spektralen Darstellung eines ursprünglichen Signals arbeitet;

Fig. 8

ein Blockschaltbild eines bekannten Decodierers zum Decodieren von durch den in Fig. 7 gezeigten Codierer codierten Signalen; und

Fig. 9

eine beispielhafte Fenstersequenz mit einem Überlappungsgrad von 50%.

Preferred exemplary embodiments of the present invention are described in detail below with reference to the accompanying figures. Show it:

Fig. 1

a block diagram of an inventive device for determining an encoding block grid;

Fig. 2

a flowchart of a method according to the invention for determining a coding block raster;

Fig. 3

a schematic diagram of a decoded signal to illustrate different coding block raster offsets;

Fig. 4

a spectral representation of a portion of the decoded signal with a raster offset of one sample to the left;

Fig. 5

a spectral representation of a portion of the decoded signal without raster offset;

Fig. 6

a spectral representation of a portion of the decoded signal with a raster offset of one sample to the right;

Fig. 7

a block diagram of a known encoder, which works on the basis of a spectral representation of an original signal;

Fig. 8

a block diagram of a known decoder for decoding signals encoded by the encoder shown in Fig. 7; and

Fig. 9

an exemplary window sequence with a degree of overlap of 50%.

Fig. 1 shows a block diagram of an inventive Device for determining a coding block raster, the is based on a decoded signal. The decoded Signal is at an input 10 in the invention Device is fed in and reaches a device 11 for picking out a section from the decoded Signal. The one picked out by the device 11 Section is in a device 12 in a spectral Representation of the same implemented. The spectral representation the selected section is then in a Device 13 with respect to a predetermined criterion rated to an evaluation result for the singled out Get section. The evaluation result is then into a device 14 for browsing and output a plurality of evaluation results are input to an output 15 of the device according to the invention decoded signal at input 10 of the invention Output device-based coding block grid. The device shown in Fig. 1 operates iteratively, such that the device 11 for picking out depending on a section control signal 16 one Can pick out section of the decoded signal that differs from a previously selected section. The inventive device for determining a coding block grid is thus arranged to one A plurality of sections of the decoded signal, which at different output samples start to pick out implement and determine a plurality of Get evaluation results. From this majority of The device 14 then determines the evaluation results selected section, which meets the criterion, which the Assessment is based, best corresponds, or the depending on the criterion least corresponds to it give an indication of the coding block grid.

In the following, reference is made to FIG Structure of a decoded signal at input 10 shown in FIG. 1 device according to the invention and the various To represent coding block raster offsets. The decoded signal generally consists of a sequence 30 of discrete-time samples, for example that of FIG. 8 has shown the decoder generated at its audio output 84. In particular, the sequence 30 consists of discrete-time Samples of the decoded signal from samples 31a, 31b, 31c, 31d, .... In Fig. 3 is also outlined in bold Coding block 32 drawn from samples, which defines the coding block grid division that the decoded signal 30 is originally based. Fig. 3 represents the case where no overlap is used, while Fig. 9, which will be discussed below, a Window sequence representing a 50% overlap used.

The coding block grid is in the sense of the present Description defined such that a coding block Includes samples taken from an analysis window the stream of] time samples become. The number of samples in a coding block thus corresponds to the number of samples that the Windows are used, or in other words, the Window length. Since there is no overlap of the temporal in FIG Window is present, ends before that in FIG. 3 by way of example shown coding block 32 a previous coding block and begins at the end of coding block 32 subsequent coding block.

9 shows a window sequence in which there is an overlap of 50% is used. Such a window sequence can occur with MPEG-2 AAC. Along the abscissa of Fig. 9 is the number of a discrete sample in a stream plotted from samples. Along the ordinate in Fig. 9 the relative size of the window is plotted; H. the Factor with which a sample value is weighted when windowed.

The window sequence in Fig. 9 includes a "long" window 90, a so-called start window 92, a sequence of eight "short" Windows 94, a stop window 96 and another long one Window 98.

With the standard MPEG-2 AAC, an encoder can be used to strong to be able to encode transient time signals better, from one long windows to a sequence of eight short windows switch. The window sequence in Fig. 9 is therefore for this suitable for transient time signals between sample no. 2560 and sample no. 3584 to process.

In the case shown in Fig. 9, a long one Window 2048 samples while a short window 256 Samples. The eight short windows 94 include as many samples as a long window 90 or 98. In addition, the start window 92 and the stop window 96 chosen such that after a transition from the window with long windows to a window with short ones Windows and after an opposite transition again back to windows with long windows the coding block grid of n · (1024 samples) is maintained. The Coding block grid is here through a long window defined, d. H. by the number of samples that includes a long window.

With an overlap of 50% included in the case of a sequence from long windows each new window 50% of the samples, that were windowed through the previous window and 50% "new" sampled values. Will have a higher overlap used as 50%, the number of "New" picked samples in a coding block, while the number of "old" samples increases. The total number of samples per coding block remains however, the same.

The device according to the invention for determining a coding block raster thus only needs a single coding block of the decoded signal because the coding block grid is usually fixed in a signal and, even if short windows are used, not themselves generally changes.

In Fig. 3 there are also three possible controls of the device 11 (Fig. 1) drawn for picking out, namely a first alternative 33 with an offset of one Left sample, i.e. H. an offset of -1, one second alternative 34 with an offset of 0 and a third Alternative 35 with one sample offset to the right, d. H. with an offset of +1.

In the following, Fig. 2 is discussed, which is a Flow diagram of the method according to the invention provides. First is a first via the control line 16 (Fig. 1) Relocation of the pickup 11 notified, d. H. a first offset is set (step 20). After that this section determined by the first offset, which at an output sample of the decoded Signal begins through the device 12 in its spectral range Presentation implemented, d. H. it will be a spectral analysis this section with this offset (Step 21). This is followed by the spectral representation at the output the device 12 (FIG. 1) in the device 13 (FIG. 1) rated, d. H. it will be an assessment of the spectrum to obtain an evaluation result (step 22). Then it is determined in a step 23 whether all the desired ones Offsets have already been run through, d. H. if the search area has been run through. If not, i. H. the decision in step 23 returns a "no", see above in a step 24 via the control line 16 of the device 11 communicated a new offset for picking out, thus the iteration loop again with this new offset can be run through. Once the search area has been run through, d. H. delivers the decision in step 23 "Yes", so the different evaluation results are searched, and the evaluation result is determined that regarding the other evaluation results depending on the criterion is either maximum or minimum, then an identification the basis of the decoded signal Coding block grid based on the section that had the cheapest evaluation result in a step 25 issue.

In the following, reference is made to FIGS. 4 to 6, around the evaluation performed by the device 13 or to explain step 22 of FIG. 2 in more detail. In the 4 to 6 is the coefficient number along the abscissa applied. 4 to 6 thus show graphical Representation of spectra if the coefficient number with multiplied the bandwidth of a spectral coefficient becomes. Along the ordinate of that shown in FIGS. 4 to 6 graphs is the absolute amount of Logarithmic spectral coefficients applied.

4 shows the spectral representation of a selected section with an offset of minus a sample, which corresponds to alternative 33 of FIG. 3. There is a clearly smeared spectrum, in which no well-defined spectral coefficients are present, and in addition only a very small one Number of spectral coefficients equal to 0 or less than a predetermined threshold.

For comparison, a spectral representation of one is selected Section shown, no grid offset has, d. H. Alternative 34 of Fig. 3. It is clear to recognize defined spectrum, in which a multitude of Spectral lines depending on the quantization from the psychoacoustic masking threshold 0 or very are small, and furthermore all spectral lines are clean have a defined structure.

Finally, in Fig. 6 is a spectral representation of a selected section shown, which one Has a grid offset of plus one sample, d. H. which one corresponds to the third alternative 35 of FIG. 3. It is can clearly be seen that, in contrast to FIG. 5, the spectrum 6 is heavily smeared again.

The following is based on various evaluation criteria discussed in more detail. In principle, any property can be used as a criterion of the spectrum shown in FIG. 5 are used, which differs from a property of that shown in Figs Spectra differs. Most visible is that in the spectrum shown in Fig. 5, there is no raster offset is based on a large number of spectral lines smaller than z. B. 30 dB, i.e. H. about 70 dB below the significant spectral coefficients. Different expressed a large number of spectral lines is the same 0 or less than 30 dB. A criterion can therefore be a simply counting the spectral lines equal to 0 is used in order to evaluate the results different from 0 To use spectral lines of a selected section.

The section with the least number of different from 0 Spectral values or the largest number of spectral lines the portion of the output sample would then be 0 of the decoded signal starts (here the Sample 31c of Fig. 3), which is also the first sample used when encoding the original signal Analysis window is. There is therefore no grid offset here in front.

Alternatively, a decision threshold can also be used as a predetermined criterion used to be used as the evaluation result either the spectral values with an amount above the threshold or an amount below the threshold issue.

Alternatively, a predetermined criteria can be used to determine the correct coding block grid also on the evaluation the rapid or abrupt fluctuation of the z. B. logarithmic Amplitude of the spectral coefficients are based. in the The squared difference between two spectral coefficients is averaged 4 and 6 (with grid offset) lower be as in Fig. 5 (without grid offset). Like in The first example can also be a decision threshold can be used to determine a "fluctuation rate" of the spectrum with an amount above the threshold or an amount below the threshold issue.

At this point it should be noted that a spectrum as shown in Fig. 5, only becomes visible when next the parameters of the analysis filter bank with the correct grid offset 71 (Fig. 7) match. Such parameters are for example the filter bank type (e.g. DFT, DCT, MDCT), the Coding block length and the window shape. In the in the 4 to 6, the example shown was, for example Filter bank according to MPEG-2 AAC, as a window form a KBD window (KBD = Kaiser-Bessel-Derived) and as coding block length a long block (only long sequence) is set.

Often the case is actually such that it is from the outset decoded signal is known to encode according to MPEG-2 AAC and has been decoded again. Even if it does is not known, the per se iterative concept can the present invention, as shown in Figs. 1 and 2 is easily modified such that also the device 12 for converting into the spectral representation (Fig. 1) is operated iteratively to implement different implementation parameters in the spectral representation to be based on in a double iteration loop in connection with the control of the section that is picked out, in addition to the coding block grid to determine the coding algorithm used. It will noted that only a limited number of Encoder candidate is practically relevant, which is why the invention Concept even in a limited time to one Result comes, albeit the encoder that does the present generated decoded signal is still unknown.

In general, as has already been stated, the Determination of a single coding block 32 (FIG. 3) in order the entire coding block raster that the decoded Signal is to be determined generally. To also the Switching from long coding blocks to short coding blocks or maybe even on other grid divisions To be able to understand, the inventive Methods are modified so that the length of a Section which the device 11 for picking out is to be communicated, is also varied in order to achieve that in FIG shown iterative methods for different coding block lengths to repeat. When using short windows facilities 12 and 13 will also be informed. This means that from some of the grid points found entire grid can be extrapolated or, as it is Example of the short coding blocks has been shown, even in its possible fine structures broken down become.

Have been used in the encoding that underlies the decoded signal additional coding "tools" are used, so can by an advanced search or by additional Calculations also determine these configurations.

If M / S stereo coding when generating the decoded signal (J.D. Johnston, A.J. Ferreira: "Sum Difference Stereo Transform Coding ", IEEE ICASSP 1992, p. 569 - 571), also called center / side coding or as Sum / difference coding is used the iterative determination of the Coding block raster not on the decoded signal executed itself, but on the sum or difference of Spectral values. A significant one is then shown, for example Number of vanishing (sum and difference) Spectral coefficients, M / S coding is concluded, and any subsequent bills will then be included the sum and difference spectral coefficients. The predetermined criterion can be modified here that individual criteria of the sum signal and the difference signal weighted together in a suitable manner, so that the predetermined criterion on both the sum signal as well as based on the difference signal.

If TNS coding when generating the decoded signal (TNS = Temporal Noise Shaping = temporal noise shaping) (J. Herre, J.D. Johnston: "Enhancing the Performance of Perceptual Audio Coders by Using Temporal Noise Shaping (TNS)) has been used, the coding block grid based on the "low frequency" spectral coefficients which are usually not one TNS coding. Usually spectral coefficients not subjected to TNS coding below 1 kHz. However, this value can of course vary from Vary from case to case.

Although the inventive concept for determining a coding block grid described using an audio coding concept , it should be noted that this Concept is also applicable to video encoders. The invention Concept generally applies to all coding algorithms applicable to all signals if these coding algorithms have the property that they are on a build up spectral representation of the signal to be encoded. Whenever this is the case, different coding block divisions can be used for the decoded signal one generated spectral representation of the selected section then the spectral representation with regard to to evaluate a predetermined criterion.

Finally, it should be noted that the device according to the invention not necessarily to determine an encoding block grid must work in series so that an evaluation result after another is generated, d. H. that about the Control lines 16 (Fig. 1) the device 11 for picking out is controlled to gradually one for example to pick out 1 shifted section. ever according to implementation constraints, the invention Device also completely or partially parallel be implemented so that, for example, 1024 evaluation results can be generated in one processing step. Also mixed serial / parallel options are possible so that For example, there are eight parallel branches, which then exist correspondingly often work in series to cover an entire search area to be able to cover.

At this point it should also be noted that not always necessarily run through a whole predetermined search area must become. If, as in the present case, the distinction between the spectrum without grid offset and one Spectrum with a minimal grid offset possible so clearly the iteration shown in FIG. 2 can be canceled even if a predetermined one Criterion is met, since there is actually no longer any doubt about it is that the one tested here is the Section is a section that is related to the original Coding block grid is synchronous.

In addition, it should be noted that the coding block grid by any definition can be identified, and not just by the initial sample of a coding block. Of course, each sample can be one Coding blocks of samples are used to to define the coding block grid. Finally, can the coding block grid also deviates from the number of samples per window are defined such that two grid points of the coding block grid around the z. B. twice the number of samples of a window from each other are spaced.

Claims (11)

1. A device for determining a coding block raster on which a decoded signal is based, in which the decoded signal is produced from an original signal by coding and decoding according to a coding algorithm including a coding block generating step, a conversion step and a data reducing step, said coding block generating step of the coding algorithm including partitioning the original signal according to the coding block raster into coding blocks with a specific number of time-discrete signal values, said conversion step including generating from a coding block a spectral representation of the same, and said data reducing step including removing information from the spectral representation of the original signal, said device comprising:

   means (11) for picking out a segment of the decoded signal, said segment beginning at an output sampling value of the decoded signal;

   means (12) for performing the conversion step on said segment of the decoded signal so as to provide a spectral representation of said segment;

   means (13) for evaluating the spectral representation of said segment with respect to a predetermined criterion in order to obtain an evaluation result for the segment,

   said device for determining a coding block raster being further arranged to pick out, convert and evaluate a plurality of segments of the decoded signal that begin at different output sampling values in order to obtain a plurality of evaluation results; and

   means (14) for searching the evaluation results and for outputting an identification for the coding block raster underlying the decoded signal, on the basis of the segment that has an extreme evaluation result with respect to other evaluation results.
2. A device according to claim 1,
   wherein said means (13) for evaluating is arranged to use as predetermined criterion the number of spectral coefficients of the spectral representation that is smaller than a predetermined threshold value.
3. A device according to claim 1,
   wherein said means (13) for evaluating is arranged to use as predetermined criterion a measure for a fluctuation of preferably logarithmic amplitudes of spectral coefficients of the spectral representation.
4. A device according to any of the preceding claims,
   wherein said means (13) for evaluating is arranged to examine only a segment of the spectral representation from the smallest frequency to a limit frequency with respect to said criterion.
5. A device according to any of the preceding claims,
   wherein the coding algorithm is one of a plurality of different coding algorithms, and wherein said means for performing said conversion step further comprises:

   a memory means for storing a set of coding parameters of its own for each coding algorithm, said set of coding parameters being selected to define at least the conversion step of the corresponding coding algorithm; and

   means for retrieving another set of coding parameters from said memory means in order to provide evaluation results for an additional coding algorithm.
6. A device according to claim 5,
   wherein said set of coding parameters for a coding algorithm defines a filter bank underlying the same as well as a window used by the same for coding block formation.
7. A device according to any of the preceding claims,
   wherein the decoded signal is a stereo signal and wherein said device further comprises:

   means for stereo processing the decoded signal in order to provide at least one processed stereo signal.
8. A device according to claim 7,
   wherein said means for stereo processing performs mid/side processing such that the means for converting (12) acts at least on a mid signal or on a side signal.
9. A device according to any of the preceding claims,
   further comprising:

   a write means coupled to said means (14) for searching and outputting, in order to provide the decoded signal with a mark comprising at least coding block raster information.
10. A device according to any of the preceding claims which is arranged to process as decoded signal an audio signal or a video signal, wherein the data reducing step, in case of the audio signal, comprises a quantization depending on a psychoacoustic model and, in case of a video signal, comprises a quantization depending on a psychooptic model.
11. A method for determining a coding block raster on which a decoded signal is based, in which the decoded signal is produced from an original signal by coding and decoding according to a coding algorithm including a coding block generating step, a conversion step and a data reducing step, said coding block generating step of the coding algorithm including partitioning the original signal according to the coding block raster into coding blocks with a specific number of time-discrete signal values, said conversion step including generating from a coding block a spectral representation of the same, and said data reducing step including removing information from the spectral representation of the original signal, said method comprising the steps of:

    picking out (11) a segment of the decoded signal, said segment beginning at an output sampling value of the decoded signal;

    performing (12) the conversion step on said segment of the decoded signal so as to provide a spectral representation of said segment;

    evaluating (13) the spectral representation of said segment with respect to a predetermined criterion in order to obtain an evaluation result for the segment,

    said steps of picking out (11), performing (12) and evaluating (13) being carried out a plurality of times in order to pick out, convert and evaluate a plurality of segments of the decoded signal that begin at different output sampling values in order to obtain a plurality of evaluation results; and

    searching (14) the evaluation results and outputting an identification for the coding block raster underlying the decoded signal, on the basis of the segment that has an extreme evaluation result with respect to other evaluation results.

Images