EP1371055B1 - Device for the analysis of an audio signal with regard to the rhythm information in the audio signal using an auto-correlation function - Google Patents

Abstract

An apparatus for analyzing an audio signal with regard to rhythm information of the audio signal by using an autocorrelation function comprises a filter bank for separating the audio signal into at least two sub-band signals. The sub-band signals are examined with regard to periodicities by an autocorrelation function, to obtain rhythm raw-information for the at least two sub-band signals. To reduce or eliminate the ambiguities of the autocorrelation function for periodical signals, the rhythm raw-information is postprocessed to obtain post-processed rhythm raw-information for the sub-band signal. The rhythm information of the audio signal is established based on the postprocessed rhythm raw-information. By the sub-band-wise ACF postprocessing, ACF ambiguities are already eliminated where they originate, and rhythm portions are added at double tempi, which an autocorrelation function processing does normally not provide, so that, as a result, a more robust determination of the rhythm information of the audio signal arises.

Description

The present invention relates to signal processing concepts and in particular to the analysis of audio signals in terms of of rhythm information.

In recent years, the availability of multimedia data, such as As audio or video data, has risen sharply. This is due to a number of technical factors in particular the broad availability of the Internet, of powerful computer hardware and software as well as of efficient methods for data compression, d. H. Source coding, of audio and video procedures.

The huge amounts of audiovisual data, for example, on available on the Internet, demand concepts that which make it possible to judge these data according to content criteria, catalog, etc. to be able to. There is a desire to be able to target multimedia data by specifying to search for and find meaningful criteria.

This requires so-called "content-based" techniques the audiovisual data so-called characteristics, which in professional circles Also referred to as "features", which extract important ones represent characteristic properties of the signal. Based on such features or combinations of these features can similarity relationships or similarities between Audio or video signals are derived. This process is done by comparing or relating the extracted Feature values from the different signals, which Also simply referred to as "pieces".

Of particular interest is the determination or extraction of Characteristics that are not just signal theory, but possible have immediate semantic meaning, d. H. directly from the listener represent perceived properties.

This allows the user to easily and intuitively To formulate search queries to pieces from the entire existing To find the database of an audio signal database. Similarly, semantically relevant features allow similarity relationships to model between pieces that are human Get close to sensation. The use of features that semantic meaning, also allows for example a automatic suggestion of interesting for a particular user Pieces, if his preferences are known.

In the field of music analysis, the tempo is an important musical one Parameter that has semantic meaning. The pace will be usually measured in beats per minute (BPM). The automatic Extraction of the tempo and the center of gravity of the "Beats" or generally the automatic extraction of Rhythm information, is an example of winning one semantically important feature of a piece of music.

Furthermore, there is a desire that the feature extraction, d. H. extracting rhythm information an audio signal, can be robust and computationally efficient. Robustness means that it does not matter if that is Piece has been source coded and decoded again, whether the piece played over a loudspeaker and received by a microphone whether it is from an instrument or an instrument Plural of instruments is played.

For the determination of the center of gravity and thus of the tempo, d. H. for the determination of rhythm information the term "beat tracking" has also become established in the professional circles. It is already known from the prior art, a Beat tracking on the basis of a note-like or transcribed Signal representation, z. B. in midi format to perform. The goal, however, is not to require such metadata but an analysis directly with a z. B. PCM coded or generally speaking to make digitally present audio signal.

The specialist publication "Tempo and Beat Analysis of Acoustic Musical Signals "by Eric D. Scheirer, J. Acoust. 103: 1, (Jan 1998), pages 588-601, discloses a method for automatic extraction of a rhythmic pulse from musical Excerpts. The input signal is by means of a filter bank split into a number of subbands, for example in 6 subbands with crossover frequencies of 200 Hz, 400 Hz, 800 Hz, 1600 Hz and 3200 Hz. For the first subband, low pass filtering is used carried out. For the last subband is a High pass filtering is done, for the rest, in between bandpass filtering is described. each Subband is processed as follows. The subband signal is first rectified. In other words the absolute value of the samples is determined. The resulting n values are then smoothed, for example with a Averaging over a suitable window to get an envelope signal to obtain. To reduce the computational complexity, the envelope signal be subsampled. The envelope signals will be differentiated, d. H. sudden changes in signal amplitude are preferably forwarded by the differentiation filter. The result is then limited to non-negative values. Each envelope signal is then converted into a bank of resonant filters, d. H. Oscillators, given, each one filter for each Tempo range so that matches the musical tempo Filter is most excited. For each filter is the energy of the output signal as a measure of the match the tempo of the input signal at the tempo associated with the filter calculated. The energies for each tempo eventually become summed over all subbands, with the largest amount of energy the tempo delivered as a result, d. H. the rhythm information, features. Unlike autocorrelation methods is found to be advantageous that the oscillator bank also with output signals at the double, triple, etc. of the Tempos or even at rational multiples (eg 2/3, 4/3) of the Tempos reacts to a stimulus. An autocorrelation function does not have this property, it only provides output signals at the halved, thirded, etc. pace.

A major disadvantage of this method is the large Computing and storage complexity especially for realization the large number of parallel vibrating "oscillators", of which ultimately only one is selected. This makes an efficient implementation, for example, for real-time applications almost impossible.

The specialist publication "Pulse Tracking with a Pitch Tracker" by Eric D. Scheirer, Proc. 1997 Workshop on Applications of Signal Processing to Audio and Acoustics, Mohonk, NY, Oct 1997, describes a comparison of the "oscillator concept" described above with an alternative concept that is on the use of autocorrelation functions for extraction of periodicity from an audio signal, d. H. the rhythm information of a signal, builds. An algorithm for modeling the human Pitch perception, d. H. the pitch, for the "beat Tracking "used.

The known algorithm is shown in Fig. 3 as a block diagram shown. The audio signal is sent via an audio input 300 an analysis filter bank 302 supplied. The analysis filter bank generates from the audio input a number n of channels, i. H. of individual subband signals. Each subband signal contains a certain range of frequencies of the audio signal. The Analysis Filter Bank filters are selected to match the Approximate the selection characteristic of the human inner ear. Such an analysis filter bank is also called a gamma-tone filter bank designated.

In the devices 304a to 304c, the rhythm information becomes each subband signal evaluated. For every input signal will initially be an envelope-like output signal calculated (according to a so-called "Inner Hair Cell" processing in the ear) and subsampled. Out of this result An autocorrelation function (AKF) is calculated to determine the periodicity the signal as a function of the delay, d. H. of To get "lag".

At the output of devices 304a-304c is then for each Subband signal before an autocorrelation function, which the Represents rhythm information of each subband signal.

The individual autocorrelation functions of the subband signals are then combined in a means 306 by summation, to obtain a sum auto-correlation function (SAKF) which Aspects of the rhythm information of the signal at the audio input 300 plays. This information may be at a tempo output 308 are issued. Great values in the sum auto-correlation indicate that for a peak assigned to the SAKF Delay (lag) a high periodicity of note beginnings is present. Therefore, for example, the largest value of Sum auto correlation function within the musically meaningful Delays searched.

Musically meaningful delays are, for example, the tempo range between 60 bpm and 200 bpm. The device 306 may be further arranged to provide a delay time in tempo information implement. For example, a peak corresponds a delay of one second, a tempo of 60 beats per minute. Minor delays indicate higher tempo while larger delays occur at slower speeds than 60 bpm clues.

This method has over the former method an advantage in that no oscillators with large Computing and memory costs must be implemented. on the other hand the concept is disadvantageous in that the quality The results depend very much on the type of audio signal. Is an audio signal, for example, a dominant To hear out rhythm instrument, so that described in Fig. 3 Concept work well. If, on the other hand, the voice is dominant, which are not very clear rhythm information will provide, so the rhythm determination will be ambiguous. In The audio signal could well be a band that only Contains rhythm information, such as B. a higher Frequency band in which, for example, a hihat a drums is positioned, or a low frequency band in which the big drum of a drum set on the frequency scale is. Due to the combination of individual information however, the reasonably clear information of this special subbands of the ambiguous information of others Subbands superimposed or "diluted".

Another problem with using autocorrelation functions for extracting the periodicity of a subband signal is that the sum autocorrelation function, obtained by means 306 is ambiguous. The Sum autocorrelation function at output 306 is to this effect ambiguous that even with the multiple of a delay an autocorrelation function peak is produced. This is out of it understandable that a sine component with a period of t0 when subjected to autocorrelation function processing becomes, beside the desired maximum at t0 also maxima Multiples of the delays, d. H. at 2t0, 3t0, etc. generated.

The technical publication "A Computationally Efficient Multipitch Analysis Model ", by Tolonen and Karjalainen, IEEE Transactions on Speech and Audio Processing, Vol. 8, No. 6, Nov. 2000 a time-efficient model for a periodicity analysis of complex audio signals. The calculation model shares the signal in two channels, in a channel below 1000 Hz and a channel over 1000 Hz. This will result in an autocorrelation of the lower channel and an autocorrelation of the envelope of the the upper channel. Finally, the two autocorrelation functions summed. The ambiguities of the sum auto-correlation function eliminates the sum auto-correlation function further processed to a so-called Enhanced Summary Autocorrelation Function (ESACF) (Further Development Sum auto-correlation function). This post-processing The sum auto-correlation function includes a repeated one Subtracting from splayed with integer factors Versions of the autocorrelation function from the sum autocorrelation function with subsequent limitation to non-negative Values.

The disadvantage of this concept is the fact that through the autocorrelation functions in the subbands per subband obtained ambiguities only in the cumulative autocorrelation function eliminated, but not directly where they occur, namely in the individual subbands.

Another disadvantage of this concept is the fact that the Autocorrelation function in itself no indication of the double, triple, ... of the tempo that gives an autocorrelation peak assigned.

The object of the present invention is to provide a Apparatus and method for analyzing an audio signal in terms of rhythm information using a To create autocorrelation function that is robust and computing time efficient.

This object is achieved by a device for analyzing a Audio signal according to claim 1 or by a Method for analyzing an audio signal according to claim 7 solved.

The present invention is based on the finding that a post-processing of an autocorrelation function in a partial band can be performed to the ambiguities of the autocorrelation function for periodic signals to eliminate or Tempo information that does not autocorrelation processing provides the information obtained by an autocorrelation function to be added. According to one aspect of the present Invention will be an autocorrelation function post-processing the subband signals already used the ambiguities eliminate "at the root", or "missing" rhythm information to add.

According to another aspect of the present invention is a Post-processing of the sum auto-correlation function performed, re-edited rhythm raw information for the audio signal so that in the reworked rhythm raw information a signal component at an integer fraction a delay that has an autocorrelation function peak is assigned, is added. This makes it possible to rhythm information not obtained by an autocorrelation function at double, triple etc. tempos or at rational Multiply by calculating by an integer factor or by a rational factor compressed versions of the Autocorrelation function and by adding these versions to the to produce the original autocorrelation function. In contrast to the prior art, in which for this purpose a complex oscillator bank is required, this is done according to the invention with easy to implement weighting and addition routines.

According to another aspect of the present invention, the Sum auto-correlation function also post-processed by a with a factor greater than zero and less than one, weighted to an integer factor greater than one splayed Version of the raw rhythm information about the autocorrelation function is subtracted. This has the advantage of a Elimination of AKF ambiguities in integer multiples the delay associated with an autocorrelation peak is. While in the prior art no weighting of the splayed Versions of the autocorrelation function before subtraction carried out, and an elimination of ambiguities thus only achieved in the theoretically optimal case, in which the rhythm repeats itself ideally cyclically, the weighted subtraction the ability to choose by appropriate choice Weighting factors, which can be done empirically, for example, Rhythm information that does not repeat itself ideally cyclically, to take into account.

According to a preferred embodiment of the present invention becomes an autocorrelation function post-processing performed by means of an autocorrelation function certain rhythm raw information with compressed and / or spread versions of the same can be combined. In case of Use of spread versions of the rhythm raw information become the splayed versions of the rhythm raw information subtracted while in the case of integer Factors compressed versions of the autocorrelation function these compressed versions to the rhythm raw information be added.

In a preferred embodiment of the invention, the compressed / spread version before adding or subtracting weighted by a factor between zero and one.

According to another preferred embodiment of the present invention Invention is a quality assessment of the rhythm raw information, to get a significance measure, on the basis the postprocessed rhythm raw information performed, such that the quality rating is no longer due to autocorrelation function artifacts being affected. This will be a safe quality assessment possible, reducing the robustness determining rhythm information of the audio signal can be further increased.

Alternatively, the quality assessment can be carried out before the AKF post-processing occur. This has the advantage that if a flat course of the rhythm raw information found is, i. no pronounced rhythm information on the AKF post-processing for this subband signal can be dispensed with can, because this subband due to its little meaningful Rhythm information when determining the rhythm information the audio signal will not matter anyway. In this manner and way, the computational and memory overhead can be further reduced become.

In the individual frequency bands, d. H. the subbands, lie often different favorable conditions for finding of rhythmic periodicities. While, for example, at Pop music often in the middle, for example around 1 kHz, the signal is dominated by vocals not corresponding to the beat is often present in the higher frequency ranges all percussion sounds present, such. B. the hihat of the drums, which is a very good extraction of rhythmic regularities allow. In other words, different ones Frequency bands depending on the audio signal a different amount in rhythmic information or have a different Quality or significance for the rhythm information of the audio signal.

The audio signal is therefore first decomposed into subband signals. Each subband signal is examined for its periodicity, around raw rhythm information for each subband signal to obtain. This is according to a preferred embodiment The present invention provides an evaluation of quality the periodicity of each subband signal is performed by a significance measure for each subband signal. A high degree of significance indicates that in this subband signal clear rhythm information is available, while a low Significance indicates that in this subband signal less clear rhythm information is available.

According to a preferred embodiment of the present invention is considered when examining a subband signal Its periodicities initially a modified envelope of the subband signal and then an autocorrelation function the envelope is calculated. The autocorrelation function the envelope represents the raw rhythm information. Unique Rhythm information is present when the autocorrelation function has distinct maxima, while less definite Rhythm information is present when the autocorrelation function the envelope of the subband signal less pronounced Signal peaks or no signal peaks at all. An autocorrelation function, which has significant signal peaks, therefore receive a high degree of significance while an autocorrelation function, which has a relatively flat course, a low one Significance is obtained. The artifacts of autocorrelation functions are, as stated above, according to the invention eliminated.

The individual rhythm raw information of the individual subband signals So they are not simply combined "blindly" but rather taking into account the significance measure for each subband signal used to get the rhythm information of the audio signal to obtain. If a subband signal has a high degree of significance, then it is preferred in determining the rhythm information, while a subband signal, that is a low significance measure has, d. H. this is a low quality in terms of rhythm information in determining the rhythm information the audio signal hardly or in extreme cases not at all is taken into account.

This can be computationally good by a weighting factor be implemented, which depends on the significance measure. While a subband signal that is good quality for the rhythm information has, d. H. which has a high degree of significance, one Weighting factor of 1 will get another Subband signal having a smaller significance measure, a weighting factor less than 1 received. In extreme cases, a Subband signal, which is a completely flat autocorrelation function has to have a weighting factor of 0. The weighted Autocorrelation functions, d. H. the weighted rhythm raw information are then simply added up. If only a subband signal of all subband signals good rhythm information while the other subband signals provide autocorrelation functions can have a flat course this weighting in extreme cases cause all subband signals except for the one subband signal, a weighting factor received from 0, d. H. in determining the rhythm information not be considered at all, so the rhythm information the audio signal only from a single Subband signal can be determined.

The inventive concept is advantageous in that it allows a robust determination of the rhythm information because subband signals with no unique or even different Rhythm information, d. H. if the song is another Has rhythm as the actual beat of the piece, the Rhythm information of the audio signal does not "dilute" or "distort". In addition, very noisy subband signals, which will complete a system autocorrelation function provide a flat waveform, the signal-to-noise ratio in determining the rhythm information does not worsen. Exactly this would happen, however, if, as in the state of Technique, just all the autocorrelation functions of the subband signals be summed up with the same weight.

Another advantage of the method described is that that with a small additional computational effort a significance measure can be determined, and that the evaluation of the rhythm raw information with the significance measure and the subsequent Summation without large storage and computational time efficient can be performed, which is the concept of the invention especially recommended for real-time applications.

Fig. 1

ein Blockschaltbild einer Vorrichtung zum Analysieren eines Audiosignals mit einer Qualitätsbewertung der Rhythmus-Rohinformationen;

Fig. 2

ein Blockschaltbild einer Vorrichtung zum Analysieren eines Audiosignals unter Verwendung von Gewichtungsfaktoren auf der Basis der Signifikanzmaße;

Fig. 3

ein Blockschaltbild einer bekannten Vorrichtung zum Analysieren eines Audiosignals hinsichtlich von Rhythmusinformationen;

Fig. 4

ein Blockschaltbild einer Vorrichtung zum Analysieren eines Audiosignals hinsichtlich von Rhythmusinformationen unter Verwendung einer Autokorrelationsfunktion mit einer teilbandweisen Nachbearbeitung der Rhythmus-Rohinformationen; und

Fig. 5

ein detailliertes Blockschaltbild der Einrichtung zum Nachbearbeiten von Fig. 4.

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

Fig. 1

a block diagram of an apparatus for analyzing an audio signal with a quality evaluation of the rhythm raw information;

Fig. 2

a block diagram of an apparatus for analyzing an audio signal using weighting factors based on the significance measures;

Fig. 3

a block diagram of a known device for analyzing an audio signal in terms of rhythm information;

Fig. 4

a block diagram of an apparatus for analyzing an audio signal with respect to rhythm information using an autocorrelation function with a partial bandwise post-processing of the rhythm raw information; and

Fig. 5

a detailed block diagram of the device for reprocessing of Fig. 4.

Fig. 1 shows a block diagram of an apparatus for analyzing an audio signal regarding rhythm information. The audio signal is passed through an input 100 of a device 102 for decomposing the audio signal into at least two subband signals 104a and 104b supplied. Each subband signal 104a, 104b is sent to a device 106a or 106b for examining it in terms of periodicities in the subband signal, around raw rhythm information 108a and 108b for each To obtain subband signal. The rhythm raw information will be then a device 110a or 110b for rating a quality the periodicity of each of the at least two subband signals supplied to a significance measure 112a, 112b for each of the at least to obtain two subband signals. Both the rhythm raw information 108a, 108b as well as the significance measures 112a, 112b are a means 114 for determining the rhythm information supplied to the audio signal. The device 114 takes into account in determining the rhythm information of the audio signal the significance measures 112a, 112b for the subband signals and the rhythm raw information 108a, 108b of at least one Sub-band signal.

For example, the device 110a has a quality evaluation found that in the subband signal 104a no special Periodicity is present, then the significance measure 112a becomes very be small, or equal to 0. In this case, the set up 114 for determining the rhythm information that the significance measure 112a is equal to zero, so that the rhythm raw information 108a of the subband signal 104a in the determination the rhythm information of the audio signal not at all more must be considered. The rhythm information of the Audio signals are then solely and exclusively based the rhythm raw information 108b of the subband signal 104b.

In the following, reference is made to FIG. 2 for a specific one Embodiment of the device of Fig. 1 received. As a device 102 for decomposing the audio signal may be a usual Analysis filter bank, the output one of provides a user selectable number of subband signals. each Subband signal is then processed by the facilities 106a, 106b and 106c, respectively, and then through the facilities 110a to 110c of each rhythm raw information significance measures be determined. The device 114 includes at the illustrated in Fig. 2 preferred embodiment a Means 114a for calculating weighting factors for each Subband signal based on the significance measure for this Subband signal and optionally also the other subband signals. In The device 114b then finds a weighting of the rhythm raw information 108a to 108c with the weighting factor for this subband signal instead, whereupon then, also in the Means 114b combining weighted rhythm raw information, z. B. summed up to be at the tempo output 116th to get the rhythm information of the audio signal.

The inventive concept thus arises as follows After the evaluation of the rhythmic information of the Individual bands, which are produced, for example, by enveloping, Smoothing, differentiating, limiting to positive values and forming the autocorrelation function can take place (facilities 106a to 106c), finds a rating of the value or the Quality of these intermediate results in the facilities 110a to 110c instead. This is achieved by means of an evaluation function, which the reliability of the individual results valued with a significance measure. From the significance measures All subband signals become a weighting factor for each Volume derived for the extraction of rhythm information. The overall result of the rhythm extraction will then be in the facility 114b by combining the bandwise individual results achieved taking into account their respective weighting factors.

As a result, such an algorithm implemented for rhythm analysis a good ability to rhythmic information reliable in a signal even under unfavorable conditions to find. The concept according to the invention is therefore distinguished through a high robustness.

In a preferred embodiment, the rhythm raw information becomes 108a, 108b, 108c, the periodicity of the respective Subband signal represent, by means of an autocorrelation function certainly. In this case, it is preferable that Significance measure to determine by a maximum of the autocorrelation function by an average of the autocorrelation function is divided, and then the value 1 is subtracted. It was It should be noted that any autocorrelation function is always included a delay of 0 is a local maximum, i. H. a peak, which represents the energy of the signal. This local Maximum should be disregarded, so that the quality determination is not distorted.

Furthermore, the autocorrelation function is intended only in a special Tempo range are considered, d. H. from a maximum Delay corresponding to the smallest interest rate of interest, to a minimum delay, which is the highest one of interest Tempo corresponds. A typical tempo area lies between 60 bpm and 200 bpm.

Alternatively, as a measure of significance, the ratio between the arithmetic mean of the autocorrelation function in the interest Tempo range and the geometric mean of the Autocorrelation function determined in the tempo of interest become. It is known that if all values of the autocorrelation function are the same, d. H. if the autocorrelation function has a flat course, the geometric mean the autocorrelation function and the arithmetic mean the autocorrelation function are the same. In that case would have the significance measure has a value equal to 1, which means that the Rhythm raw information is not significant.

In the case of a system autocorrelation function with strong peaks would the ratio of arithmetic mean to geometric Mean value will be greater than 1, which means that the autocorrelation function has good rhythm information. ever smaller, however, is the ratio between arithmetic mean and geometric mean, the flatter is the autocorrelation function and the less periodicities it contains they, which in turn means that the rhythm information of this Subband signal less significant, d. H. a lower one Quality, resulting in a low or a weighting factor of 0 will express.

With regard to the weighting factors, there are various possibilities. Preferred is a relative weight, such that all weighting factors of all subband signals to Add 1, d. H. that determines the weighting factor of a band is divided as the significance value of this band by the sum of all significance values. In this case, a relative Weighting before the summation of the weighted rhythm raw information performed to the rhythm information of the To get audio signal.

As has already been stated, it is preferred to carry out the evaluation the rhythm information using an autocorrelation function perform. This case is in FIG. 4 shown. The audio signal is sent via the audio signal input 100 in the device 102 for decomposing the audio signal in Subband signals 104a and 104b are fed. Each subband signal is then in the device 106a or 106b, as stated investigated using an autocorrelation function, to determine the periodicity of the subband signal. The rhythm raw information then lies at the output of the device 106a or 106b 108a, 108b. These are in a facility 118a and 118b, respectively, by the autocorrelation function rhythm raw information output from the device 116a rework. This will u. a. ensured that the ambiguities of the autocorrelation function, d. H. that at integer multiples of the delays as well Signal peaks occur, are eliminated on a band-by-band basis, to post-processed rhythm raw information 120a and 120b, respectively receive.

This has the advantage that the ambiguities of the autocorrelation functions, d. H. the rhythm raw information 108a, 108b, already be eliminated partially bandwise, and not first, as in Prior art, according to the summation of the individual autocorrelation functions. In addition, the single band allows Elimination of ambiguities in autocorrelation functions by means 118a, 118b that the rhythm raw information the subband signals are handled independently of each other can be. You can, for example, a quality assessment by means of the device 110a for the rhythm raw information 108a or by means 110b for the Rhythm raw information 108b.

As shown by the dashed lines in FIG. 4, However, the quality assessment can also be based on the post-processed Rhythm raw information take place, these the latter option is preferred since the quality assessment on the basis of the reworked raw rhythm information, that the quality of an information is assessed which is no longer ambiguous.

The determination of the rhythm information by the device 114 then takes place on the basis of postprocessed rhythm information a channel and preferably also on the base the significance measure for this channel.

If a quality assessment based on the rhythm raw information, that is, the signal before the device 118a is performed, this is advantageous in that if it is determined that the significance measure is 0, d. H. that the autocorrelation function has a flat course on the Post-processing by means 118a completely omitted to save computing time resources.

In the following, reference is made to Fig. 5 to a more detailed Construction of a device 118a or 118b for post-processing to represent the rhythm raw information. First For example, the subband signal becomes 104a into the device 106a for examining the periodicity of the subband signal by means of an autocorrelation function fed to raw rhythm information To receive 108a. To the ambiguities Teilbandweise to eliminate, just as in the prior art, a spread autocorrelation function by means of a device 121, wherein the device 121 is arranged is to calculate the spread autocorrelation function so that it is spread by an integer multiple. An institution 122 is arranged in this case to the spread Autocorrelation function from the original autocorrelation function, d. H. subtract the rhythm raw information 108a. In particular, it is preferred, first on the Double spread autocorrelation function in the facility 121 and then from the raw rhythm information 108a to subtract. Then, in the next step, one around the Factor 3 spread autocorrelation function in the device 121 and from the result of the previous subtraction subtracted again, so that gradually all ambiguities be eliminated from the rhythm raw information.

In addition, the splayed versions of the rhythm raw information 108a before subtracting to be weighted Here, too, a flexibility in the sense of a high degree of robustness to reach.

By the method, the periodicity of a subband signal can examine the basis of an autocorrelation function So a further improvement can be achieved if the properties the autocorrelation function and are involved the post-processing using means 118a or 118b is performed. So creates a periodic sequence of Grade starts with a distance t0 not just an AKF peak at a delay t0 but also at 2t0, 3t0, etc. This will to an ambiguity in tempo detection, d. H. the search significant maxima in the autocorrelation function. The ambiguities can thereby be eliminated when around integer factors splayed versions of the AKF from baseline deducted part-bandwise (weighted).

In addition, the compressed versions of the rhythm raw information 108a before adding with a factor unequal One to be weighted, here too flexibility in the sense to achieve a high degree of robustness.

Further, in the autocorrelation function, there is the problem that they have no information at t0 / 2, t0 / 3 ... etc., that is double, Triple, etc. of the "basic tempo" delivers what special can lead to wrong results if two instruments, which lie in different subbands, together the Defining the rhythm of the signal. This thing is taken into account that edged by integer factors the autocorrelation function are calculated and then to weighted or unweighted added to the rhythm raw information become.

The AKF post-processing thus takes place on a part-band basis, with for at least one subband signal, an autocorrelation function is calculated and this with stretched or spread versions this feature is combined.

According to another aspect of the present invention, first generates the sum autocorrelation function of the subbands, whereupon compressed by integer factors versions the sum auto-correlation function is added preferably weighted be to the shortcomings of the autocorrelation function at the double, triple, etc. pace clear.

In another aspect, the post-processing of the sum auto-correlation function, to the ambiguities at the half, to eliminate the third part, the fourth part etc. of the tempo, carried out by the spread by integer factors Versions of the sum auto correlation function not be subtracted just before subtracting with a Factor not equal to one and preferably less than one and be weighted greater than zero and then subtracted. This will make a more robust determination of rhythm information possible because the unweighted subtracting only for ideal sinusoidal signals complete elimination of the AKF ambiguity supplies.

Claims (7)

1. Apparatus for analyzing an audio signal with regard to rhythm information of the audio signal by using an autocorrelation function, comprising:

   means (102) for dividing the audio signal into at least two sub-band signals (104a, 104b);

   means for examining (106a, 106b) at least one sub-band signal with regard to a periodicity in the at least one sub-band signal by an autocorrelation function, to obtain rhythm raw-information (108a) for the sub-band signal, wherein a delay is associated to a peak of the autocorrelation function;

   means (118a) for postprocessing the rhythm raw-information (108a) for the sub-band signal (104a) determined by the autocorrelation function, to obtain postprocessed rhythm raw-information (120a) for the sub-band signal, so that in the postprocessed rhythm raw-information an ambiguity in an integer plurality of a delay, to which an autocorrelation function peak is associated, is reduced, or a signal portion is added at an integer fraction of a delay, to which an autocorrelation function peak is associated; and

   means (114) for establishing the rhythm information of the audio signal by using the postprocessed rhythm raw-information (120a) of the sub-band signal and by using another sub-band signal of the at least two sub-band signals.
2. Apparatus according to claim 1, wherein the means for postprocessing (118a, 118b) comprises:

   means (121) for calculating a version of the rhythm raw-information (108a) of a sub-band signal spread by an integer factor; and

   means (122) for subtracting the version of the rhythm raw-information (108a) of the sub-band signal spread by an integer factor larger than one, or a version of the rhythm raw-information (108a) of the sub-band signal derived from this version, to obtain the postprocessed rhythm raw-information (120a) for the sub-band signal.
3. Apparatus according to claim 2, wherein means (122) for subtracting is disposed to perform, prior to subtracting, a weighting of the spread version with a factor between zero and one, to generate the derived version.
4. Apparatus according to claim 1, wherein means for postprocessing (118a) comprises:

   means (121) for calculating a version of the rhythm raw-information (108a) compressed by an integer factor larger than one; and

   means (122) for adding the compressed version of the rhythm raw-information of the sub-band signal or a version derived therefrom to the rhythm raw-information (108a) of the sub-band signal, to obtain the postprocessed rhythm raw-information (120a) for the sub-band signal.
5. Apparatus according to claim 4, wherein the means (122) for adding is disposed to perform, prior to adding, a weighting of the compressed version of the rhythm raw-information by a factor between zero and one, such that a weighted compressed version of the rhythm raw-information is added to the rhythm raw-information of the sub-band signal to generate the derived version.
6. Apparatus according to one of the previous claims, further comprising:

   means (110a, 110b) for evaluating a quality of the periodicity of the postprocessed rhythm raw-information (120a), to obtain a significance measure for the sub-band signal,

   wherein means (114) for establishing is further disposed to establish the rhythm information of the audio signal by considering the significance measure of the sub-band signal.
7. Method for analyzing an audio signal with regard to rhythm information of the audio signal by using an autocorrelation function, comprising:

   dividing (102) the audio signal into at least two sub-band signals (104a, 104b),

   examining (106a, 106b) at least one sub-band signal with regard to a periodicity in the at least one sub-band signal by an autocorrelation function, to obtain rhythm raw-information (108a) for the sub-band signal, wherein a delay is associated to a peak of the autocorrelation function;

   postprocessing (118a) the rhythm raw-information (108a) for the sub-band signal (104a) determined by the autocorrelation function, to obtain postprocessed rhythm raw-information (120a) for the sub-band signal, so that in the postprocessed rhythm raw-information an ambiguity in the integer plurality of a delay, to which an autocorrelation function peak is associated, is reduced, or a signal portion is added at an integer fraction of a delay, to which an autocorrelation function peak is associated; and

   establishing (114) the rhythm information of the audio signal by using the postprocessed rhythm raw-information (120a) of the sub-band signal and by using a further sub-band signal of the at least two sub-band signals.

Images