EP1377924B1 - Method and device for extracting a signal identifier, method and device for creating a database from signal identifiers and method and device for referencing a search time signal - Google Patents

Abstract

In a method of extracting a signal identifier from a time signal, the temporal occurrence of signal edges in the time signal is detected (12), wherein a signal edge has a specified temporal length. In addition, the temporal interval between two selected detected signal edges is determined (14). From the temporal interval determined, a frequency value is calculated (16), the frequency value being associated with a time of occurrence of the frequency value in the time signal so as to obtain a coordinate tuple from the frequency value and the time of occurrence for this frequency value. A signal identifier is created from a plurality of coordinate tuples (18), each coordinate tuple including a frequency value and a time of occurrence, which is why the signal identifier includes a sequence of signal identifier values reproducing the temporal form of the time signal. The extracted signal identifier is based on signal edges of the time signal and thus reproduces the temporal form of the time signal. The signal identifier is therefore characteristic of the time signal, on the one hand, and robust towards changes in the time signal, on the other hand.

Description

The present invention relates to processing time signals that have a harmonic component, and in particular on the generation of a signal identifier for a Time signal to get the time signal using a database in of a plurality of signal identifiers for a plurality of Time signals is stored to be able to describe.

Concepts by which time signals with a harmonic component, such as B. audio data, identifiable and referenced are useful for many users. In particular in a situation where there is an audio signal, whose title and author are unknown, it is often desirable find out from whom the corresponding song comes. There is a need for this, for example, if the Wish is present, e.g. B. a CD of the artist in question to acquire. If the present audio signal only includes the time signal content, but no name above the interpreter, the music publisher etc. is an identification the origin of the audio signal or by whom Song comes from, not possible. The only hope was then in it, the audio piece including reference data regarding the Author or the source where the audio signal can be purchased, heard again to get the title you want to be able to.

On the internet it is not possible to use audio data conventional search engines to look for since the search engines can only deal with textual data. Audio signals or more generally, time signals that one can have a harmonious share through such search engines not be processed if they are not textual Include search information.

A realistic inventory of audio files is several thousand saved audio files up to hundreds of thousands of audio files. Music database information can be stored on a central Internet server, and potential Search queries could be made over the Internet. Alternatively, with today's hard drive capacities the central music databases on local hard disk systems conceivable by users. It is desirable to have such music databases to be able to search for reference data via to learn an audio file, of which only the file itself, but no reference data are known.

In addition, it is equally desirable to have music databases search using predetermined criteria to be able to, for example, similar To be able to find out pieces. Similar pieces are for example, the pieces with a similar melody, one similar set of instruments, or simply with similar ones Noises such as B. sound of the sea, twittering of birds, male voices, female voices, etc.

U.S. Patent No. 5,918,223 discloses a method and a device for content-based analysis, storage, Recovery and segmentation of audio information. This procedure relies on several acoustic Extract features from an audio signal. Measured are volume, bass, pitch, brightness and melency-based Cepstral coefficients in a time window certain length in periodic intervals. Everyone Measurement data set consists of a sequence of measured feature vectors. Each audio file is through the complete sentence the characteristic sequences calculated for each characteristic. Furthermore, the first derivatives for each sequence of Feature vectors calculated. Then statistical values how mean and standard deviation are calculated. This Set of values is in an N vector, i.e. H. a vector with N elements, saved. This procedure is based on a variety of audio files applied to each audio file derive an N vector. This will gradually according to a database made up of a large number of N vectors. An unknown audio file then becomes Extract a search n vector using the same procedure. In the case of a search query, a distance calculation is then made of the given N vector and that in the database stored N vectors determined. Eventually the N vector is output, which is the minimum distance to that Search N vector. The output N vector is data assigned via the author, title, source of supply, etc. so an audio file regarding its origin can be identified.

This method has the disadvantage that several characteristics are calculated and arbitrary heuristics to calculate of the parameters are introduced. By mean and Standard deviation calculations across all feature vectors for an entire audio file, the information generated by the time course of the feature vectors is given, reduced to a few feature sizes. This leads to one high loss of information.

The object of the present invention is a Method and device for extracting a signal identifier to create a time signal that is meaningful Identification of a time signal without too large Enable loss of information.

This task is accomplished by a method of extracting a Signal identifier from a time signal according to claim 1 or by a device for extracting a signal identifier solved from a time signal according to claim 19.

Another object of the present invention is therein, a method and an apparatus for generating a Database of signal identifiers and a method and a Device for referencing a search time signal by means of to create such a database.

This task is accomplished by a method for generating a Database according to claim 13, a device for generating a database according to claim 20, a method for referencing a search time signal according to claim 14 or a device for referencing a Search time signal according to claim 21 solved.

The present invention is based on the finding that that with time signals that have a harmonic component, the time course of the time signal can be used can to a signal identifier of the time signal from the time signal to extract the one hand a good fingerprint for the time signal, but on the other hand is manageable in terms of their amount of data in order to be an efficient one Search a variety of signal identifiers in to enable a database. An essential quality of time signals with a harmonic component recurring signal edges in the time signal, z. B. two successive signal edges with the same or Similar length, the specification of a period and thus a frequency in the time signal with high temporal and enable frequency resolution, if not only that Presence of the signal edges per se, but also that temporal occurrence of the signal edges in the time signal is taken into account becomes. It is therefore possible to provide a description to obtain the time signal in that the time signal is off consecutive frequencies. Exemplary of an audio signal, the audio signal is thus characterized that a tone, that is, a frequency, at a certain point Time is present, and that this sound, i.e. H. this frequency, another one later Sound, d. H. another frequency follows.

According to the invention, the description of the time signal by a sequence of temporal samples into one Description of the time signal using coordinate tuples Frequency and time of occurrence of the frequency passed. The signal identifier or, in other words, the feature vector (MV) used to describe the time signal thus comprises a sequence of signal identification values, the more or less roughly depending on the embodiment reproduces the time course of the time signal. The time signal is therefore not based on, as in the prior art characterized its spectral properties, but based on the temporal sequence of frequencies in the time signal.

To calculate a frequency value from the detected signal edges are at least two detected signal edges needed. The selection of these two signal edges the total detected signal edges, on the basis thereof Frequency values are calculated is varied. First can have two consecutive signal edges of essentially same length can be used. The frequency value is then the reciprocal of the time interval between them Flanks. Alternatively, a selection can also be made according to the amplitude of the detected signal edges. So two successive signal edges can be the same Amplitude can be taken to get a frequency value determine. However, it doesn't always have to be two consecutive Signal edges are taken, but z. B. always the second, third, fourth, ... signal edge of the same amplitude or length. Finally, it should be noted that also any two signal edges can be taken to under Using statistical methods and based on the Superposition laws to get the coordinate tuple. At the Example of a flute, it becomes clear that a flute sound is two Signal edges with high amplitude provides between those there is a wave crest with a lower amplitude. Around for example, to determine the root note of the flute a selection of the two detected signal edges after the Amplitude are taken.

The temporal sequence represents in particular for audio signals of tones the most natural way of characterizing because there is the easiest way to recognize it from music signals is the essence of the audio signal in the chronological sequence of tones. The most immediate Sensation that a listener receives from a music signal is the sequence of tones. Not just in the classic Music where works are always about a certain one Build up subject that is in various modifications through the whole work, but also with songs of popular or other contemporary music exists a catchy melody that generally consists of one episode consists of simple tones, the subject or simple melody essential to the recognition ability regardless of rhythm, pitch, any instrument accompaniment etc. shapes.

The concept according to the invention is based on this knowledge and provides a signal identifier that consists of a temporal Sequence of frequencies exists or, depending on the embodiment, from a chronological sequence of frequencies, d. H. Tones derived from statistical methods.

An advantage of the present invention is that the signal identifier as a chronological sequence of frequencies High information content fingerprint for time signals with a harmonic component and to a certain extent the essential or the core of a time signal.

Another advantage of the present invention is in that the signal identifier extracted according to the invention represents a strong compression of the time signal, however, the timing of the time signal continues is based on the natural view of time signals, e.g. B. pieces of music is adjusted.

Another advantage of the present invention is in that due to the sequential nature of the signal identifier from the distance calculation referencing algorithms in State of the art can be gone and for referencing the time signal used in a database algorithms can be known from DNA sequencing are and that in addition also similarity calculations can be performed using DNA sequencing algorithms with replace / insert / delete operations be used.

Another advantage of the present invention is in that to detect the time occurrence of Signal edges in the time signal in a favorable manner the Hough transform can be used for that efficient algorithms from image processing and image recognition exist.

Another advantage of the present invention is in that the signal identifier extracted according to the invention of a time signal is independent of whether the search signal identifier from the entire time signal or only from one Section of the time signal is derived because according to the DNA sequencing algorithms one step at a time Comparison of the search signal identifier with a reference signal identifier can be carried out, due to the temporal sequential comparison of the section of the identifying time signal to a certain extent automatically the reference time signal is identified where the highest match between search signal identifier and Reference signal identifier exists.

Fig. 1

ein Blockschaltbild der erfindungsgemäßen Vorrichtung zum Extrahieren einer Signalkennung aus einem Zeitsignal;

Fig. 2

ein Blockschaltbild eines bevorzugten Ausführungsbeispiels, in dem eine Vorverarbeitung des Audiosignals dargestellt ist;

Fig. 3

ein Blockschaltbild eines Ausführungsbeispiels für die Signalkennungserzeugung;

Fig. 4

ein Blockschaltbild für eine erfindungsgemäße Vorrichtung zum Erzeugen einer Datenbank und zum Referenzieren eines Such-Zeitsignals in der Datenbank.

Fig. 5

graphische Darstellung eines Ausschnitts von Mozart KV 581 durch Frequenz-Zeit-Koordinaten-Tupel.

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

Fig. 1

a block diagram of the inventive device for extracting a signal identifier from a time signal;

Fig. 2

a block diagram of a preferred embodiment, in which a preprocessing of the audio signal is shown;

Fig. 3

a block diagram of an embodiment for the signal identification generation;

Fig. 4

a block diagram for an inventive device for generating a database and for referencing a search time signal in the database.

Fig. 5

graphic representation of a section of Mozart KV 581 by frequency-time-coordinate tuple.

Fig. 1 shows a block diagram of an extracting device a signal identifier from a time signal. The device comprises a device 12 for performing a Signal edge detection, a device 14 for determining the distance between two selected detected edges, a device 16 for frequency calculation and a device 18 for signal identification generation using from output from the device 16 for frequency calculation Coordinate tuples, each a frequency value and have an appearance time for this frequency value.

At this point it should be noted that, although in subsequent spoken of an audio signal as a time signal is, the concept of the invention not only for audio signals is suitable, but for all time signals, which have a harmonious part, because the signal identification is based on the fact that a time signal from a time sequence of frequencies, using the example of the audio signal of tones, consists.

The device 12 for detecting the occurrence of time of signal edges in the time signal preferably carries one Hough transformation through.

The Hough transform is described in U.S. Patent No. 3,069,654 by Paul V. C. Hough. The Hough transformation is used to recognize complex structures and especially for the automatic detection of complex Lines in photographs or other images. The Hough transformation is thus generally a technique that can be used to create features with special shape inside extract an image.

In their application according to the present invention the Hough transform used to do this from the time signal Extract signal edges with specified time lengths. A signal edge is initially determined by its temporal Length specified. Ideally a sine wave would be a signal edge through the rising edge of the sine function defined from 0 to 90 °. Alternatively, one could Signal edge also due to the increase in the sine function of -90 ° to + 90 ° can be specified.

Is the time signal as a result of time samples ("Samples"), the length of time corresponds to one Signal edge taking into account the sampling frequency, with the samples were generated, a certain number of samples. The length of a signal edge can thus by simply specifying the number of samples, specified to encompass the signal edge become.

In addition, it is preferred to have a signal edge only then to be detected as a signal edge if the same is continuous and is predominantly monotonous, i.e. in In the case of a positive signal edge a predominantly monotone has increasing course. Of course you can too negative signal edges, i.e. monotonously falling signal edges can be detected.

Another criterion for the classification of signal edges is that a signal edge is only a signal edge is detected when it has a certain level range exceeds. To hide noise disturbances, it is preferred to use a minimum for a signal edge Specify level range or amplitude range, being monotonous rising signal edges below this level range cannot be detected as signal edges.

According to a preferred embodiment of the present Invention is used for referencing audio signals further restriction that only Signal edges are searched, their specified temporal length greater than a minimum limit length and is less than a maximum time limit. This in other words means that only signal edges are searched that are on frequencies less than an upper cutoff frequency and greater than a lower cutoff frequency Clues. It is preferred for pieces of music, just Detect signal edges based on frequencies in the Frequency range from 27.5 Hz (tone A2) to 4186 Hz (tone c5) Clues. This frequency range is defined by a The usual piano provided overtones. This has for signal identifications of pieces of music Sound range highlighted as sufficient.

The signal edge detection unit 12 thus provides one Signal edge and the time of occurrence of the signal edge. It is irrelevant whether the time of the signal occurrence the signal edge the time of the first sample the signal edge, the time of the last sample the signal edge or the time of any sample is taken within the signal edge as long Signal edges are treated the same.

The device 14 for determining a time interval between two successive signal edges whose time lengths apart from a predetermined tolerance value are the same, examined by the device 12 output signal edges and extracted two consecutive Signal edges that are the same or within one certain predetermined tolerance value essentially are the same. If a simple sinus tone is considered, such is a period of the sine tone through the time interval two successive equally long z. B. more positive Given quarter waves. The facility is based on this 16 for calculating a frequency value from the determined time interval. The frequency value corresponds to the inverse the determined time interval.

By doing this, high time and at the same time frequency resolution a representation of a Time signal by specifying those occurring in the time signal Frequencies and by specifying the with the frequencies Corresponding performance times can be delivered. If the results of the device 16 for frequency calculation are graphically represented, is a diagram according to Fig. 5 obtained.

Fig. 5 shows a section of about 13 seconds in length of the clarinet quintet in A major, Larghetto, KV 581 by Wolfgang Amadeus Mozart as it exit 16 would appear for frequency calculation. In this excerpt a clarinet sounds, which is a melody leading Solo part plays as well as an accompanying string quartet. The coordinate tuples shown in FIG. 5 result, as provided by the device 16 for frequency calculation could be generated.

The device 18 is finally used from the results to generate a signal identifier for the device 16, the cheap and suitable for a signal identification database is. The signal identifier is generally made up of a plurality generated from coordinate tuples, each coordinate tuple includes a frequency value and an occurrence time, so that the signal identifier is a sequence of signal identifier values includes the time course of the time signal.

As will be explained later, the device 18 serves to from the frequency-time diagram of Fig. 5, which by the device 16 could be generated, the essential Extract information to a fingerprint of the To generate time signal that is compact on the one hand, and on the other hand, the time signal is sufficiently precise and distinguishable can differ from other time signals.

2 shows an extraction device according to the invention a signal identifier according to a preferred embodiment of the present invention. As a time signal an audio file 20 is input to an audio I / O handler. The audio I / O handler 22 reads the audio file, for example from a hard drive. The audio data stream can can also be read directly via a sound card. To reading a portion of the audio data stream device 22 retrieves the audio file and loads the next one audio file to be edited or terminates the import process. The sequence of PCM samples (PCM = pulse code Modulated), as received for example from a CD are then transferred to a device 24 for preprocessing of the audio signal entered. The device 24 serves on the one hand, if necessary a sample rate conversion perform, or a volume modification of the Audio signal. Audio signals are on different Media in different sampling frequencies in front. However, as has already been stated, the Time of occurrence of a signal edge in the audio signal used to describe the audio signal so that the sampling rate must be known to the times of occurrence of signal edges to be detected correctly and above also correctly detect frequency values. alternative can do a sample rate conversion by decimation or interpolation be performed to the audio signals different Bring sampling rates to the same sampling rate.

In a preferred embodiment of the present Invention that is said to be suitable for multiple sampling rates the device 24 is therefore provided for setting a sampling rate perform.

The PCM samples also become automatic Level adjustment also undergone in the facility 24 is provided. In the device 24 is for automatic Level adjustment in a look-ahead buffer is the middle one Signal power of the audio signal determined. The audio signal section, the between two signal power minima is multiplied by a scaling factor that the product of a weighting factor and the quotient from full scale and maximum level within the segment is. The length of the look-ahead buffer is variable.

The audio signal preprocessed in this way is then converted into the device 12 is fed by a signal edge detection performs as described with reference to FIG. 1 has been. For this purpose, the Hough transform used. A circuit technology Realization of the Hough transformation is in the WO 99/26167.

The amplitude of one determined by the Hough transformation Signal edge and the time of detection of a signal edge are then transferred to the device 14 of FIG. 1. In this unit there are two successive ones Subtract detection times from each other, where the reciprocal of the difference in performance times as a frequency value Is accepted. This task is accomplished by the 1 causes and leads when a piece of music is processed accordingly to the frequency-time diagram of Fig. 5, in which the obtained frequency-time coordinate tuple are graphically represented by Mozart, Köchel-Directory 581.

According to the invention, the representation of FIG. 5 could already be can be used as a signal identifier for the time signal since the temporal sequence of the coordinate tuples the temporal Reproduces the course of the time signal.

In one embodiment, however, it is preferred to use one Postprocess to get out of the frequency-time diagram 5 to extract the essential information, one for signal referencing if possible small and yet meaningful fingerprint deliver for the time signal.

For this purpose, the signal identification generator 18 can be used as in FIG. 3 be constructed shown. The device 18 is structured into a device 18a for determining the cluster areas, into a device 18b for grouping, into a Device 18c for averaging over a group, in a Device 18d for setting intervals, in a device to quantize 18e and finally into a device 18f to switch the signal identifier for the time signal receive.

As can be clearly seen in FIG. 5, the device 18a characteristic distribution point clouds to determine the cluster areas, which are referred to as clusters or clusters are worked out. This is done by isolating everyone Frequency-time tuples are deleted that have a given Minimum distance to the closest spatial neighbor exceed. Such isolated frequency-time tuples are for example, the dots in the top right corner of the 5. This leaves a so-called pitch contour strip band left that in Fig. 5 with the reference symbol 50 is outlined. The Pitch Contour strip tape consists of clusters of certain frequency latitude and longitude, whereby these clusters are caused by played notes. These tones are shown in Fig. 5 by horizontal lines intersect the ordinate, indicated (52), with this one shown example the tones h1, c2, cis2, d2 and h1 in the Range between about 6 and 10 seconds in the above Episode occur. The tone a1 has a frequency of 440 Hz. The tone h1 has a frequency of 494 Hz. The tone c2 has one Frequency of 523 Hz, the tone cis2 has a frequency of 554 Hz, while the tone d2 has a frequency of 587 Hz.

In the case of polyphonic sounds, there are wider stripes. The stripe width for single tones also depends of a vibrato of the musical instrument producing the single tones from.

In the device 18b for grouping or for forming Blocks become the coordinate tuples of the pitch contour strip in a time window of n samples summarized in a processing block to be processed separately or grouped. The block size can be equidistant or variable. Depending on accuracy and available storage space for the signal identification a relatively rough division can be chosen, for example a one-second grid thing about the present Sampling rate of a certain number of samples corresponds to per block, or a smaller division. alternative can, in the case of pieces of music the underlying notation To take account of the grid always chosen that a tone falls into the grid. This is it is necessary to estimate the length of a sound, what by the polynomial fit function 54 shown in FIG. 5 is possible. A group or block is then created by the temporal distance between two local extreme values of the Polynomial determined. This approach delivers particularly relatively large groups for relatively monophonic sections samples as they occur between 6 and 12 seconds, while at relatively polyphonic intervals of the piece of music, where the coordinate tuples are larger than one Frequency range are distributed, such as. B. at about 2 seconds 5 in Fig. 5 or smaller at 12 seconds from Fig. 5 Groups are determined, which in turn leads to the fact that the Signal identification carried out on the basis of relatively small groups becomes smaller, so that the information compression than with solid block formation.

In block 18c for averaging over a group of samples becomes a weighted average as needed determined over all coordinate tuples present in a block. In the preferred embodiment, the Tuples outside of the Pitch Contour strip band beforehand "Hidden". Alternatively, however, this can also be done Hide are dispensed with, which leads to all coordinate tuple calculated by the device 16 at the averaging performed by the device 18c will be taken into account.

In the device 18d for setting intervals, a Jump distance to determine the middle of the next, d. H. temporally following, group of samples determined.

It should be noted that in device 18c either an arithmetic, a geometric or a median averaging can be carried out.

In the quantizer 18e, the value generated by the device 18c has been calculated in non-equidistant Grid values quantized. For pieces of music, it is preferred carry out the division according to the audio frequency scale, being the tone frequency scale, as already stated has been classified according to the frequency range, which is supplied by a standard piano and differs from 27.5 Hz (tone A2) to 4186 Hz (tone c5) and 88 tone levels includes. Is the averaged value at the output of the Device 18c between two adjacent semitones, see above he receives the value of the closest reference tone.

This results in quantization at the output of the device 18e gradually a sequence of quantized values, which together make up the signal identifier. As required can the quantized values by the device 18f be post-processed, with post-processing for example in a pitch offset correction, one Transposition into another tone scale, etc. could exist.

In the following, reference is made to FIG. 4. Fig. 4 shows schematically a device for referencing a Search time signal in a database 40, the database 40 signal identifiers of a plurality of database time signals Track_1 to Track_m, which in one preferably library 42 separate from database 40 are saved.

To reference a time signal based on database 40 , the database must first be filled, which in a "learning" mode can be achieved. To do this Audio files 41 are gradually fed to a vector generator 43, which has a reference identifier for each audio file and stored in the database so that they are recognized can to which audio file z. B. in library 42 the signal identifier belongs.

According to the assignment given in FIG. 4, the signal identifier corresponds MV11, ...., MV1n the time signal Track_1. The Signal identifier MV21, ..., MV2n belongs to the time signal Track_2. Finally, the signal identifier MVm1, ..., MVmn to the time signal Track_m.

The vector generator 43 is designed to generally perform the operations shown in FIGS Fig. 1 perform functions, and is according to a preferred embodiment as in FIG. 2nd and 3 implemented. Processed in "Learn" mode the vector generator 43 gradually different Audio files (Track_1 to Track_m) for signal identifiers for save the time signals in the database, d. H. around to fill the database.

In "search" mode, an audio file 41 is to be based on the database 40 are referenced. For this, the search time signal 41 processed by the vector generator 43 to generate a search identifier 45. The search identifier 45 will then fed into a DNA sequencer 46 to match the reference identifiers to be compared in database 40. The DNA sequencer 46 is also arranged to make a statement about the search time signal with respect to the plurality of database time signals from library 42. The DNA sequencer searches the database with the search identifier 45 40 on a matching reference identifier and passes a pointer to the corresponding one with the reference identifier associated audio files in library 42.

The DNA sequencer 46 thus makes a comparison of the search identifier 45 or parts thereof with the reference identifiers in the database. If the given sequence is available or a partial sequence thereof becomes the associated time signal referenced in library 42.

DNA sequencer 46 preferably uses a Boyer-Moore algorithm which, for example, in the specialist book "Algorithms on Strings, Trees and Sequences", Dan Gusfield, Cambridge University Press, 1997. According to A first alternative is based on exact match checked. The conclusion of a statement is therefore that it is said that the search time signal is identical to one Time signal in library 42 is. Alternatively or additionally can also see the similarity of two sequences Use of replace / insert / delete operations and a pitch offset correction (pitch offset correction) to be examined.

Preferably, database 40 is structured to: composed of the chaining of signal identification sequences is, the end of each vector signal identifier of a time signal is specified by a separator so the search does not continue beyond time signal file limits. If multiple matches are found, all of them referenced time signals specified.

By using the replace / insert / delete operations (Replace / Insert / Delete) a similarity measure can be introduced with the time signal referenced in library 42 is that the search time signal 41 based on a predetermined Similarity measure is most similar. Furthermore, it prefers to measure the similarity of the search audio signal and then determine multiple signals in the library the most similar n sections in library 42 in Order of descending similarity.

Claims (22)

1. Method for extracting a signal identifier from a time signal having a harmonic portion, the method comprising:

   detecting (12) the temporal occurrence of signal edges in the time signal;

   determining (14) a temporal interval between two selected detected signal edges;

   calculating (16) a frequency value from the temporal interval determined, and associating the frequency value with a time of occurrence of the frequency value in the time signal to obtain a coordinate tuple from the frequency value and the time of occurrence for this frequency value; and

   creating (18) the signal identifier from a plurality of coordinate tuples, each coordinate tuple including a frequency value and a time of occurrence, whereby the signal identifier includes a sequence of signal-identifier values which reflects the temporal form of the time signal.
2. Method as claimed in claim 1, wherein in the step of detecting (12), a signal-flank is detected as a signal-flank only if same has, over its specified temporal length, an amplitude larger than a predetermined amplitude threshold value.
3. Method as claimed in claim 1 or 2,
   wherein in the step of detecting (12), a signal-flank is detected as a signal-flank only if its specified temporal length is longer than a minimum cut-off length and shorter than a maximum cut-off length.
4. Method as claimed in claim 3, wherein the time signal is an audio signal, and wherein the minimum temporal cut-off length is specified by means of a maximum audible cut-off frequency, and the maximum temporal cut-off length is specified by means of a minimum audible cut-off frequency.
5. Method as claimed in claim 3, wherein the time signal is an audio signal, and wherein the minimum temporal cut-off length is specified by means of a maximum tone frequency that may be created by an instrument, and the maximum temporal cut-off length is specified by means of a minimum tone frequency which may be created by an instrument.
6. Method as claimed in any one of the previous claims, wherein the step of creating (18) the signal identifier comprises:

   eliminating (18a) coordinate tuples spaced apart by more than a predetermined threshold distance from an adjacent coordinate tuple in a frequency-time diagram so as to determine clusters of coordinate tuples.
7. Method as claimed in claim 5 or 6, wherein the step of creating (18) comprises:

   grouping (18b) coordinate tuples in successive temporal intervals into blocks of coordinate tuples.
8. Method as claimed in claim 7, wherein the successive temporal intervals have a fixed and/or a variable length.
9. Method as claimed in claim 7 or 8, wherein the step of creating (18) the signal identifier comprises:

   averaging (18c) the frequency values of coordinate tuples in the temporal intervals to obtain a sequence of averaged frequency values for a sequence of temporal intervals, the sequence of averaged frequency values representing a feature vector.
10. Method as claimed in claim 9, wherein step (18) of creating the signal identifier comprises:

    quantizing (18e) the feature vector to obtain a quantized feature vector.
11. Method as claimed in claim 10, wherein the step of quantizing (18e) is performed using non-equidistantly distributed raster points, distances between two adjacent raster points being determined in accordance with a tone-frequency scale.
12. Method as claimed in any one of the previous claims, wherein in step (12) of detecting signal edges, a Hough transformation is employed.
13. Method for creating a database (40) from reference signal identifiers for a plurality of time signals, comprising:

    extracting a first signal identifier for a first time signal by the method as claimed in any one of claims 1 to 12;

    extracting a second signal identifier for a second time signal by means of a method as claimed in any one of claims 1 to 12; and

    storing the extracted first signal identifier in association with the first time signal in the database (40); and

    storing the extracted second signal identifier in association with the second time signal in the database (40).
14. Method of referencing a search time signal using a database (40), the database comprising reference signal identifiers of a plurality of database time signals, a reference signal identifier of a database time signal having been determined by a method as claimed in any one of claims 1 to 12, the method comprising:

    providing at least one portion of a search time signal (41);

    extracting (43) a search signal identifier from the search time signal by a method as claimed in any one of claims 1 to 12; and

    comparing (46) the search signal identifier with the plurality of reference signal identifiers, and, in response to the step of comparing, making a statement about the search time signal with regard to the plurality of database time signals.
15. Method as claimed in claim 14, wherein in the step of making a statement, a search time signal is identified as a reference time signal if the search signal identifier matches at least a portion of a reference signal identifier.
16. Method as claimed in claim 14, wherein in the step of making a statement, a similarity between a search time signal and a database time signal is established if the search signal identifier and/or at least a portion of database signal identifier may be made to match by means of a reproducible manipulation.
17. Method as claimed in any one of claims 14 to 16,
    wherein the database signal identifier comprises a sequence of database signal identifier values reproducing the temporal form of the database time signal,
    wherein the search signal identifier comprises a search sequence of search signal identifier values reproducing the temporal form of the search time signal,
    wherein the length of the database sequence is longer than the length of the search sequence, and
    wherein the search sequence is sequentially compared to the database sequence.
18. Method as claimed in claim 17, wherein during the sequential comparing of the search sequence with the database sequence, a correction of the values of the search and/or the database signal identifier is performed by a replace, insert or delete operation of at least one value of the search and/or the database signal identifier to determine a similarity of the search time signal and the database time signal.
19. Method as claimed in any one of claims 14 to 18,
    wherein the step of comparing (46) is performed using a DNA sequencing algorithm and/or using the Boyer-Moore algorithm.
20. Apparatus for extracting a signal identifier from a time signal having a harmonic portion, the apparatus comprising:

    means for detecting (12) the temporal occurrence of signal edges in the time signal;

    means for determining (14) a temporal interval between two selected detected signal edges;

    means for calculating (16) a frequency value from the temporal interval determined, and for associating the frequency value with a time of occurrence of the frequency value in the time signal to obtain a coordinate tuple from the frequency value and the time of occurrence for this frequency value; and

    means for creating (18) the signal identifier from a plurality of coordinate tuples, each coordinate tuple including a frequency value and a time of occurrence, whereby the signal identifier includes a sequence of signal-identifier values which reflects the temporal form of the time signal.
21. Apparatus for creating a database (40) from reference signal identifiers for a plurality of time signals, comprising:

    means for extracting a first signal identifier for a first time signal as claimed in claim 20;

    means for extracting a second signal identifier for a second time signal as claimed in claim 20; and

    means for storing the extracted first signal identifier in association with the first time signal in the database (40); and

    means for storing the extracted second signal identifier in association with the second time signal in the database (40).
22. Apparatus for referencing a search time signal using a database (40), the database comprising reference signal identifiers of a plurality of database time signals, a reference signal identifier of a database time signal having been determined by a method as claimed in any one of claims 1 to 12, the apparatus comprising:

    means for providing at least one portion of a search time signal (41);

    means for extracting (43) a search signal identifier as claimed in claim 20; and

    means for comparing (46) the search signal identifier with the plurality of reference signal identifiers, and, in response to the step of comparing, making a statement about the search time signal with regard to the plurality of database time signals.

Images