Method and system for analyzing ditigal audio files

1. A method for generating a fingerprint of an audio waveform, comprising:

defining a codebook which represents a multivariate vector of one or more spectral features with one of a plurality of codes, each code corresponding to a bin and representing a portion of a predetermined audio waveform;

dividing the audio waveform into a plurality of bins,

for each bin of the plurality of bins, computing one or more spectral features for the bin,

wherein the computing comprises:

(a) computing the one or more spectral features for a first group of data points within the bin;

(b) shifting some number of data points within the bin; and

(c) computing the one or more spectral features for a second group of data points within the bin;

mapping the one or more spectral features obtained by the computing to a string of codes from the codebook;

generating a fingerprint by representing the audio waveform with the string of codes obtained by the mapping; and

aligning the string of codes in a time series.

2. The method of claim 1 , wherein one or more data points in the first group overlap with one or more data points in the second group.

3. The method of claim 1 , wherein each code is a hash code.

4. The method of claim 1 , further comprising:

compressing the string of codes from the codebook to form a compressed string, wherein the codes of the compressed string are aligned in a time series.

5. A method for creating a signature of an audio waveform, comprising:

dividing the audio waveform into a plurality of bins;

for the plurality of the bins, selecting a first group of data points within each bin, and computing one or more spectral features for each bin based upon the first group of points within the plurality of bins,

and for the plurality of bins, selecting a second group of data points within each bin, and computing one or more spectral features for each bin based upon the second set of data points within the bins of the plurality;

mapping the one or more spectral properties for the plurality of bins to a string of codes by referencing a codebook of multivariate vectors; and

creating one or more signatures representing the audio waveform with the string of codes obtained by the mapping.

6. The method of claim 5 , wherein each code in the string represents the waveform over a portion of the waveform, and wherein the codes are aligned in a time series.

7. The method of claim 6 , further comprising:

compressing the string such that the time series alignment is maintained.

8. The method of claim 5 , further comprising:

comparing a signature of the one or more signatures that initiates at a given time with a representation of an audio segment.

9. The method of claim 8 , further comprising:

defining a codebook which represents a vector of one or more spectral features with a code, prior to referencing said codebook.

10. The method of claim 5 , wherein one or more signatures is created for each bin.

11. A method for creating a signature for an audio waveform, comprising:

dividing the audio waveform into a plurality of bins;

for the plurality of the bins, selecting a first group of data points within each bin, and computing one or more spectral properties for each bin based upon the first group of points within the plurality of bins,

and for the plurality of bins, selecting a second group of data points within each bin, and computing one or more spectral properties for each bin based upon the second set of data points within the plurality of bins;

mapping the one or more spectral properties for the plurality of bins to a string of hash codes by referencing a codebook of hash values representing a multivariate vector; and

creating one or more signatures representing the audio waveform with the string of hash values obtained by the mapping.

12. The method of claim 11 , wherein each code is a hash code.

13. A computer-readable memory medium having computer-executable code stored therein for causing a computer to generate a fingerprint of an audio waveform, said computer-executable code comprising:

code for causing the computer to define a codebook which represents a multivariate vector of one or more spectral features with one of a plurality of codes, each code corresponding to a bin and representing a portion of a predetermined audio waveform;

code for causing the computer to divide the audio waveform into a plurality of bins;

code for causing the computer to compute one or more spectral features for a first group of data points within each bin of the plurality of bins;

code for causing the computer to shift some number of data points within each bin of the plurality of bins; and

code for causing the computer to compute one or more spectral features for a second group of data points within each bin of the plurality of bins;

code for causing the computer to map the one or more spectral features obtained by the computing to a string of codes from the codebook;

code for causing the computer to generate a fingerprint by representing the audio waveform with the string of codes; and

code for causing the computer to align the string of codes in a time series.

14. The memory medium of claim 13 , wherein one or more data points in the first group overlap with one or more data points in the second group.

15. The memory medium of claim 13 , wherein each code is a hash code.

16. The memory medium of claim 13 , wherein said computer-executable code , further comprises:

code for causing the computer to compress the string of codes from the codebook to form a compressed string, wherein the codes of the compressed string are aligned in a time series.

17. An apparatus for fingerprinting an audio waveform, comprising:

means for defining a codebook which represents a multivariate vector of one or more spectral features with one of a plurality of codes, each code corresponding to a bin and representing a portion of a predetermined audio waveform;

means for dividing the audio waveform into a plurality of bins;

means for computing one or more spectral features for a first group of data points within each bin of the plurality of bins;

means for shifting some number of data points within each bin of the plurality of bins; and

means for computing one or more spectral features for a second group of data points within each bin of the plurality of bins;

means for mapping the one or more spectral features for the first and second groups of data points obtained by the means for computing to a string of codes from the codebook;

means for generating a fingerprint by representing the audio waveform with the string of codes obtained by the mapping; and

means for aligning the string of codes in a time series.

18. A computer-readable memory medium having computer-executable code stored therein for causing a computer to create a signature of an audio waveform, said computer-executable code comprising:

code for causing the computer to divide the audio waveform into a plurality of bins;

code for causing the computer to select a first group of data points within each bin;

code for causing the computer to compute one or more spectral features for each bin based upon the first group of points within the plurality of bins;

code for causing the computer to select a second group of data points within each bin;

code for causing the computer to compute one or more spectral features for each bin based upon the second set of data points within the bins of the plurality;

code for causing the computer to map the one or more spectral features for the plurality of bins to a string of codes by referencing a codebook representing at least one multivariate vector; and

code for causing the computer to create one or more signatures representing the audio waveform with the string of codes.

19. The memory medium of claim 18 , wherein each code in the string represents the waveform over a portion of the waveform, and wherein the codes are aligned in a time series.

20. The memory medium of claim 19 , wherein said computer-executable code further comprises:

code for causing the computer to compress the string such that the time series alignment is maintained.

21. The memory medium of claim 19 , wherein said computer-executable code further comprises:

code for causing the computer to compare a signature of the one or more signatures that initiates at a given time with a representation of an audio segment.

22. The memory medium of claim 21 , wherein said computer-executable code further comprises:

code for causing the computer to define a codebook which represents a vector of one or more spectral features with a code, prior to referencing said codebook.

23. The memory medium of claim 18 , wherein one or more signatures is created for each bin.

24. The memory medium of claim 18 , wherein each code is a hash code.

25. An apparatus for creating a signature of an audio waveform, comprising:

means for dividing the audio waveform into a plurality of bins;

means for selecting a first group of data points within each bin;

means for computing one or more spectral features for each bin based upon the first group of points within the plurality of bins;

means for selecting a second group of data points within each bin;

means for computing one or more spectral features for each bin based upon the second set of data points within the bins of the plurality;

means for mapping the one or more spectral features for the plurality of bins to a string of codes by referencing a codebook, wherein the codebook represents at least one multivariate vector; and

means for creating one or more signatures representing the audio waveform with the string of codes.

26. A method for fingerprinting an audio waveform, comprising:

defining a codebook which represents a multivariate vector of one or more spectral features with a corresponding one of a plurality of codes;

dividing the audio waveform into a plurality of bins;

computing one or more spectral features for each bin; and

generating a fingerprint by mapping the one ore more spectral features to a string of codes from the codebook to represent the audio waveform based on the computed one or more spectral features for each bin.

27. A computer-readable memory medium having computer-executable code stored therein for causing a computer to create a signature of an audio waveform, said computer-executable code comprising:

code for causing the computer to define a codebook which represents a multivariate vector of one or more spectral features with a corresponding one of a plurality of codes;

code for causing the computer to divide the audio waveform into a plurality of bins;

code for causing the computer to compute one or more spectral features for each bin; and

code for causing the computer to map the one ore more spectral features to a string of codes from the codebook to generate a fingerprint representing the audio waveform based on the computed one or more spectral features for each bin.

28. An apparatus for fingerprinting an audio waveform, comprising:

means for defining a codebook which represents a multivariate vector of one or more spectral features with a corresponding one of a plurality of codes;

means for dividing the audio waveform into a plurality of bins;

means for computing one or more spectral features for each bin;

means for mapping the one ore more spectral features to a string of codes from the codebook; and

generating a fingerprint by representing the audio waveform with the string of codes.