Automated cover song identification

1. An apparatus comprising:

memory; and

one or more processors to execute instructions to:

identify query audio from a content source based on a search query using rights metadata associated with the query audio;

execute a constant Q transform on query time slices of the query audio;

binarize the constant Q transformed query time slices;

execute a two-dimensional Fourier transform on query time windows within the binarized and constant Q transformed query time slices to generate two-dimensional Fourier transforms of the query time windows;

generate a query data structure based on a sequential order of the two-dimensional Fourier transforms;

select a subset of a reference database based on the rights metadata, the subset including reference audio; and

identify the query audio as a cover rendition of the reference audio based on a comparison between the query data structure and a reference data structure associated with the reference audio.

2. The apparatus of claim 1 , wherein the rights metadata includes at least one of an artist, a publisher, license information, right holder information, royalty information, or a title of the query audio, and the one or more processors are to:

obtain the reference audio via a user interface in communication with a network; and

register the reference audio based on a first storage of the reference audio in the reference database and a second storage of the rights metadata in a rights database.

3. The apparatus of claim 1 , wherein the content source is at least one of (i) a stream of a live broadcast, (ii) a music sharing site, (iii) a video sharing site, or (iv) a social networking feed, a post, update, or a tweet of a social network.

4. The apparatus of claim 1 , wherein the one or more processors are to select the content source based on at least one of information provided by a right holder associated with the query audio, the rights metadata, a popularity of the query content source, or a likelihood of the query content source having potentially unlicensed cover songs.

5. The apparatus of claim 1 , wherein the one or more processors are to:

execute a constant Q transform on reference time slices of the reference audio;

binarize the constant Q transformed reference time slices;

execute a two-dimensional Fourier transform on reference time windows within the binarized and constant Q transformed reference time slices to generate two-dimensional Fourier transforms of the reference time windows; and

generate the reference data structure by sequentially ordering the two-dimensional Fourier transforms of the reference time windows.

6. The apparatus of claim 1 , wherein the one or more processors are to:

generate a similarity matrix that indicates degrees to which reference portions of the reference data structure are similar to query portions of the query data structure;

compute a distance measure between the query data structure and the reference data structure based on the similarity matrix; and

store an association in a database between the reference audio and the query audio based on a computed distance measure, the association identifying the query audio as the cover rendition.

7. The apparatus of claim 6 , wherein the one or more processors are to:

convolve the similarity matrix with a checkerboard kernel to generate a first convolved similarity matrix, the first convolved similarity matrix including positive elements and negative elements; and

replace the negative elements with zeros to generate a second convolved similarity matrix; and wherein:

the computing of the distance measure between the query data structure and the reference data structure is based on the second convolved similarity matrix.

8. The apparatus of claim 1 , wherein the one or more processors are to:

group the binarized and constant Q transformed query time slices of the query audio into the query time windows prior to the executing of the two-dimensional Fourier transform on the query time windows, the query time windows including overlapping query time windows of uniform duration; and

apply a blur algorithm to the two-dimensional Fourier transforms of the query time windows prior to the sequential ordering of the two-dimensional Fourier transforms in the query data structure.

9. A non-transitory machine-readable medium comprising instructions that, when executed, cause one or more processors to at least:

identify query audio from a content source based on a search query using rights metadata associated with the query audio;

execute a constant Q transform on query time slices of the query audio;

binarize the constant Q transformed query time slices;

execute a two-dimensional Fourier transform on query time windows within the binarized and constant Q transformed query time slices to generate two-dimensional Fourier transforms of the query time windows;

generate a query data structure based on a sequential order of the two-dimensional Fourier transforms;

select a subset of a reference database based on the rights metadata, the subset including reference audio; and

identify the query audio as a cover rendition of the reference audio based on a comparison between the query data structure and a reference data structure associated with the reference audio.

10. The non-transitory machine-readable medium of claim 9 , wherein the rights metadata includes at least one of an artist, a publisher, license information, right holder information, royalty information, or a title of the query audio, and the instructions, when executed, cause the one or more processors to:

obtain the reference audio via a user interface in communication with a network; and

register the reference audio based on a first storage of the reference audio in the reference database and a second storage of the rights metadata in a rights database.

11. The non-transitory machine-readable medium of claim 9 , wherein the content source is at least one of (i) a stream of a live broadcast, (ii) a music sharing site, (iii) a video sharing site, or (iv) a social networking feed, a post, update, or a tweet of a social network.

12. The non-transitory machine-readable medium of claim 9 , wherein the instructions, when executed, cause the one or more processors to select the content source based on at least one of information provided by a right holder associated with the query audio, the rights metadata, a popularity of the query content source, or a likelihood of the query content source having potentially unlicensed cover songs.

13. The non-transitory machine-readable medium of claim 9 , wherein the instructions, when executed, cause the one or more processors to:

execute a constant Q transform on reference time slices of the reference audio;

binarize the constant Q transformed reference time slices;

execute a two-dimensional Fourier transform on reference time windows within the binarized and constant Q transformed reference time slices to generate two-dimensional Fourier transforms of the reference time windows; and

generate the reference data structure by sequentially ordering the two-dimensional Fourier transforms of the reference time windows.

14. The non-transitory machine-readable medium of claim 9 , wherein the instructions, when executed, cause the one or more processors to:

generate a similarity matrix that indicates degrees to which reference portions of the reference data structure are similar to query portions of the query data structure;

compute a distance measure between the query data structure and the reference data structure based on the similarity matrix; and

store an association in a database between the reference audio and the query audio based on a computed distance measure, the association identifying the query audio as the cover rendition.

15. The non-transitory machine-readable medium of claim 14 , wherein the instructions, when executed, cause the one or more processors to:

convolve the similarity matrix with a checkerboard kernel to generate a first convolved similarity matrix, the first convolved similarity matrix including positive elements and negative elements; and

replace the negative elements with zeros to generate a second convolved similarity matrix; and wherein:

the computing of the distance measure between the query data structure and the reference data structure is based on the second convolved similarity matrix.

16. The non-transitory machine-readable medium of claim 9 , wherein the instructions, when executed, cause the one or more processors to:

group the binarized and constant Q transformed query time slices of the query audio into the query time windows prior to the executing of the two-dimensional Fourier transform on the query time windows, the query time windows including overlapping query time windows of uniform duration; and

apply a blur algorithm to the two-dimensional Fourier transforms of the query time windows prior to the sequential ordering of the two-dimensional Fourier transforms in the query data structure.

17. A method comprising:

identifying, using one or more processors, query audio from a content source based on a search query using rights metadata associated with the query audio;

executing, using the one or more processors, a constant Q transform on query time slices of the query audio;

binarizing, using the one or more processors, the constant Q transformed query time slices;

executing, using the one or more processors, a two-dimensional Fourier transform on query time windows within the binarized and constant Q transformed query time slices to generate two-dimensional Fourier transforms of the query time windows;

generating, using the one or more processors, a query data structure based on a sequential order of the two-dimensional Fourier transforms;

selecting, using the one or more processors, a subset of a reference database based on the rights metadata, the subset including reference audio; and

identifying, using the one or more processors, the query audio as a cover rendition of the reference audio based on a comparison between the query data structure and a reference data structure associated with the reference audio.

18. The method of claim 17 , wherein the rights metadata includes at least one of an artist, a publisher, license information, right holder information, royalty information, or a title of the query audio, and further including:

obtaining the reference audio via a user interface in communication with a network; and

registering the reference audio based on storing the reference audio in the reference database and storing the rights metadata in a rights database.

19. The method of claim 17 , wherein the content source is at least one of (i) a stream of a live broadcast, (ii) a music sharing site, (iii) a video sharing site, or (iv) a social networking feed, a post, update, or a tweet of a social network.

20. The method of claim 17 , wherein the one or more processors are to select the content source based on at least one of information provided by a right holder associated with the query audio, the rights metadata, a popularity of the query content source, or a likelihood of the query content source having potentially unlicensed cover songs.

21. The method of claim 17 , further including:

executing a constant Q transform on reference time slices of the reference audio;

binarizing the constant Q transformed reference time slices;

executing a two-dimensional Fourier transform on reference time windows within the binarized and constant Q transformed reference time slices to generate two-dimensional Fourier transforms of the reference time windows; and

generating the reference data structure by sequentially ordering the two-dimensional Fourier transforms of the reference time windows.

22. The method of claim 17 , further including:

generating a similarity matrix that indicates degrees to which reference portions of the reference data structure are similar to query portions of the query data structure;

computing a distance measure between the query data structure and the reference data structure based on the similarity matrix; and

storing an association in a database between the reference audio and the query audio based on a computed distance measure, the association identifying the query audio as the cover rendition.

23. The method of claim 22 , further including:

convolving the similarity matrix with a checkerboard kernel to generate a first convolved similarity matrix, the first convolved similarity matrix including positive elements and negative elements; and

replacing the negative elements with zeros to generate a second convolved similarity matrix; and wherein:

the computing of the distance measure between the query data structure and the reference data structure is based on the second convolved similarity matrix.

24. The method of claim 17 , further including:

grouping the binarized and constant Q transformed query time slices of the query audio into the query time windows prior to the executing of the two-dimensional Fourier transform on the query time windows, the query time windows including overlapping query time windows of uniform duration; and

applying a blur algorithm to the two-dimensional Fourier transforms of the query time windows prior to the sequential ordering of the two-dimensional Fourier transforms in the query data structure.