Method and apparatus for identifying an unknown work
View Patent ↗A system for determining an identity of a received work. The system receives audio data for an unknown work. The audio data is divided into segments. The system generates a signature of the unknown work from each of the segments. Reduced dimension signatures are then generated at least a portion of the signatures. The reduced dimension signatures are then compared to reduced dimensions signatures of known works that are stored in a database. A list of candidates of known works is generated from the comparison. The signatures of the unknown works are then compared to the signatures of the known works in the list of candidates. The unknown work is then identified as the known work having signatures matching within a threshold.
1. A method, implemented by a computing system programmed to perform the following, comprising:
providing, by the computing system, a reference database containing a plurality of reference works, wherein each reference work corresponds to a reference signature comprising a plurality of reference segments;
receiving, by the computing system, an unknown work, wherein the unknown work comprises a plurality of unknown segments;
calculating, by the computing system, at least one distance from one of the plurality of unknown segments to one of the plurality of reference segments;
if the at least one calculated distance does not meet a predefined threshold, repeating, by the computing system, a calculation for a distance from each subsequent segment of the plurality of unknown segments to each subsequent segment of the plurality of reference segments until the predefined threshold is met; and
identifying, by the computing system, the unknown work as one of the plurality of reference works based on the calculated distances.
2. The method of claim 1 , wherein the predetermined threshold includes one calculated distance greater than 20.
3. The method of claim 1 , wherein the predetermined threshold includes two calculated distances greater than 15 .
4. The method of claim 1 , wherein the predetermined threshold includes four calculated distances greater than 10.
5. The method of claim 1 , wherein calculating the at least one distance from one of the plurality of unknown segments to one of the plurality of reference segments comprises:
calculating a mel frequency cepstral coefficients (MFCC) for the one of the plurality of unknown segments;
calculating a MFCC for the one of the plurality of reference segments; and
calculating a sum of squared differences of the MFCCs for the unknown segment and the
6. A non-transitory computer readable storage medium, comprising executable instructions which when executed on a processing system cause the processing system to perform a method comprising:
providing a reference database containing a plurality of reference works, wherein each reference work corresponds to a reference signature comprising a plurality of reference segments, the segments defining a hop size;
receiving an unknown work having an unknown signature;
calculating a distance from the unknown signature to one of the plurality of reference signatures at a point in time in said reference signature;
if the calculated distance is greater than a predefined threshold, repeating a calculation for a distance from the unknown signature to the reference signature at least one offset point in time in the reference signature until the calculated distance is less than the predefined threshold;
calculating a distance from the unknown signature to the reference signature at one or more points in time neighboring the point in time where the distance calculated is less than the predefined threshold; and
identifying the unknown signature as one of the plurality of reference works based on the distances calculated at the one or more neighboring points in time.
7. The non-transitory computer readable storage medium of claim 6 , wherein the offset point in time is a point in time one or more multiples of the hop size from the last point in time where the distance was calculated.
8. The non-transitory computer readable storage medium of claim 6 , wherein the offset point in time is a point in time three hop sizes from the last point in time where the distance was calculated.
9. The non-transitory computer readable storage medium of claim 6 , wherein the one or more neighboring points in time is one or more point in time one hop size from the point in time in the reference signature where the distance calculated is less than the predefined threshold.
10. The non-transitory computer readable storage medium of claim 6 , wherein the predefined threshold is 8.0.
11. A computer-implemented apparatus comprising:
a memory to store a reference database containing a plurality of reference works, wherein each reference work corresponds to a reference signature comprising a plurality of reference segments;
a processor coupled to the memory
to receive an unknown work, wherein the unknown work comprises a plurality of unknown segments;
to calculate at least one distance from one of the plurality of unknown segments to one of the plurality of reference segments;
if the at least one calculated distance does not meet a predefined threshold, to repeat a calculation for a distance from each subsequent segment of the plurality of unknown segments to each subsequent segment of the plurality of reference segments until the predefined threshold is met; and
to identify the unknown work as one of the plurality of reference works based on the calculated distances.
12. The system of claim 11 , wherein the predetermined threshold includes one calculated distance greater than 20.
13. The system of claim 11 , wherein the predetermined threshold includes two calculated distances greater than 15.
14. The system of claim 11 , wherein the predetermined threshold includes four calculated distances greater than 10.
15. The system of claim 11 , wherein the processor is to calculate the at least one distance from one of the plurality of unknown segments to one of the plurality of reference segments by:
calculating a mel frequency cepstral coefficients (MFCC) for the one of the plurality of unknown segments;
calculating a MFCC for the one of the plurality of reference segments; and
calculating a sum of squared differences of the MFCCs for the unknown segment and the reference segment.