Interest points density control for audio matching
Systems and methods are provided herein relating to audio matching. The density and quality of interest points can be controlled to assure a small but uniform number of high quality interest points. By scoring interest points based on quality and comparing them over time, those interest points that maintain a high quality when compared with a varying number of neighboring interest points can be retained, while those interest points that do not maintain a high quality can be discarded. Thus, the scalability of an audio matching system can be improved while retaining accuracy.
1. A system comprising:
a memory that has stored thereon computer executable components; and
a processor that executes the following computer executable components stored in the memory:
an input component that receives an audio sample;
a spectrogram component that generates a spectrogram of the audio sample;
an interest point detection component that generates a set of interest points based on local maxima over respective time/frequency windows of the spectrogram, wherein interest points in the set of interest points are associated with time and frequency positions;
a quality component that generates and associates respective quality scores with interest points in the set of interest points, wherein a quality score for an interest point is based on a magnitude of the interest point compared to an average magnitude of points of a time/frequency window associated with the interest point; and
an interest point filter component that filters the set of interest points based on the respective quality scores associated with the interest points in the set of interest points and a desired density.
2. The system of claim 1 , wherein the quality score has positive correlation with a likelihood that the interest point will not change in presence of distortion.
3. The system of claim 1 , wherein the interest point detection component adjusts sizes of the respective time/frequency windows based on the desired density.
4. The system of claim 1 , wherein the interest point filter component filters the set of interest points further based on a quality ranking and a distribution threshold.
5. The system of claim 4 , wherein the quality ranking and the distribution threshold are set to maximize quality.
6. The system of claim 4 , wherein the quality ranking and the distribution threshold are set to maximize distribution.
7. A method comprising:
receiving, by a system including a processor, an audio sample;
generating, by the system, a spectrogram of the audio sample;
generating, by the system, a set of interest points based on local maxima over respective time/frequency windows of the spectrogram, wherein interest points in the set of interest points are associated with a time and frequency position;
generating and associating, by the system, respective quality scores with interest points in the set of interest points, wherein a quality score for an interest point is based on a magnitude of the interest point compared to an average magnitude of points of a time/frequency window associated with the interest point; and
filtering, by the system, the set of interest points based on the respective quality scores associated with the interest points in the set of interest points and a desired density.
8. The method of claim 7 , wherein the quality score has positive correlation with a likelihood that the interest point will not change in presence of distortion.
9. The method of claim 7 , further comprising adjusting sizes of the respective time/frequency windows based on the desired density.
10. The method of claim 7 , wherein filtering the set of interest points is further based on a quality ranking and a distribution threshold.
11. The method of claim 10 , wherein the quality ranking and the distribution threshold are set to maximize quality.
12. The method of claim 10 , wherein the quality ranking and the distribution threshold are set to maximize distribution.
13. A non-transitory computer-readable medium having instructions stored thereon that, in response to execution, cause a system including a processor to perform operations comprising:
generating a spectrogram of an audio sample;
generating a set of interest points based on local maxima over respective time/frequency windows of the spectrogram wherein interest points in the set of interest points are associated with a time and frequency position;
associating respective quality scores with interest points in the set of interest points, wherein a quality score for an interest point is based on a magnitude of the interest point compared to an average magnitude of points of a time/frequency window associated with the interest point; and
filtering the set of interest points based on the respective quality scores associated with the interest points in the set of interest points and a desired density.
14. The non-transitory computer-readable medium of claim 13 , wherein the quality score has positive correlation with a likelihood that the interest point will not change in presence of distortion.
15. The non-transitory computer-readable medium of claim 13 , the operations further comprising adjusting sizes of the respective time/frequency windows based on the desired density.
16. The non-transitory computer-readable medium of claim 13 , the operations further comprising filtering the set of interest points further based on a quality ranking and a distribution threshold.
17. The non-transitory computer-readable medium of claim 16 , wherein the quality ranking and the distribution threshold are set to maximize quality.
18. The non-transitory computer-readable medium of claim 16 , wherein the quality ranking and the distribution threshold are set to maximize distribution.
19. The non-transitory computer-readable medium of claim 13 , the operations further comprising generating a set of descriptors for the audio sample based on the set of interest points.
20. The non-transitory computer-readable medium of claim 19 , the operations further comprising generating and associating metadata with the descriptor, wherein the metadata includes at least one of the desired density, a quality ranking, and a distribution threshold.
21. The non-transitory computer-readable medium of claim 19 , the operations further comprising identifying a set of potential matches based on comparing the set of descriptors with a set of reference descriptors.
22. The non-transitory computer-readable medium of claim 21 , the operations further comprising identification of the set of potential matches.
23. The non-transitory computer-readable medium of claim 21 , wherein the set of reference descriptors are located in a cloud computer.
24. The non-transitory computer-readable medium of claim 23 , the operations further comprising identifying the set of potential matches by transmitting the descriptors to the cloud computer, wherein the cloud computer employs the descriptors to identify a set of potential matches based on comparing the descriptors to the set of reference descriptors.