Methods and apparatus to fingerprint an audio signal via exponential normalization
Methods, apparatus, systems, and articles of manufacture are disclosed to fingerprint an audio signal via exponential normalization. An example apparatus includes an audio segmenter to divide an audio signal into a plurality of audio segments including a first audio segment and a second audio segment, the first audio segment including a first time-frequency bin, the second audio segment including a second time-frequency bin, a mean calculator to determine a first exponential mean value associated with the first time frequency bin based on a first magnitude of the audio signal associated with the first time frequency bin and a second exponential mean value associated with the second time frequency bin based on a second magnitude of the audio signal associated with the second time frequency bin and the first exponential mean value. The example apparatus further includes a bin normalizer to normalize the first time-frequency bin based on the second exponential mean value and a fingerprint generator to generate a fingerprint of the audio signal based on the normalized first time-frequency bins.
1. An apparatus, comprising:
signal transformer circuitry to transform an audio signal into a frequency domain including a plurality of time-frequency bins, wherein each of the time-frequency bins corresponds to an intersection of a frequency bin and a time bin and contains a portion of the audio signal;
audio segmenter circuitry to divide the audio signal into a plurality of audio segments including a first audio segment and a second audio segment, the first audio segment including a second group of time-frequency bins having a first time-frequency bin, the second audio segment including a first group of time-frequency bins having a second time-frequency bin;
mean calculator circuitry to determine:
a first exponential mean value associated with the first time-frequency bin based on a first magnitude of the audio signal associated with the first time-frequency bin; and
a second exponential mean value associated with the second time-frequency bin based on a second magnitude of the audio signal associated with the second time-frequency bin and the first exponential mean value;
bin normalizer circuitry to normalize the first time-frequency bin based on the second exponential mean value;
fingerprint generator circuitry to generate a fingerprint of the audio signal based on normalized time-frequency bins, wherein the normalized time-frequency bins include the normalized first time-frequency bin;
subfingerprint circuitry to generate a subfingerprint by selecting energy extrema of the normalized time-frequency bins, wherein selecting the energy extrema comprises selecting a group of the normalized time-frequency bins with the highest normalized energy values, and the fingerprint including the subfingerprint; and
memory manager circuitry to, after the second exponential mean value is determined, discard the first group of the time-frequency bins from memory.
2. The apparatus of claim 1 , wherein the first time-frequency bin and the second time-frequency bin are in a same frequency band of the audio signal.
3. The apparatus of claim 1 , wherein the bin normalizer circuitry is further to normalize the first time-frequency bin based a third exponential mean value associated with a third time-frequency bin, the third time-frequency bin in the second audio segment and proximate to the second time-frequency bin.
4. The apparatus of claim 1 , wherein the memory manager circuitry is further to, after the mean calculator circuitry determines the second exponential mean value, discard the first exponential mean value from the memory.
5. The apparatus of claim 1 , wherein each time-frequency bin of the first group of time-frequency bins is a unique combination of (1) a time period of the audio signal and (2) a frequency band of the audio signal.
6. A method comprising:
transforming, via at least one processor, an audio signal into a frequency domain including a plurality of time-frequency bins, each of the time-frequency bins corresponding to an intersection of a frequency bin and a time bin and contains a portion of the audio signal;
dividing, via at least one processor, the audio signal into a plurality of audio segments, each audio segment including a plurality of time-frequency bins;
determining, via at least one processor, a first audio segment having a first group of time-frequency bins including a first time-frequency bin;
determining, via at least one processor, a first exponential mean value associated with the first time-frequency bin based on a first magnitude of the audio signal associated with the first time-frequency bin;
determining, via at least one processor, a second audio segment having a second group of time-frequency bins including a second time-frequency bin;
determining, via at least one processor, a second exponential mean value associated with the second time frequency bin based on a second magnitude of the audio signal associated with the second time frequency bin and the first exponential mean value;
normalizing, via at least one processor, the first time-frequency bin based on the second exponential mean value;
generating, via at least one processor, a fingerprint of the audio signal based on normalized time-frequency bins, wherein the normalized time-frequency bins include the normalized first time-frequency bin;
generating, via at least one processor, a subfingerprint by selecting energy extrema of the normalized time-frequency bins, wherein selecting the energy extrema comprises selecting a group of the normalized time-frequency bins with the highest normalized energy values, and the fingerprint including the subfingerprint; and
discarding, via at least one processor, from memory the first group of time-frequency bins in response to determination of the normalized the first time-frequency bin.
7. The method of claim 6 , wherein the first time-frequency bin and the second time-frequency bin are in a same frequency band of the audio signal.
8. The method of claim 6 , wherein the normalization of the first time-frequency bin is further based a third exponential mean value associated with a third time-frequency bin, the third time-frequency bin in the second audio segment and proximate to the second time-frequency bin.
9. The method of claim 6 , wherein each time-frequency bin of the first group of time-frequency bins is a unique combination of (1) a time period of the audio signal and (2) a frequency band of the audio signal.
10. The method of claim 6 , further comprising discarding the first exponential mean value in response to the determination of the second exponential mean value.
11. A non-transitory computer readable medium comprising instructions, which when executed, cause a processor to:
transform an audio signal into a frequency domain including a plurality of time-frequency bins, each of the time-frequency bins corresponding to an intersection of a frequency bin and a time bin and contains a portion of the audio signal;
determine a first audio segment having a first group of time-frequency bins including a first time-frequency bin;
determine a first exponential mean value associated with the first time-frequency bin based on a first magnitude of the audio signal associated with the first time-frequency bin;
determine a second audio segment having a second group of time-frequency bins including a second time-frequency bin;
determine a second exponential mean value associated with the second time-frequency bin based on a second magnitude of the audio signal associated with the second time-frequency bin and the first exponential mean value;
normalize the first time-frequency bin based on the second exponential mean value;
generate a fingerprint of the audio signal based on normalized time-frequency bins, wherein the normalized time-frequency bins include the normalized first time-frequency bin;
generate a subfingerprint by selecting energy extrema of the normalized time-frequency bins, wherein selecting the energy extrema comprises selecting a group of the normalized time-frequency bins with the highest normalized energy values, and the fingerprint including the subfingerprint; and
discard from memory the first group of time-frequency bins in response to determination of the normalized first time-frequency bin.
12. The non-transitory computer readable medium of claim 11 , wherein the first time-frequency bin and the second time-frequency bin are in a same frequency band of the audio signal.
13. The non-transitory computer readable medium of claim 11 , wherein the normalization of the first time-frequency bin is further based a third exponential mean value associated with a third time-frequency bin, the third time-frequency bin in the second audio segment and proximate to the second time-frequency bin.
14. The non-transitory computer readable medium of claim 11 , wherein the instructions further cause the processor to discard the first exponential mean value in response to the determination of the second exponential mean value.