Audio identification using wavelet-based signatures

摘要

A method is described that includes producing an audio spectrogram from a target sample, generating a number of fingerprints based on the audio spectrogram, comparing the series of fingerprints to samples in a data repository using wavelet coefficients, and identifying the target sample based on the matches found in the data repository.

主权项

1. A method comprising:
generating, by a device including a processor, a spectrogram of a first audio sample, wherein the spectrogram is generated using a first time separation window that is constrained to a predefined range of values from a second time separation window value employed in other spectrograms of other audio samples; computing, by the device, wavelet coefficients for the spectrogram of the first audio sample; comparing, by the device using a first searching algorithm, the computed wavelet coefficients for the spectrogram of the first audio sample to stored wavelet coefficients for the other spectrograms of other audio samples to identify a subset of the other spectrograms, respectively having a plurality of wavelet coefficients that match a plurality of the wavelet coefficients of the spectrogram of the first audio sample, in which to employ a second searching algorithm; and identifying, by the device using the second searching algorithm, from the subset of the spectrograms, a second audio sample that corresponds to one of the subset of the other spectrograms that matches the spectrogram.