Audio fingerprint extraction method and device

1. An audio fingerprint extraction method, comprising:

converting an audio signal to a spectrogram;

determining one or more characteristic points in the spectrogram;

in the spectrogram, determining one or more masks for the one or more characteristic points; wherein each of the one or more masks comprise a plurality of spectrum regions;

determining mean energy of each of the plurality of spectrum regions;

determining one or more audio fingerprint bits according to the mean energy of the plurality of spectrum regions in the one or more masks;

judging credibility of the one or more audio fingerprint bits to determine one or more weight bits;

combining the one or more audio fingerprint bits and the one or more weight bits to obtain an audio fingerprint;

wherein the plurality of spectrum regions comprised in the one or more masks are symmetrical distributed; and

wherein the plurality of spectrum regions comprised in the one or more masks have an identical frequency range, and/or an identical time range, and/or are centrosymmetric with the characteristic point as a center.

2. The audio fingerprint extraction method according to claim 1 , wherein the converting an audio signal to a spectrogram comprises: converting the audio signal to a two-dimensional time-frequency spectrogram by fast Fourier transformation; wherein a value of each point in the spectrogram indicates energy of the audio signal.

3. The audio fingerprint extraction method according to claim 2 , wherein the converting an audio signal to a spectrogram further comprises: processing the spectrogram by MEL transformation.

4. The audio fingerprint extraction method according to claim 2 , wherein the converting an audio signal to a spectrogram further comprises: processing the spectrogram by human auditory system filtering.

5. The audio fingerprint extraction method according to claim 2 , wherein each of the one or more characteristic points is a fixed point in the spectrogram.

6. The audio fingerprint extraction method according to claim 5 , wherein each of the one or more characteristic points is a point whose frequency is equal to one of a plurality of pre-set frequency values.

7. The audio fingerprint extraction method according to claim 2 , wherein each of the one or more characteristic points is a maximal energy point in the spectrogram, or each of the one or more characteristic points is a minimal energy point in the spectrogram.

8. The audio fingerprint extraction method according to claim 1 , wherein the mean energy of each of the plurality of spectrum regions is a mean value of energy values of all points comprised in the spectrum region.

9. The audio fingerprint extraction method according to claim 1 , wherein the determining one or more audio fingerprint bits according to mean energy of the plurality of spectrum regions in the one or more masks comprises:

determining one of the one or more audio fingerprint bits according to a difference value of the mean energy of the plurality of spectrum regions comprised in one mask.

10. The audio fingerprint extraction method according to claim 9 , wherein the judging credibility of the one or more audio fingerprint bits to determine one or more weight bits comprises:

judging whether an absolute value of the difference value attains or exceeds a pre-set weight bit threshold, if it attains or exceeds the pre-set weight bit threshold, determining the audio fingerprint bit to be a strong bit, otherwise determining the audio fingerprint bit to be a weak bit; determining the weight bit according to the audio fingerprint bit is the strong bit or the weak bit.

11. The audio fingerprint extraction method according to claim 10 , wherein the weight bit threshold is a fixed value, or a value based on the difference value, or a scale value.

12. The audio fingerprint extraction method according to claim 1 , wherein the audio fingerprint extraction method further comprises:

dividing the audio signal into a plurality of audio sub-signals according to time;

extracting the audio fingerprints of the plurality of audio sub-signals; and

combining the audio fingerprints of the plurality of audio sub-signals to obtain the audio fingerprint of the audio signal.

13. The audio fingerprint extraction method according to claim 1 , wherein after combining the one or more audio fingerprint bits and the one or more weight bits to obtain an audio fingerprint, the audio fingerprint extraction method further comprises:

storing the audio fingerprint in an audio fingerprint library.

14. An audio fingerprint extraction device, comprising:

a processor; and

a memory communicatively coupled to the processor and storing executable instructions that upon execution by the processor cause the processor to perform operations, the operations comprising:

converting an audio signal to a spectrogram;

determining one or more characteristic points in the spectrogram;

in the spectrogram, determining one or more masks for the one or more characteristic points; wherein each of the one or more masks comprise a plurality of spectrum regions;

determining mean energy of each of the plurality of spectrum regions;

determining one or more audio fingerprint bits according to the mean energy of the plurality of spectrum regions in the one or more masks;

judging credibility of the one or more audio fingerprint bits to determine one or more weight bits;

combining the one or more audio fingerprint bits and the one or more weight bits to obtain an audio fingerprint;

wherein the plurality of spectrum regions comprised in the one or more masks are symmetrical distributed; and

wherein the plurality of spectrum regions comprised in the one or more masks have an identical frequency range, and/or an identical time range, and/or are centrosymmetric with the characteristic point as a center.

15. The audio fingerprint extraction device according to claim 14 , wherein the mean energy of each of the plurality of spectrum regions is a mean value of energy values of all points comprised in the spectrum region.

16. The audio fingerprint extraction device according to claim 14 , wherein the determining one or more audio fingerprint bits according to mean energy of the plurality of spectrum regions in the one or more masks comprises:

determining one of the one or more audio fingerprint bits according to a difference value of the mean energy of the plurality of spectrum regions comprised in one mask.

17. The audio fingerprint extraction device according to claim 16 , wherein the judging credibility of the one or more audio fingerprint bits to determine one or more weight bits comprises:

judging whether an absolute value of the difference value attains or exceeds a pre-set weight bit threshold, if it attains or exceeds the pre-set weight bit threshold, determining the audio fingerprint bit to be a strong bit, otherwise determining the audio fingerprint bit to be a weak bit; determining the weight bit according to the audio fingerprint bit is the strong bit or the weak bit.

18. A non-transitory computer readable storage medium, configured for storing computer readable instructions that upon execution; by a computer cause the computer to perform operations, the operations comprising:

converting an audio signal to a spectrogram;

determining one or more characteristic points in the spectrogram;

in the spectrogram, determining one or more masks for the one or more characteristic points; wherein each of the one or more masks comprise a plurality of spectrum regions;

determining mean energy of each of the plurality of spectrum regions;

determining one or more audio fingerprint bits according to the mean energy of the plurality of spectrum regions in the one or more masks;

judging credibility of the one or more audio fingerprint bits to determine one or more weight bits;

combining the one or more audio fingerprint bits and the one or more weight bits to obtain an audio fingerprint;

wherein the plurality of spectrum regions comprised in the one or more masks are symmetrical distributed; and

wherein the plurality of spectrum regions comprised in the one or more masks have an identical frequency range, and/or an identical time range, and/or are centrosymmetric with the characteristic point as a center.