Process for identifying audio content

• US 6,453,252 B1
• Filed: 05/15/2000
• Issued: 09/17/2002
• Est. Priority Date: 05/15/2000
• Status: Expired due to Term

• Current Assignee: Creative Technology Ltd.
• Original Assignee: Creative Technology Ltd.
• Inventors: Laroche, Jean
• Primary Examiner(s): Hilten, John S.
• Assistant Examiner(s): Washburn, Douglas N
• Application Number:
  US09/570,679
• Time in Patent Office: 855 Days
• Field of Search: 702/57, 702/71, 702/73, 702/75, 734/321, 116/1, 324/76.12, 324/76.19, 381/124, 386/75, 386/96
• US Class Current: 702/75
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence

Abstract

A fingerprint of an audio signal is generated based on the energy content in frequency subbands. Processing techniques assure a robust identification fingerprint that will be useful for signals altered subsequent to the generation of the fingerprint. The fingerprint is compared to a database to identify the audio signal.

Claims

1. A method of identifying a digital audio signal by monitoring the content of the audio signal, said method comprising the acts of:

selecting a set of frequency subbands of said audio signal, with each frequency having a selected frequency range;

for each subband, generating subband energy signal having a magnitude, in decibels (dB), equal to signal energy in the subband;

forming an energy flux signal for each subband having a magnitude equal to the difference between subband energy signals of neighboring frames;

determining the magnitude of frequency components bins of the energy flux signal for each subband;

forming a fingerprint comprising the magnitudes of the frequency component bins of the energy flux signal for all subbands; and

comparing the fingerprint for the audio file to fingerprints in a database to identify the audio file.

2. The method of claim 1 where said step of generating a subband energy signal comprises the acts of:

for each subband, filtering the audio signal to obtain a filtered signal having only frequency components in the subband; and

calculating the power of the filtered signal.

3. A method of generating a fingerprint for identifying an audio signal, said method comprising the acts of:

selecting a set of frequency subbands of said audio signal, with each frequency having a selected frequency range;

for each subband, generating subband energy signal having a magnitude, in decibels (dB), equal to signal energy in the subband;

forming an energy flux signal for each subband having a magnitude equal to the difference between subband energy signals of neighboring frames;

determining the magnitude of frequency components bins of the energy flux signal for each subband; and

forming a fingerprint comprising the magnitudes of the frequency component bins of the energy flux signal for all subbands.

4. The method of claim 3 where said step of generating a subband energy signal comprises the acts of:

dividing a segment of the signal into overlapping frames;

for each frame, determining the magnitude of frequency bins at different frequencies;

selecting a set of frequency subbands of a desired frequency range;

for each subband and each frame, summing the frequency bins of the frame located within the subband to form a subband energy signal having a magnitude expressed in decibels (dB) for the given frame.

5. The method of claim 3 where said step of generating a subband energy signal comprises the acts of:

for each subband, filtering the audio signal to obtain a filtered signal having only frequency components in the subband; and

calculating the power of the filtered signal.

6. A method of identifying a digital audio signal by monitoring the content of the audio signal, said method comprising the acts of:

dividing a segment of the signal into overlapping frames;

for each frame, determining the magnitude of frequency bins at different of frequencies;

selecting a set of frequency subbands of a desired frequency range;

for each subband and each frame, summing the frequency bins of the frame located within the subband to form a subband energy signal having a magnitude expressed in decibels (dB) for the given frame;

forming an energy flux signal for each subband having a magnitude equal to the difference between subband energy signals of neighboring frames;

determining the magnitude of frequency components bins of the energy flux signal for each subband;

forming a fingerprint comprising the magnitudes of the frequency component bins of the energy flux signal for all subbands;

comparing the fingerprint for the audio file to fingerprints in a database to identify the audio file.

7. The method of claim 6 further comprising the acts of:

smoothing the subband energy signal for each subband to compensate for subsequent alterations of the audio signal.

8. The method of claim 6 further comprising the acts of:

generating local maxima of a parameter of the audio signal; and

locating a fingerprint monitoring period near a local maxima.

9. The method of claim 8 where said act of generating comprises:

generating local maxima of the energy content of the audio signal.

10. A computer program product comprising:

a computer readable storage medium having computer program code embodied therein for forming a fingerprint for identifying an audio file, said computer program code comprising;

program code for causing a computing system to select a set of frequency subbands of said audio signal, with each frequency having a selected frequency range;

for each subband, program code for causing a computing system to generate subband energy signal having a magnitude, in decibels (dB), equal to signal energy in the subband;

program code for causing a computing system to form an energy flux signal for each subband having a magnitude equal to the difference between subband energy signals of adjacent frames;

program code for causing a computing system to determine the magnitude of frequency components bins of the energy flux signal for each subband; and

program code for causing a computing system to form a fingerprint comprising the magnitudes of the frequency component bins of the energy flux signal for all subbands.