Method and apparatus for identifying sound in a composite sound signal

• US 6,182,018 B1
• Filed: 08/25/1998
• Issued: 01/30/2001
• Est. Priority Date: 08/25/1998
• Status: Expired due to Fees

• Current Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Original Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Inventors: Lei, Sheau-Fang, Charette, Francois, Hsueh, Keng D., Tran, Vy
• Primary Examiner(s): Hoff, Marc S.
• Assistant Examiner(s): VO, HIEN XUAN
• Application Number:
  US09/140,071
• Time in Patent Office: 889 Days
• Field of Search: 702/66, 702/70, 702/71, 702/73, 702/57, 702/191, 702/112, 702/124, 702/189, 702/190, 735/84, 736/02, 706/20, 706/22, 324/76.11, 324/76.12
• US Class Current: 702/66
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 2218/06 by applying a scale-space analysis, e.g. using wavelet analysis

Abstract

Impulsive sound components in a composite sound signal are separated using wavelet analysis and sorting of wavelet coefficient sets according to statistical parameters of each respective coefficient set. Each entire coefficient set is either included or excluded from each respective separated component based on the statistical parameter. Once the impulsive sound component is isolated, it is compared with reference sound information in order to classify the sound according to its potential causes.

Claims

1. A method of identifying a sound in a composite sound signal, comprising the steps of:

converting said composite sound signal to an electrical signal;

decomposing said electrical signal using a wavelet transform to produce a plurality of sets of wavelet coefficients, each set of wavelet coefficients corresponding to a respective time-versus-frequency span;

determining a respective statistical parameter for each set of wavelet coefficients;

re-synthesizing an impulsive signal using an inverse wavelet transform applied to selected ones of said sets of wavelet coefficients, said selected ones being selected in response to said respective statistical parameters; and

comparing said impulsive signal with a plurality of reference signals to classify said sound with respect to said reference signals.

2. The method of claim 1 wherein said plurality of reference signals includes sounds caused by squeaks and rattles in motor vehicles.

3. The method of claim 1 wherein said comparing step is comprised of signature analysis wherein similarities between said impulsive signal and each of said reference signals are ranked.

4. The method of claim 3 wherein said signature analysis is comprised of determining one or more statistical parameters of said impulsive sound and comparing said statistical parameters with corresponding statistical parameters of said reference signals.

5. The method of claim 1 wherein said comparing step is comprised of a neural network classification.

6. A method of identifying a sound in a composite sound signal, comprising the steps of:

converting said composite sound signal to an electrical signal;

decomposing said electrical signal using a wavelet transform to produce a plurality of sets of wavelet coefficients, each set of wavelet coefficients corresponding to a respective time-versus-frequency span;

determining a respective statistical parameter for each set of wavelet coefficients;

identifying selected sets of wavelet coefficients as impulsive sets in response to said respective statistical parameters; and

comparing said impulsive sets with a plurality of reference sets of wavelet coefficients to classify said sound with respect to said reference sets of wavelet coefficients, wherein each reference set of wavelet coefficients corresponds to a respective sound classification.

7. The method of claim 6 wherein said plurality of reference sets of wavelet coefficients includes decompositions of sounds caused by squeaks and rattles in motor vehicles.

8. The method of claim 6 wherein said comparing step is comprised of signature analysis wherein similarities between said impulsive sets and each of said reference sets of wavelet coefficients are ranked.

9. The method of claim 8 wherein said signature analysis is comprised of determining one or more statistical parameters of said impulsive sets and comparing said statistical parameters with corresponding statistical parameters of said reference sets of wavelet coefficients.

10. The method of claim 6 wherein said comparing step is comprised of a neural network classification.

11. Apparatus for identifying a sound in a composite sound signal, comprising the steps of:

a transducer converting said composite sound signal to an electrical signal;

a wavelet transformer decomposing said electrical signal into a plurality of wavelet coefficient sets;

a statistical parameter calculator calculating a statistical parameter for each wavelet coefficient set;

a separator forming an impulsive group of wavelet coefficient sets in response to said statistical parameters;

an inverse wavelet transformer for synthesizing an impulsive signal from said impulsive group of wavelet coefficient sets; and

a classifier comparing said impulsive signal with a plurality of reference signals to classify said sound with respect to said reference signals.

12. The apparatus of claim 11 wherein said classifier is comprised of a signature analyzer wherein similarities between said impulsive signal and each of said reference signals are ranked.

13. The apparatus of claim 12 wherein said signature analyzer comprises:

statistics means for determining statistical parameters of said impulsive signal; and

comparing means for comparing said determined statistical parameters with statistical parameters of said reference signals.

14. The apparatus of claim 11 wherein said classifier is comprised of a neural network.