Acoustic signature recognition and identification

• US 6,173,074 B1
• Filed: 09/30/1997
• Issued: 01/09/2001
• Est. Priority Date: 09/30/1997
• Status: Expired due to Term

• Current Assignee: WSOU Investments, LLC (WSOU Holdings, LLC)
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Russo, Anthony Peter
• Primary Examiner(s): MEHTA, BHAVESH M
• Application Number:
  US08/940,321
• Time in Patent Office: 1,197 Days
• Field of Search: 382/190, 382/194, 382/209, 382/100, 382/191, 382/210, 367/135, 367/140, 367/21-23, 367/68, 367/70-73, 367/118, 367/136, 73/579, 73/602, 73/649, 73/658, 73/659, 600/443
• US Class Current: 382/190
• CPC Class Codes: G01H 3/08 Analysing frequencies present in complex vibrations, e.g. comparing harmonics present acoustic presence detection G01V1/001

Abstract

An acoustic signature recognition and identification system receives signals from a sensor placed on a designated piece of equipment. The acoustic data is digitized and processed, via a Fast Fourier Transform routine, to create a spectrogram image of frequency versus time. The spectrogram image is then normalized to permit acoustic pattern recognition regardless of the surrounding environment or magnitude of the acoustic signal. A feature extractor then detects, tracks and characterizes the lines which form the spectrogram. Specifically, the lines are detected via a KY process that is applied to each pixel in the line. A blob coloring process then groups spatially connected pixels into a single signal object. The harmonic content of the lines is then determined and compared with stored templates of known acoustic signatures to ascertain the type of machinery. An alert is then generated in response to the recognized and identified machinery.

Claims

1. A system for recognizing and identifying acoustic signals from a sensor, comprising:

a pre-processor for generating a gray-level image of the acoustic signals;

a feature extractor for detecting, tracking and characterizing at least one line of the gray-level image, said feature extractor calculating a fundamental frequency for each of said at least one line and grouping said at least one line into harmonically related lines based on said fundamental frequency; and

a post-processor for determining a harmonic content of said at least one line and comparing said harmonic content against known acoustic signatures by calculating a weighted score based on said fundamental frequency.

2. The system according to claim 1, wherein said feature extractor includes a line detector for labeling a pixel with a first state and a second state, said first state and said second state being mutually exclusive, wherein said pixel is labeled with said first state if a grey-level value of said pixel is above a first threshold and a KY value of said pixel is above a second threshold.

4. The system according to claim 2, wherein said feature extractor includes a blob-coloring processor for forming into tracked signals all pixels having said first state by taking all spatially connected pixels and grouping them into a single signal object.

5. The system according to claim 2, wherein said feature extractor further includes a signal extender for extending a signal in a given direction by labeling pixels with said second state with said first state when said pixels have a KY value greater than a third threshold.

6. The system according to claim 1, wherein said post-processor includes a classifier for computing a classification score for each group of lines output by a harmonic set extractor by comparing said each group of lines against at least one acoustic signature template.

3. A system for recognizing and identifying acoustic signals from a sensor, comprising:

a pre-Processor for generating a gray-level image of the acoustic signals;

a feature extractor for detecting, tracking and characterizing at least one line of the gray-level image, said feature extractor calculating a fundamental frequency for each of said at least one line and grouping said at least one line into harmonically related lines based on said fundamental frequency; and

a post-processor for determining a harmonic content of said at least one line and comparing said harmonic content against known acoustic signatures

wherein said feature extractor includes a line detector for labeling a pixel with a first state and a second state, said first state and said second state being mutually exclusive, wherein said pixel is labeled with said first state if a grey-level value of said pixel is above a first threshold and a KY value of said pixel is above a second threshold; and

wherein said pixel has an upper set of line segments and a lower set of line segments extending radial from said pixel, said line detector further including;

a first summer for summing grey-level values of all pixels along each of said upper set of line segments and said lower set of line segments, said first summer determining a upper line segment maximum value and a lower line segment maximum value; and

a second summer for summing said upper line segment maximum value and said lower line segment maximum value to determine said KY value.

7. An apparatus for recognizing and identifying acoustic signals from a sensor, comprising:

a spectrum analyzer for generating a spectrogram image of the acoustic signals;

a feature extractor for detecting, tracking and characterizing at least one line of said spectrogram image, said feature extractor calculating a fundamental frequency for each of said at least one line and grouping said at least one line into harmonically related lines based on said fundamental frequency;

a harmonic set extractor for determining a harmonic content of said at least one line; and

a classifier for comparing said harmonic content against known acoustic signature templates by calculating a weighted score based on said fundamental frequency.

8. The apparatus according to claim 7, wherein said feature extractor includes a detector for labeling a pixel with a ON state and a OFF state, wherein said pixel is labeled with said ON state if a grey-level value of said pixel is above a first threshold and a KY value of said pixel is above a second threshold.

10. The apparatus according to claim 7, wherein said feature extractor includes a blob-coloring processor for forming into tracked signals all pixels having a ON state by taking all spatially connected pixels and grouping them into a single signal object.

11. The apparatus according to claim 7, wherein said feature extractor further includes a signal extender for extending a signal in a given direction by labeling pixels having a OFF state with a ON state label when said pixels have a KY value greater than a threshold.

12. The apparatus according to claim 7, wherein said classifier computes a classification score for each group of lines output by said harmonic set extractor by comparing said each group of lines against at least one acoustic signature template.

9. An apparatus for recognizing and identifying acoustic signals from a sensor, comprising:

a spectrum analyzer for generating a spectrogram image of the acoustic signals;

a feature extractor for detecting, tracking and characterizing at least one line of said spectrogram image, said feature extractor calculating a fundamental frequency for each of said at least one line and rouping said at least one line into harmonically related lines based on said fundamental frequency;

a harmonic set extractor for determining a harmonic content of said at least one line; and

a classifier for comparing said harmonic content against known acoustic signature templates

wherein said feature extractor includes a detector for labeling a pixel with a ON state and a OFF state, wherein said pixel is located with said ON state if a grey-level value of said pixel is above a first threshold and a KY value of said pixel is above a second threshold; and

wherein said pixel has an upper set of line segments and a lower set of line segments extending radial from said pixel, said detector further including;

a first summer for summing grey-level values of all pixels along each of said upper set of line segments and said lower set of line segments, said first summer determining a upper line segment maximum value and a lower line segment maximum value; and

a second summer for summing said upper line segment maximum value and said lower line segment maximum value to determine said KY value.

13. A method for recognizing and identifying acoustic signals from a sensor, comprising:

generating a spectrogram image of the acoustic signals;

detecting, tracking and characterizing at least one line of said spectrogram image by comparing grey-level values against a first threshold and KY values against a second threshold;

calculating a fundamental frequency for each of said at least one line;

grouping said at least one line into harmonically related lines based on said fundamental frequency;

determining a harmonic content of said at least one line; and

classifying said harmonic content against known acoustic signature templates by calculating a weighted score using said fundamental frequency.

14. The method according to claim 13, wherein said step of detecting includes the step of labeling a pixel with a first state and a second state, said first state and said second state being mutually exclusive, wherein said pixel is labeled with said first state if a grey-level value of said pixel is above a first threshold and a KY value of said pixel is above a second threshold.

16. The method according to claim 13, wherein said step of detecting includes the step of forming into tracked signals all pixels having a ON state by taking all spatially connected pixels and grouping them into a single signal object.

17. The method according to claim 13, wherein said step of detecting includes the step of extending a signal in a given direction by labeling pixels with a OFF state with a ON state label when said pixels have a KY value greater than a third threshold.

18. The method according to claim 13, wherein said step of classifying includes the step of computing a classification score for each group of lines output by a harmonic set extractor by comparing said each group of lines against at least one acoustic signature template.

15. A method for recognizing and identifying acoustic signals from a sensor, comprising:

generating a spectrogram image of the acoustic signals;

detecting, tracking and characterizing at least one line of said spectrogram image by comparing grey-level values against a first threshold and KY values against a second threshold;

calculating a fundamental frequency for each of said at least one line;

grouping said at least one line into harmonically related lines based on said fundamental frequency;

determining a harmonic content of said at least one line; and

classifying said harmonic content against known acoustic signature templates

wherein a pixel has an upper set of line segments and a lower set of line segments extending radial from said pixel, said step of detecting further including the step of;

summing grey-level values of all pixels along each of said upper set of segments and said lower set of line segments;

determining a upper line segment maximum value and a lower line segment maximum value; and

summing said upper line segment maximum value and said lower line segment maximum value to determined said KY value.

19. A method for recognizing and identifying acoustic signals from a sensor, comprising:

generating a spectrogram image of the acoustic signals;

detecting, tracking and characterizing at least one line of said spectrogram image by comparing grey-level values against a first threshold and KY values against a second threshold;

determining a harmonic content of said at least one line;

classifying said harmonic content against known acoustic signature templates;

summing grey-level values of all pixels along an upper set of line segments and a lower set of line segments;

determining an upper line segment maximum value and a lower line segment maximum value; and

summing said upper line segment maximum value and said lower line segment maximum value to determine a KY value.