Acoustic signature recognition of running vehicles using spectro-temporal dynamic neural network

• US 8,111,174 B2
• Filed: 10/03/2008
• Issued: 02/07/2012
• Est. Priority Date: 10/03/2007
• Status: Expired due to Fees

• Current Assignee: University of Southern California
• Original Assignee: University of Southern California
• Inventors: Berger, Theodore W., Dibazar, Alireza, Lu, Bing
• Primary Examiner(s): La, Anh V
• Application Number:
  US12/245,575
• Publication Number:
  US 20110169664A1
• Time in Patent Office: 1,222 Days
• Field of Search: 340/943, 340/933, 340/988, 340934-937, 340/907, 340/905, 340/917, 340/941, 340/942, 701/117, 701/118, 701/119, 701/120, 367/128, 367/129, 367/906
• US Class Current: 340/943
• CPC Class Codes: G01H 3/08 Analysing frequencies present in complex vibrations, e.g. comparing harmonics present acoustic presence detection G01V1/001
• CPC Class Codes: G08G 1/015 with provision for distinguishing between two or more types of vehicles, e.g. between motor-cars and cycles
• CPC Class Codes: G08G 1/04 using optical or ultrasonic detectors
• CPC Class Codes: G10L 25/30 using neural networks
• CPC Class Codes: G10L 25/51 for comparison or discrimination

Abstract

A method and apparatus for identifying running vehicles in an area to be monitored using acoustic signature recognition. The apparatus includes an input sensor for capturing an acoustic waveform produced by a vehicle source, and a processing system. The waveform is digitized and divided into frames. Each frame is filtered into a plurality of gammatone filtered signals. At least one spectral feature vector is computed for each frame. The vectors are integrated across a plurality of frames to create a spectro-temporal representation of the vehicle waveform. In a training mode, values from the spectro-temporal representation are used as inputs to a Nonlinear Hebbian learning function to extract acoustic signatures and synaptic weights. In an active mode, the synaptic weights and acoustic signatures are used as patterns in a supervised associative network to identify whether a vehicle is present in the area to be monitored. In response to a vehicle being present, the class of vehicle is identified. Results may be provided to a central computer.

Claims

1. An apparatus for identifying running vehicles using acoustic signatures, comprising:

an input sensor configured to capture an acoustic waveform produced by a vehicle source in an area to be monitored and convert the waveform into a digitized electrical signal; and

a processing system configured to

divide the digitized electrical signal into a plurality of frames;

compute at least one spectral feature vector for each frame;

integrate said spectral feature vectors over the plurality of frames to produce a spectro-temporal representation of said acoustic waveform, and

apply values obtained from said spectro-temporal representation as inputs to a learning function to determine an acoustic signature of the vehicle source.

2. The apparatus of claim 1 wherein said learning function comprises nonlinear Hebbian learning (NHL).

3. The apparatus of claim 1 wherein said computing at least one spectral feature vector comprises filtering each frame using gammatone filterbanks (GTF).

4. The apparatus of claim 1 wherein the dividing the electrical signal into a plurality of frames comprises using a Hamming window.

5. The apparatus of claim 1 wherein

the input sensor is configured to capture a plurality of acoustic waveforms produced by a respective plurality of vehicle sources;

and

the processing system is further configured, in a training mode, to determine at least one acoustic signature for a respective each one of said plurality of captured acoustic waveforms.

6. The apparatus of claim 5 wherein said each at least one acoustic signature is associated in said training mode with a corresponding identity of a different vehicle class.

7. The apparatus of claim 6 wherein:

said input sensor is configured, in an active mode, to capture a further acoustic waveform from an acoustic source in said area; and

said processing system is configured in said active mode to

determine, based on comparing information in said further acoustic waveform with one or more of said at least one acoustic signatures, whether said further acoustic waveform is representative of a running vehicle; and

identify, in response to determining that said further acoustic waveform is representative of a running vehicle, an associated vehicle class.

8. The apparatus of claim 2 wherein said Nonlinear Hebbian learning function is performed at least in part as the following steps:

9. The apparatus of claim 8 wherein

10. The apparatus of claim 1 further comprising a wireless transmitter, wherein the processor is configured to send, to a central computer via the wireless transmitter, information sufficient to identify a class of vehicle.

11. The apparatus of claim 1 wherein a collective duration of said plurality of frames is at least 200 ms.

12. The apparatus of claim 7 wherein said determining whether said acoustic waveform is representative of a running vehicle comprises using a radial basis function neural network.

13. A method for identifying running vehicles using acoustic signatures, comprising:

capturing an acoustic waveform produced by a vehicle source in an area to be monitored;

amplifying the acoustic waveform;

converting the waveform into a digitized electrical signal;

dividing the digitized electrical signal into a plurality of frames;

computing at least one spectral feature vector for each frame;

integrating said spectral feature vectors over the plurality of frames to produce a spectro-temporal representation of said acoustic waveform; and

applying values obtained from said spectro-temporal representation as inputs to a learning function to determine an acoustic signature of the vehicle source.

14. The method of claim 13 wherein said learning function comprises Nonlinear Hebbian learning (NHL).

15. The method of claim 13 wherein said computing at least one spectral feature vector comprises filtering each frame using gammatone filterbanks (GTF).

16. The method of claim 13 wherein said dividing the digitized electrical signal into a plurality of frames comprises using a Hamming window.

17. The method of claim 14 wherein said nonlinear Hebbian learning function is performed at least in part as the following steps:

18. The method of claim 17 wherein

19. The method of claim 13 further comprising sending, to a central computer via a wireless transmitter, information sufficient to identify a class of vehicle.

20. The method of claim 13 further comprising

capturing a plurality of acoustic waveforms from a respective plurality of vehicle sources;

determining at least one acoustic signature for a respective each one of said plurality of captured acoustic waveforms.

21. The method of claim 20 wherein said each at least one acoustic signature is associated in a training mode with a corresponding identity of a different vehicle class.

22. The method of claim 21 further comprising

capturing, in an active mode, a further acoustic waveform from an acoustic source in said area;

determining, based on comparing information in said further acoustic waveform with one or more of said at least one acoustic signatures, whether said further acoustic waveform is representative of a running vehicle; and

identifying, in response to determining that said further acoustic waveform is representative of a running vehicle, an associated vehicle class.

23. The method of claim 13 wherein a collective duration of said plurality of frames is at least 200 milliseconds.

24. The method of claim 22 wherein said determining whether said acoustic waveform is representative of a running vehicle comprises using a radial basis function neural network.

25. A system for identifying running vehicles in an area to be monitored using acoustic signatures, comprising:

at least one local sensor, the local sensor comprising

an input sensor configured to capture an acoustic waveform produced by a vehicle source, and convert the waveform into an electrical signal,

a processing system configured to divide the electrical signal into frames;

compute a spectral feature vector for each frame;

integrate said spectral feature vectors over the plurality of frames to produce a spectro-temporal representation of said acoustic waveform, apply values obtained from said spectro-temporal representation as inputs to a learning function to determine an acoustic signature of the vehicle source, and identify, based on the determined acoustic signature, the vehicle source; and

a command center comprising

a central computer configured to receive a message from said at least one local sensor, said message comprising information sufficient to identify said source.

26. The system of claim 25 wherein the central computer is further configured to identify, in response to said message, a location of said at least one local sensor, and to perform an action in response to said message.

27. The system of claim 26 wherein said action comprises triggering an alarm.

28. The system of claim 25 wherein said at least one local sensor comprises a wireless transmitter for transmitting said message to said central computer.

29. The system of claim 25 wherein said at least one local sensor comprises a plurality of local sensors, and wherein said plurality of local sensors are in different locations on said area to be monitored.

30. An apparatus for identifying running vehicles using acoustic signatures, comprising:

input sensor means for capturing an acoustic waveform produced by a vehicle source in an area to be monitored and converting the waveform into a digitized electrical signal; and

processing means for dividing the digitized electrical signal into a plurality of frames, computing at least one spectral feature vector for each frame, integrating said spectral feature vectors over the plurality of frames to produce a spectro-temporal representation of said acoustic waveform, and applying values obtained from said spectro-temporal representation as inputs to a learning function to determine an acoustic signature of the vehicle source.