Biologically-based signal processing system applied to noise removal for signal extraction

• US 20020023066A1
• Filed: 06/25/2001
• Published: 02/21/2002
• Est. Priority Date: 06/26/2000
• Status: Active Grant

• Current Assignee: Lawrence Livermore National Security LLC (Government of the United States of America)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Petrich, Loren I., Fu, Chi Yung
• Granted Patent: US 6,763,339 B2
• Time in Patent Office: Days
• US Class Current: 706/22
• CPC Class Codes: G06F 2218/06 by applying a scale-space analysis, e.g. using wavelet analysis
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: H03H 17/0211 using specific transformation algorithms, e.g. WALSH functions, Fermat transforms, Mersenne transforms, polynomial transforms, Hilbert transforms correlation computation G06F17/156
• CPC Class Codes: H03H 2222/04 using neural networks

Abstract

The method and system described herein use a biologically-based signal processing system for noise removal for signal extraction. A wavelet transform may be used in conjunction with a neural network to imitate a biological system. The neural network may be trained using ideal data derived from physical principles or noiseless signals to determine to remove noise from the signal.

Claims

1. A method comprising:

receiving a signal corrupted with noise;

decomposing said signal using a wavelet transform;

modifying wavelet coefficients of said wavelet tranform to reject noise;

re-synthesizing said decomposed signal; and

inputting said re-synthesized signal into a neutral network to further filter out the noise from the signal and recover a clean signal.

2. A method comprising:

receiving a signal corrupted with noise;

decomposing said signal using a wavelet transform to produce a plurality of wavelet coefficients;

evaluating each of said plurality of wavelet coefficients separately and determining acceptance of each of the plurality of wavelet coefficients independently;

re-synthesizing said signal using an inverse transform; and

inputting said decomposed signal into a neutral network to recover a clean signal.

3. A method comprising:

receiving a signal corrupted with noise;

transforming said signal into the wavelet domain at substantially full resolution;

thresholding said signal;

iteratively determining a self-consistent transform of said signal to act as a filter;

recover said signal using an inverse transform; and

inputting said signal into a neutral network to further recover a clean signal.

4. The method of claim 3,

wherein said thresholding step may be implemented with at least two values.

5. The method of claim 4,

wherein said at least two values include a threshold value at a lowest level and a multiplier for finding a level'"'"'s threshold from the previous level.

6. The method of claim 3,

wherein said thresholding step uses threshold autoscaling for each threshold level.

7. The method of claim 3, wherein said neural network is a projection neural network.

8. The method of claim 3, further comprising:

a Savitzky-Golay filter may filter the signal after processing by the neural network and identify first and second derivatives for further processing.

9. The method of claim 3, further comprising:

training said neural network using a set of data to enable the neural network to recognize signals.

10. The method of claim 9, wherein said training data may include noiseless signal data or ideal data derived from physical principles.

11. A system comprising:

a wavelet transformer capable of decomposing a signal; and

a neural network operatively coupled to said wave transformer and together capable of filtering out noise from the signal and outputting a clean signal.

12. The system of claim 11,

wherein said neural network has been trained using noiseless signal data or ideal data derived from physical principle and such trained neural network after sufficient training will generalize its capability to recover signal from unseen corrupted data.

13. The system of claim 11,

wherein said wavelet transformer is capable of producing a plurality of wavelet coefficients and evaluating each of said wavelet coefficients separately and determining acceptance of each of the plurality of wavelet coefficients independently.

14. The system of claim 11,

wherein said neural network does not function as a classifier.

16. The system of claim 15, wherein said wavelet transformer is capable of producing a plurality of wavelet coefficients in the wavelet domain and evaluating each of said wavelet coefficients separately and determining acceptance of each of the plurality of wavelet coefficients independently.

17. The system of claim 15, wherein said neural network functions as filter.

18. The system of claim 15, further comprising:

a Savitzky-Golay filter operatively coupled to said neural network.

15. A system comprising:

a wavelet transformer capable of transforming a first signal into the wavelet domain, thresholding said first signal, finding a self-consistent transform of said first signal through a plurality of iterations, and producing a filtered signal from said first signal through an inverse transform; and

a neural network capable of processing said filtered signal to obtain a clean signal.

19. A system comprising:

a means for decomposing a signal using a wavelet transform;

a means to modify wavelet coefficients of said wavelet transform to remove noise;

a means for re-synthesizing said decomposed signal; and

a means for inputting said re-synthesized signal into a neutral network to filter out the noise from the signal and recover a clean signal.

20. A system comprising:

a means for receiving a signal corrupted with noise;

a means for decomposing said signal using a wavelet transform to produce a plurality of wavelet coefficients;

a means for evaluating each of said plurality of wavelet coefficients separately and determining acceptance of each of the plurality of wavelet coefficients independently;

a means for re-synthesizing said signal using an inverse transform; and

a means for inputting said decomposed signal into a neutral network to recover a clean signal.