METHOD AND RHYTHM EXTRACTOR FOR DETECTING AND ISOLATING RHYTHMIC SIGNAL FEATURES FROM AN INPUT SIGNAL USING THE WAVELET PACKET TRANSFORM

• US 20100114813A1
• Filed: 10/20/2009
• Published: 05/06/2010
• Est. Priority Date: 10/20/2008
• Status: Active Grant

• Current Assignee: Neurochip Corporation
• Original Assignee: Neurochip Corporation
• Inventors: Bardakjian, Berj L., Zalay, Osbert C.
• Granted Patent: US 8,315,970 B2
• Time in Patent Office: Days
• US Class Current: 706/58
• CPC Class Codes: A61B 5/374 Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
• CPC Class Codes: A61B 5/4821 Determining level or depth of anaesthesia A61B5/1106 takes precedence
• CPC Class Codes: A61B 5/7203 for noise prevention, reduction or removal
• CPC Class Codes: A61B 5/726 using Wavelet transforms
• CPC Class Codes: G06F 17/148 Wavelet transforms

Abstract

A method of detecting and isolating at least one rhythmic component from a discrete-time input signal, comprises subjecting the input signal to discrete wavelet packet transform multi-resolution analysis; applying wavelet packet basis selection criteria to the result of the analysis to evaluate rhythmic signal features of the input signal; and isolating at least one rhythmic signal component from the input signal based on the evaluation.

Claims

1. A method of detecting and isolating at least one rhythmic signal component from a discrete-time input signal, comprising:

subjecting the input signal to discrete wavelet packet transform multi-resolution analysis;

applying wavelet packet basis selection criteria to the result of the analysis, at each stage of the analysis, to evaluate the wavelet packet transform decomposed signal components of the input signal;

detecting at least one rhythmic signal feature from the input signal based on the evaluation; and

isolating the detected at least one rhythmic signal feature from the input signal.

2. The method of claim 1 wherein said selection criteria are formulated to evaluate the degree of rhythmicity of the signal components and the average power of the signal components.

3. The method of claim 2 wherein the wavelet packet transform and selection criteria are applied recursively and wherein the degree of rhythmicity of the signal components and the average power of the signal components are evaluated at each level of recursion.

4. The method of claim 3 wherein the degree of rhythmicity is based on interpeak interval variance.

5. The method of claim 4 wherein the degree of rhythmicity is based on normalized interpeak interval variance.

6. The method of claim 1 further comprising processing the detected at least one rhythmic signal feature to isolate individual rhythms or rhythmic components therein.

7. The method of claim 5 wherein said processing comprises pruning wavelet packet coefficient nodes or individual wavelet packet coefficients.

8. The method of claim 1 further comprising detecting and isolating at least two rhythmic signal components from the input signal.

9. The method of claim 8 further comprising processing each detected and isolated rhythmic signal component to isolate individual rhythms or rhythmic components therein.

10. The method of claim 9 wherein said processing comprises pruning wavelet packet coefficient nodes or individual wavelet packet coefficients.

11. The method of claim 9 wherein said processing comprises performing independent component analysis on the isolated at least two rhythmic signal components.

12. The method of claim 9 further comprising combining the isolated at least two rhythmic signal components to form a contiguous rhythm.

13. The method of claim 12 further comprising subjecting at least two contiguous rhythms to independent component analysis.

14. The method of claim 11 wherein independent component analysis is applied to separate at least one rhythmic artifact and at least one genuine source rhythm, if present together.

15. The method of claim 1 further comprising prior to the subjecting, attenuating large spikes or artifacts in the input signal.

16. A method of isolating at least one rhythmic signal component from a discrete-time input signal comprising:

decomposing the input signal using the discrete wavelet packet transform to create a wavelet packet tree;

pruning the wavelet packet tree; and

following pruning, selecting one or more coefficients or coefficient nodes of the wavelet packet tree containing rhythmic signal features thereby to isolate at least one rhythmic signal component of the input signal.

17. The method of claim 16 wherein during pruning, coefficient nodes of the wavelet packet tree below a threshold value are isolated.

18. The method of claim 17 wherein said threshold value is representative of temporal variability of a given time series of the input signal.

19. The method of claim 18 wherein said threshold value is based on the variance of peak-to-peak intervals in the given time series.

20. The method of claim 16 further comprising reconstructing a time series from each selected coefficient or coefficient node.

21. The method of claim 16 further comprising:

ordering the selected coefficients or coefficient nodes according to a metric;

partitioning the ordered coefficients or coefficient nodes;

combining reconstructions of the coefficients or coefficient nodes of the partitions into time series; and

using the time series to construct virtual channels for independent component analysis.

23. The method of claim 16 further comprising prior to the subjecting, attenuating large spikes or artifacts in the input signal.

22. The method of claim 22 wherein said metric is frequency.

24. A method of isolating rhythmic signal features of a discrete-time signal comprising:

subjecting the input signal to wavelet packet transforming and thresholding to evaluate the coefficients or coefficient nodes to be pruned with respect to rhythmicity; and

separating coefficients or coefficient nodes for independent preservation and time series reconstruction of rhythmic signal features.

25. The method of claim 24 wherein said thresholding is based on interpeak interval variance.

26. A computer readable medium embodying computer readable code for detecting and isolating at least one rhythmic component from a discrete-time input signal, said computer readable code comprising:

program code for subjecting the input signal to discrete wavelet packet transform multi-resolution analysis;

program code for applying wavelet packet basis selection criteria to the result of the analysis, at each stage of the analysis, to evaluate the wavelet packet decomposed signal components of the input signal;

program code for detecting at least one rhythmic signal component based on the evaluation; and

program code for isolating the detected at least one rhythmic signal component from the input signal.

27. A computer readable medium embodying computer readable code for isolating at least one rhythmic signal component from a discrete-time input signal comprising:

program code for decomposing the input signal using the discrete wavelet packet transform to create a wavelet packet tree;

program code for pruning the wavelet packet tree; and

program code for, following pruning, selecting one or more coefficients or coefficient nodes of the wavelet packet tree containing rhythmic signal features thereby to isolate at least one rhythmic signal component of the input signal.

28. A computer readable medium embodying computer readable code for isolating rhythmic signal features of a discrete-time signal comprising:

program code for subjecting the input signal to wavelet packet transforming and thresholding to evaluate the coefficients or coefficient nodes to be pruned with respect to rhythmicity; and

program code for separating coefficients or coefficient nodes for independent preservation and time series reconstruction of rhythmic signal features.

29. A rhythmic signal extractor for detecting and isolating at least one rhythmic signal component from a discrete-time input signal, comprising:

a discrete wavelet packet transform multi-resolution analyzer for analyzing the input signal and evaluating rhythmic signal features of the input signal; and

a signal component isolator for isolating at least one rhythmic signal component from the input signal based on the evaluation.