Method and apparatus for adaptively reducing the level of noise in an acquired signal

• US 5,924,980 A
• Filed: 03/11/1998
• Issued: 07/20/1999
• Est. Priority Date: 03/11/1998
• Status: Expired due to Term

• Current Assignee: Siemens AG
• Original Assignee: Siemens Corporate Research Incorporated (Siemens AG)
• Inventors: Coetzee, Frans Martin
• Primary Examiner(s): Kamm, William E.
• Application Number:
  US09/038,334
• Time in Patent Office: 496 Days
• Field of Search: 128/901, 128/924-925, 600/300, 600/415, 600/475, 600/508, 600/509
• US Class Current: 600/300
• CPC Class Codes: A61B 5/14551 for measuring blood gases
• CPC Class Codes: A61B 5/7257 using Fourier transforms
• CPC Class Codes: G06F 2218/02 Preprocessing
• CPC Class Codes: Y10S 128/901 Suppression of noise in electric signal

Abstract

An adaptive filtering method and apparatus for reducing the level of an undesired noise component in an acquired physiological signal having a desired signal component. The acquired physiological signal is applied to one input of the adaptive filter, and a synthetic reference signal that is modeled so as to exhibit a correlation with the desired signal component is applied to another input of the adaptive filter. Thereafter, in a feedback manner, the adaptive filter iteratively adjusts the modeled synthetic reference signal so as to progressively generate a more accurate approximation of the desired signal component in the adaptive filter, which approximation becomes a reconstruction of the acquired physiological signal wherein the level of the undesired noise component is reduced.

Claims

1. A method for reducing the level of an undesired noise component in an acquired signal having a desired signal component, comprising the following steps:

acquiring a signal;

generating a synthetic reference signal that is modeled to exhibit a correlation with the desired signal component of the acquired signal;

applying to one input of an adaptive filter arrangement said acquired signal, and to another input of said adaptive filter arrangement said synthetic reference signal; and

operating said adaptive filter arrangement in a feedback manner, so that the adaptive filter arrangement iteratively adjusts the applied synthetic reference signal so as to progressively generate as an output of the filter arrangement a more accurate approximation of the desired signal component of the acquired signal.

2. The method of claim 1, wherein said generating step comprises:

generating said reference signal by processing a predetermined kernel function that is representative of a pulsatile cycle of the desired signal component of the acquired signal.

3. The method of claim 1, wherein the operating step of said adaptive filter comprises:

shifting and weighting of said synthetic reference signal using a multistage lattice prediction filter to generate backward prediction error signals; and

combining said acquired signal with shifted and weighted version of said backward prediction error signals, to iteratively build-up a reconstruction of the desired component of said acquired signal.

4. The method of claim 1, including a further step of:

confidence processing of the acquired signal so as to identify sections thereof having a high confidence of being relatively noise-free.

5. The method of claim 4, wherein said acquiring step acquires a physiological signal from a patient.

6. The method of claim 5, including a further step of:

processing of said high confidence sections of the acquired physiological signal to develop values that are used to adjust the generation of the synthetic reference signal so that it exhibits a greater correlation with the desired signal component of the acquired physiological signal.

7. The method of claim 6, wherein:

said acquiring step comprises acquiring first and second ones of acquired physiological signals; and

said confidence processing step comprises subjecting the first and second ones of said acquired physiological signals to a correlation test, and then

identifying sections of each of said acquired physiological signals that exhibit a correlation that exceeds a predetermined threshold level as being relatively noise-free, and identifying sections of each of said acquired physiological signals that exhibit a correlation that does not exceed the predetermined threshold level as being relatively noisy.

8. The method of claim 6, wherein said processing of said high confidence sections of said acquired physiological comprises FFT processing of said high confidence sections to develop a value representative of at least a fundamental frequency component of the desired signal component of the acquired physiological signal, which value is used to adjust the modeling of the synthetic reference signal so as to increase its correlation with the desired signal component of the acquired physiological signal.

9. The method of claim 6, wherein said confidence processing comprises:

subjecting said acquired physiological signal to an amplitude threshold test; and

identifying sections of said acquired physiological signal that have an amplitude that does not exceed a predetermined amplitude threshold level as being relatively noise-free, and

identifying sections of said acquired physiological signal that have an amplitude that does exceed the predetermined amplitude threshold level as being relatively noisy.

10. The method of claim 9, wherein:

said acquiring step comprises acquiring first and second ones of acquired physiological signals; and

said confidence processing step comprises subjecting the first and second ones of said acquired physiological signals to a correlation test, and then

identifying sections of each of said acquired physiological signals that exhibit a correlation that exceeds a predetermined threshold level as being relatively noise-free, and identifying sections of each of said acquired physiological signals that exhibit a correlation that does not exceed the predetermined threshold level as being relatively noisy.

11. The method of claim 10, wherein said confidence processing step comprises the further steps of:

developing an error code stream comprising a sequence of error codes that identify corresponding successive sections of said acquired physiological signals that have been identified as being either relatively noise-free or noisy; and

morphological processing of said error code stream to increase the degree of confidence that sections of said acquired physiological signal identified as being relatively noise-free or relatively noisy have been correctly identified.

12. The method of claim 11, wherein said morphological processing step comprises:

erosion of said error code stream, by changing a sequence of contiguous error codes in said error code stream to identify as noisy those of said error codes previously identified as relatively noise-free but having a sequence length that is less than a first predetermined length M, and

dilation of said error code stream, by changing a sequence of contiguous error codes in said error code stream to identify as relatively noise-free those of said error codes previously identified as noisy but having a sequence length that is less than a second predetermined length L.

13. The method of claim 12, wherein said erosion step sets said length M to be greater than said length L.

14. Apparatus for reducing the level of an undesired noise component in an acquired physiological signal having a desired signal component, comprising:

a sensor arrangement for acquiring a physiological signal;

a signal generator for modeling a synthetic reference signal that exhibits a correlation with the desired signal component of the acquired physiological signal; and

an adaptive filter arrangement having a first input responsive to said acquired physiological signal, and a second input responsive to said synthetic reference signal, said adaptive filter arrangement being operated in a feedback manner so as to iteratively adjust the modeled synthetic reference signal and progressively generate as an output of the filter arrangement a more accurate approximation of the desired signal component.

15. The apparatus of claim 14, wherein said signal generator comprises:

a processor that is responsive to a predetermined kernel function representative of a pulsatile cycle of the desired signal component of the acquired physiological signal.

16. The apparatus of claim 14, wherein said adaptive filter comprises:

a multistage lattice prediction filter for shifting and weighting of said synthetic reference signal to generate backward prediction error signals; and

a regressor processing stage for iteratively combining said acquired physiological signal with shifted and weighted versions of said backward prediction error signals, to progressively build-up a reconstruction of the desired component of said acquired physiological signal.

17. The apparatus of claim 14, including:

a confidence processor for processing the acquired physiological signal so as to identify sections thereof having a high confidence of being relatively noise-free.

18. The apparatus of claim 17, further including:

an extraction processor responsive to the acquired physiological signal and to the confidence processor, so as to only process said high confidence sections of the acquired physiological signal, and in response to said processing develop a value that is used to adjust the generation of the synthetic reference signal so that it exhibits a greater correlation with the desired signal component of the acquired physiological signal.

19. The apparatus of claim 18, wherein:

said sensor arrangement acquires first and second ones of acquired physiological signals; and

said confidence processor comprises;

a correlation detector for subjecting the first and second ones of said acquired physiological signals to a correlation test; and

an error code tagger responsive to said correlation detector for identifying sections of each of said acquired physiological signals that exhibit a correlation that exceeds a predetermined threshold level as being relatively noise-free, and identifying sections of each of said acquired physiological signals that exhibit a correlation that does not exceed the predetermined threshold level as being relatively noisy.

20. The apparatus of claim 18, wherein said extraction processor comprises:

an FFT processor responsive to said confidence processor, for processing only those sections of the first and second ones of said acquired physiological signals that are identified to have a high confidence of being relatively noise-free, to develop a value representative of at least a fundamental frequency component of the desired signal component of the acquired physiological signal, which value is used to adjust the modeling of the synthetic reference signal so as to increase its correlation with the desired signal component of the acquired physiological signal.

21. The apparatus of claim 18, wherein said confidence processor comprises:

an amplitude detector for subjecting said acquired physiological signal to an amplitude threshold test; and

an error code tagger for identifying sections of said acquired physiological signal that have an amplitude that does not exceed a predetermined amplitude threshold level as being relatively noise-free, and identifying sections of said acquired physiological signal that have an amplitude that does exceed the predetermined amplitude threshold level as being relatively noisy.

22. The apparatus of claim 21, wherein:

said sensor arrangement acquires first and second ones of acquired physiological signals; and

said confidence processor includes;

a correlation detector for subjecting the first and second ones of said acquired physiological signals to a correlation test; and

said error code tagger is responsive to said correlation detector for developing an error code stream that identifies sections of each of said acquired physiological signals that exhibit a correlation that exceeds a predetermined threshold level as being relatively noise-free, and identifies sections of each of said acquired physiological signals that exhibit a correlation that does not exceed the predetermined threshold level as being relatively noisy.

23. The apparatus of claim 22, wherein said extraction processor comprises:

an FFT processor responsive to said error code stream and the first and second ones of said acquired physiological signals, for processing only those sections of the first and second ones of said acquired physiological signals that are identified by said error code stream to have a high confidence of being relatively noise-free, to develop a value representative of at least a fundamental frequency component of the desired signal component of the acquired physiological signals, which value is used to adjust the modeling of the synthetic reference signal so as to increase its correlation with the desired signal component of the acquired physiological signals.

24. The apparatus of claim 22, further including.

a morphological processor for processing the error code stream to increase the degree of confidence that sections of said acquired physiological signal identified as being relatively noise-free or relatively noisy have been correctly identified.

25. The apparatus of claim 24, wherein said morphological processor comprises:

an eroder for eroding said error code stream, by changing a sequence of contiguous error codes in said error code stream to identify as noisy those of said error codes previously identified as relatively noise-free but having a sequence length that is less than a first predetermined length M, and

a dilator for dilating said error code stream, by changing a sequence of contiguous error codes in said error code stream to identify as relatively noise-free those of said error codes previously identified as noisy but having a sequence length that is less than a second predetermined length L.