Heart beat signal recognition

• US 8,046,058 B2
• Filed: 08/10/2007
• Issued: 10/25/2011
• Est. Priority Date: 08/10/2007
• Status: Expired due to Fees

• Current Assignee: Salutron, Inc.
• Original Assignee: Salutron, Inc.
• Inventors: Lin, Szming, Lo, Thomas Ying-Ching
• Primary Examiner(s): Layno; Carl H
• Assistant Examiner(s): Stice; Paula J
• Application Number:
  US11/837,166
• Publication Number:
  US 20090043216A1
• Time in Patent Office: 1,537 Days
• Field of Search: 600450-505, 600508-509
• US Class Current: 600/509
• CPC Class Codes: A61B 5/0205 Simultaneously evaluating both cardiovascular conditions and different types of body conditions, e.g. heart and respiratory condition
• CPC Class Codes: A61B 5/35 by template matching
• CPC Class Codes: A61B 5/486 Bio-feedback A61B5/375 takes precedence
• CPC Class Codes: A61B 5/7214 using signal cancellation, e.g. based on input of two identical physiological sensors spaced apart, or based on two signals derived from the same sensor, for different optical wavelengths
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G06F 2218/08 Feature extraction
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

A subject'"'"'s heart rate is determined. A heart rate monitor receives a Doppler signal reflected from an artery of a target, performs demodulation and heart beat recognition techniques to determine a set of features in each frame of the signal. Pattern classification is performed to determine if the extracted feature sequence is associated with heart beats. The pattern classification may include finding the optimal state sequence by calculating the probability of each allowable state sequence based on the extracted feature sequence and heart beat models or additional noise models. Or, a heart beat candidate is determined using frame energy and dynamic thresholding followed by computing the probabilities between the feature sequence and each stored heart beat model or additional noise models. Or, heart beat candidates are determined using frame energy and dynamic thresholding which compute the similarity between the feature sequences and each of the stored heart beat templates.

Claims

1. A computer implemented method to determine a heart rate for a subject, comprising:

detecting a heartbeat signal from the subject;

extracting a set of two or more types of heartbeat features from the heartbeat signal;

accessing, from stored data, one or more heartbeat patterns and one or more noise patterns, the one or more heartbeat patterns comprising at least one of;

(a) one or more heartbeat models and (b) one or more heartbeat templates, and the one or more noise patterns comprising at least one of;

(c) one or more noise models and (d) one or more noise templates;

determining whether the set of two or more types of heartbeat features more closely matches the one or more heartbeat patterns or the one or more noise patterns;

if the set of two or more types of heartbeat features more closely matches the one or more noise patterns, determining that the set of two or more types of heartbeat features is not to be used to update the heart rate for the subject;

if the set of two or more types of heartbeat features more closely matches the one or more heartbeat patterns, determining that the set of two or more types of heartbeat features represents a heartbeat candidate;

performing post-processing to determine that the heartbeat candidate represents a heartbeat; and

updating the heart rate for the subject based on timing of the heartbeat.

2. The computer implemented method of claim 1, wherein the heartbeat signal is detected within an ultrasound signal reflected from the subject.

3. The computer implemented method of claim 1, wherein said extracting includes:

dividing the heartbeat signal into one or more frames; and

extracting the set of two or more types of heartbeat features in each frame of the heartbeat signal.

4. The computer implemented method if claim 1, wherein:

the two or more types of heartbeat features comprise one or more of signal energy values, heartbeat duration values and delta features.

5. The computer implemented method of claim 4, further comprising:

calculating the delta features based on differences among heartbeat features of the heartbeat signal.

6. The computer implemented method of claim 1, wherein said extracting includes:

performing linear predictive coding analysis on the heartbeat signal to determine a set of linear predictive coding coefficients.

7. The computer implemented method of claim 6, wherein said extracting includes:

performing cepstral analysis using the set of linear predictive coding coefficients to generate a set of cepstral coefficients.

8. The computer implemented method of claim 7, wherein said extracting includes extracting Mel Frequency Cepstral Coefficients.

9. The computer implemented method of claim 7, wherein said extracting includes:

estimating one or more formants of the heartbeat signal based on the set of the linear predictive coding coefficients.

10. The computer implemented method of claim 9, wherein the two or more types of heartbeat features include cepstral coefficients, the method further comprising:

weighting the cepstral coefficients.

11. The computer implemented method of claim 7, further comprising:

weighting a subset of the set of two or more types of features.

12. The computer implemented method of claim 1, wherein:

the one or more heartbeat models and the one or more noise models are part of at least one grammar network which defines multiple states and allowable sequences of the states, each state has an associated probability of transitioning to other states of the multiple states; and

the determining whether the set of two or more types of heartbeat features more closely matches the one or more heart beat patterns or the one or more noise patterns comprises determining a most probable sequence of the states, among the allowable sequences of the states, which is associated with the set of two or more types of heartbeat features.

13. The computer implemented method of claim 12, wherein:

the determining the most probable sequence of the states uses Hidden Markov Model techniques.

14. The computer implemented method of claim 1, wherein each of the one or more noise models represents noise generated from at least one of muscle movement and body movement.

15. The computer implemented method of claim 1, wherein:

the determining whether the set of two or more types of heartbeat features more closely matches the one or more heartbeat patterns or the one or more noise patterns comprises determining whether a probability of a match is higher between;

(a) the one or more heartbeat models and the set of two or more types of heartbeat features or (b) the one or more noise models and the set of two or more types of heartbeat features.

16. The computer implemented method of claim 1, wherein:

the determining whether the set of two or more types of heartbeat features more closely matches the one or more heartbeat patterns or the one or more noise patterns comprises determining whether a similarity of a match is higher between;

(a) the one or more heartbeat templates and the set of two or more types of heartbeat features or (b) the one or more noise templates and the set of two or more types of heartbeat features.

17. The computer implemented method of claim 1, wherein the one or more heartbeat models has at least first and second symbols, and each symbol characterizes a basic heartbeat signal pattern.

18. The computer implemented method of claim 17, wherein the first symbol is derived from a first heartbeat pattern associated with a heart contraction and the second symbol is derived from a second heartbeat pattern associated with vessel backflow.

19. The computer implemented method of claim 1, wherein the one or more heartbeat models are trained using a sampling of subjects having different ages and genders.

20. The computer implemented method of claim 1, wherein the one or more heartbeat models are trained using a sampling of subjects having different heartbeat back flow characteristics.

21. The computer implemented method of claim 1, wherein the one or more heartbeat models are trained using a sampling of subjects including a subject having an irregular heartbeat.

22. The computer implemented method if claim 1, wherein:

the two or more types of heartbeat features includes at least one of linear predictive coding coefficients, cepstrum coefficients, mel-frequency cepstrum coefficients, perceptual linear prediction coefficients, formants, signal energy, signal duration and cepstrum coefficient differences.

23. The computer implemented method if claim 1, wherein:

the one or more noise patterns comprise multiple noise patterns.

24. The computer implemented method if claim 1, further comprising:

determining a probability that the set of two or more types of heartbeat features is most likely characterized by more closely matches the one or more heartbeat models; and

determining whether the probability exceeds a threshold, the determining that identifying the set of two or more types of heartbeat features represents as representing the heartbeat candidate requires the probability to exceed the threshold.

25. The computer implemented method if claim 1, wherein:

the post-processing determines that the heartbeat candidate represents the heartbeat by determining that a time of the heartbeat candidate is within a time range threshold relative to one or more previous heartbeats.

26. The computer implemented method of claim 1, wherein:

the updating the heart rate for the subject includes transmitting the heart rate to a display device.

27. One or more processor readable storage devices having processor readable code embodied on said processor readable storage devices, said processor readable code for programming one or more processors to perform a method to determine a heart rate for a subject, the method performed comprising:

extracting a set of features from each frame of a plurality of frames of a heartbeat signal of the subject;

performing pattern classification of the set of features using one or more heartbeat patterns and one or more noise patterns, the one or more heartbeat patterns comprising at least one of;

(a) one or more heartbeat models and (b) one or more heartbeat templates, and the one or more noise patterns comprising at least one of;

(c) one or more noise models and (d) one or more noise templates;

determining whether the set of features more closely matches the one or more heartbeat patterns or the one or more noise patterns, based on the performing pattern classification;

if the set of features more closely matches the one or more noise patterns, determining that the set of features is not to be used to update the heart rate for the subject;

if the set of features more closely matches the one or more heartbeat patterns, determining that the set of features represents a heartbeat candidate;

performing post-processing to determine whether the heartbeat candidate represents a heartbeat; and

updating the heart rate for the subject based on timing of the heartbeat, if the post-processing determines that the heartbeat candidate represents the heartbeat.

28. The one or more processor readable storage devices of claim 27, wherein said extracting includes:

receiving a signal reflected from an artery of the subject;

dividing the reflected signal into frames; and

extracting the set of features for a portion of the reflected signal within each frame.

29. The one or more processor readable storage devices of claim 27, wherein the set of features includes at least one of linear predictive coding coefficients, cepstrum coefficients, mel-frequency cepstrum coefficients, perceptual linear prediction coefficients, formants, signal energy, signal duration and delta features.

30. The one or more processor readable storage devices of claim 27, wherein said performing pattern classification includes:

accessing one or more heartbeat models which characterize the one or more heartbeat patterns;

accessing one or more noise models which characterize the one or more noise patterns;

accessing at least one grammar network which defines multiple states and allowable sequences of the states, each state has an associated probability of transitioning to other states of the multiple states; and

using a Viterbi search algorithm to find a most likely sequence of the states through the at least one grammar network based on the set of features, wherein;

the set of features is determined to represent the heartbeat candidate if the most likely sequence of the states is associated with the one or more heartbeat models; and

the set of features is determined to not be used in updating the heart rate for the subject represent the noise if the most likely sequence of the states is associated with the one or more noise models.

31. The one or more processor readable storage devices of claim 27, wherein said performing pattern classification includes:

accessing one or more heartbeat models which characterize the one or more heartbeat patterns;

accessing one or more noise models which characterize the one or more noise patterns;

accessing at least one grammar network which defines multiple states and sequences of the states, each state has an associated probability of transitioning to other states of the multiple states;

processing each value in an observation sequence against each state in each heartbeat model and noise model; and

generating a total probability of match for each model by multiplying the probability of each value together for each model.

32. The one or more processor readable storage devices of claim 27, wherein said performing pattern classification includes:

accessing one or more heartbeat templates which characterize the one or more heartbeat patterns; and

generating a similarity measure of the set of features against each of the heartbeat templates using a dynamic time warping algorithm.

33. The one or more processor readable storage devices of claim 27, wherein said determining whether the set of features more closely matches the one or more heartbeat patterns or the one or more noise patterns includes:

determining the subject'"'"'s heart rate based on a matched model, selected probability or selected similarity.

34. The one or more processor readable storage devices of claim 27, wherein said step of performing pattern classification includes:

determining if a probability of a match is higher between;

(a) the one or more heartbeat patterns and the set of features, and or (b) the one or more noise patterns and the set of features; and

determining if the probability of a match exceeds a threshold value; and

the post-processing determines determining that the heartbeat candidate does not represent the heartbeat set of features represents the noise if the probability of a match does not exceed the threshold value, even if the probability of a match is higher between the one or more heartbeat patterns and the set of features than between the one or more noise patterns and the set of features.

35. The one or more processor readable storage devices of claim 27, wherein:

the one or more noise patterns comprise multiple noise patterns.

36. A heart rate monitor, comprising:

at least one transducer which detects a heartbeat signal from a subject;

at least one microcontroller associated with the at least one transducer;

at least one data store associated with the microcontroller, the data store comprising one or more heartbeat patterns and one or more noise patterns, the one or more heartbeat patterns comprising at least one of;

(a) one or more heartbeat models and (b) one or more heartbeat templates, and the one or more noise patterns comprising at least one of;

(c) one or more noise models and (d) one or more noise templates; and

an amplifier and a demodulator arranged between the at least one transducer and the at least one microcontroller;

the at least one microcontroller;

(i) extracts a set of two or more types of heartbeat features from the heartbeat signal, (ii) accesses, from the at least one data store, the one or more heartbeat patterns and one or more noise patterns, (iii) determines whether the set of two or more types of heartbeat features more closely matches the one or more heartbeat patterns or the one or more noise patterns, (iv) if the set of two or more types of heartbeat features more closely matches the one or more noise patterns, determines that the set of two or more types of heartbeat features is not to be used to update a heart rate for a subject, (v) if the set of two or more types of heartbeat features more closely matches the one or more heartbeat patterns, determines that the set of two or more types of heartbeat features represents a heartbeat candidate, and (iv) updates the heart rate for the subject based on timing of the heartbeat candidate.

37. The heart rate monitor of claim 36, further comprising:

a local display associated with the at least one microcontroller which displays the heart rate.

38. The heart rate monitor of claim 36, further comprising:

a transmitter associated with the at least one microcontroller, the transmitter transmits the heart rate in a wireless signal to a receiver of a remote display.

39. The heart rate monitor of claim 36, wherein:

the at least one transducer detects the heartbeat signal within an ultrasound signal reflected from the subject.

40. A heart rate monitor, comprising:

at least one transducer which detects a heartbeat signal from a subject;

at least one microcontroller associated with the at least one transducer; and

at least one data store associated with the microcontroller, the data store comprising one or more heartbeat patterns and one or more noise patterns, the one or more heartbeat patterns comprising at least one of;

(a) one or more heartbeat models and (b) one or more heartbeat templates, and the one or more noise patterns comprising at least one of;

(c) one or more noise models and (d) one or more noise templates;

the at least one microcontroller;

(i) extracts a set of heartbeat features from the heartbeat signal, (ii) accesses, from the at least one data store, the one or more heartbeat patterns and one or more noise patterns, (iii) determines whether the set of features more closely matches the one or more heartbeat patterns or the one or more noise patterns, (iv) if the set of features more closely matches the one or more noise patterns, determines that the set of features is not to be used to update a heart rate for a subject, (v) if the set of features more closely matches the one or more heartbeat patterns, determines that the set of features represents a heartbeat candidate, and (vi) updates the heart rate for the subject based on timing of the heartbeat candidate.

41. The heart rate monitor of claim 40, wherein:

the at least one transducer detects the heartbeat signal within an ultrasound signal reflected from the subject.