Method and apparatus for heart rate monitoring using an electrocardiogram sensor
A method and corresponding apparatus employ a time-varying spectral analysis approach for reconstructing an electrocardiogram (ECG) signal that includes motion artifacts. The motion artifacts are produced by motion of an ECG sensor relative to a sensing location. The time-varying spectral analysis based approach enables the ECG signal to be reconstructed with accuracy by suppressing the motion artifacts. Example applications for the method and corresponding apparatus include ECG-based heart rate monitoring in wearable devices for fitness tracking and health monitoring even during intense physical activities.
1. A method for reconstructing a heart-related signal output by a biomedical sensor, the method comprising:
pre-processing the heart-related signal to produce a pre-processed heart-related signal;
reconstructing a representation of the heart-related signal to produce a reconstructed representation of the heart-related signal, the reconstructing based on a time-varying spectral analysis of the pre-processed heart-related signal, the heart-related signal including motion artifacts, the motion artifacts being signal artifacts produced by movement of the biomedical sensor relative to a sensing location, the pre-processing reducing the motion artifacts in the pre-processed heart-related signal for the reconstructing, the heart-related signal being an electrocardiogram (ECG) signal; and
outputting the reconstructed representation of the heart-related signal.
2. The method of claim 1 , wherein the biomedical sensor is an electrocardiogram (ECG) sensor.
3. The method of claim 1 , wherein the reconstructing further reduces the motion artifacts reduced by the pre-processing.
4. The method of claim 1 , wherein the pre-processing includes down-sampling the heart-related signal to produce a down-sampled heart-related signal, the down-sampling being at a sampling rate less than an original sampling rate.
5. The method of claim 4 , wherein:
the time-varying spectral analysis is based on a frequency resolution; and
the down-sampling affects the frequency resolution of the time-varying spectral analysis.
6. The method of claim 4 , wherein the pre-processing includes computing a derivative of the down-sampled heart-related signal to reduce the motion artifacts.
7. The method of claim 6 , wherein the pre-processing further includes computing an absolute value of the derivative to further reduce the motion artifacts.
8. The method of claim 1 , wherein the time-varying spectral analysis includes computing a time-frequency spectrum (TFS) of the pre-processed heart-related signal.
9. The method of claim 8 , wherein the TFS computed is a 3-dimensional spectra including a time-varying amplitude or power distribution with respect to time and frequency.
10. The method of claim 8 , wherein the TFS computed is a time-varying power spectral density (PSD).
11. The method of claim 8 , further including limiting the TFS computed to a given frequency range.
12. The method of claim 11 , wherein the given frequency range is 0.5 Hz to 3 Hz.
13. The method of claim 8 , wherein:
the TFS is computed for each shift of a windowed data segment of the pre-processed heart-related signal; and
the heart related signal is reconstructed for each shift of the windowed data segment.
14. The method of claim 13 , wherein, at each shift of the windowed data segment subsequent to an initial windowed data segment, the method further includes:
retaining up to a pre-determined number of frequency spectra to produce a subset of frequency spectra; and
selecting a frequency component to produce a selected frequency component for the reconstructing, the selected frequency component selected from amongst frequency components of the subset of frequency spectra and a previous frequency component, the previous frequency component having been selected for a previous shift of the windowed data segment to reconstruct the heart-related signal for the previous shift.
15. The method of claim 1 , wherein the time-varying spectral analysis is a first time-varying spectral analysis and wherein the reconstructing is further based on a second time-varying spectral analysis of a motion signal, the motion signal output by a motion sensor and representative of the motion artifacts in the heart-related signal.
16. The method of claim 15 , further including employing the second time-varying spectral analysis of the motion signal to produce a movement classification of the movement, and wherein the reconstructing is further based on the classification of the movement.
17. The method of claim 15 , wherein the heart-related signal and the motion signal are output, concurrently.
18. The method of claim 15 , wherein the biomedical sensor and the motion sensor are co-located.
19. The method of claim 15 , wherein the motion sensor is an accelerometer.
20. The method of claim 1 , further comprising employing the reconstructed representation to determine a heart rate estimate.
21. The method of claim 1 , further comprising employing the reconstructed representation to determine a heart rate variability (HRV) estimate.
22. The method of claim 1 , further comprising employing the reconstructed representation to detect or predict a heart-related ailment, the heart-related ailment including at least one of a heart rate variability (HRV) condition, atrial fibrillation condition, congestive heart failure condition, and tachycardia condition.
23. The method of claim 1 , wherein the pre-processing, the reconstructing, and the outputting are performed in real-time with respect to outputting of the heart-related signal by the biomedical sensor.
24. The method of claim 1 , wherein the pre-processing, the reconstructing, and the outputting are performed in non-real-time with respect to outputting of the heart-related signal by the biomedical sensor.
25. An apparatus for reconstructing a heart-related signal output by a biomedical sensor, the apparatus comprising:
a pre-processing unit configured to pre-process the heart-related signal to produce a pre-processed heart-related signal;
a reconstruction unit configured to reconstruct a representation of the heart-related signal to produce a reconstructed representation of the heart-related signal, the reconstructing based on a time-varying spectral analysis of the pre-processed heart-related signal, the heart-related signal including motion artifacts, the motion artifacts being signal artifacts produced by movement of the biomedical sensor relative to a sensing location, the pre-processing unit reducing the motion artifacts in the pre-processed heart-related signal for the reconstructing, the heart-related signal being an electrocardiogram (ECG) signal; and
an output unit configured to output the reconstructed representation of the heart-related signal.
26. The apparatus of claim 25 , wherein the biomedical sensor is an electrocardiogram (ECG) sensor.
27. The apparatus of claim 25 , wherein the reconstruction unit further reduces the motion artifacts reduced by the pre-processing unit.
28. The apparatus of claim 25 , wherein pre-processing by the pre-processing unit includes down-sampling the heart-related signal to produce a down-sampled heart-related signal, the down-sampling being at a sampling rate less than an original sampling rate.
29. The apparatus of claim 28 wherein:
the time-varying spectral analysis is based on a frequency resolution; and
the down-sampling affects the frequency resolution of the time-varying spectral analysis.
30. The apparatus of claim 28 , wherein the pre-processing unit is further configured to compute a derivative of the down-sampled heart-related signal to reduce the motion artifacts.
31. The apparatus of claim 30 , wherein the pre-processing unit is further configured to compute an absolute value of the derivative to further reduce the motion artifacts.
32. The apparatus of claim 25 , wherein the reconstruction unit is further configured to perform the time-varying spectral analysis by computing a time-frequency spectrum (TFS) of the pre-processed heart-related signal.
33. The apparatus of claim 32 , wherein the TFS computed is a 3-dimensional spectra including a time-varying amplitude or power distribution with respect to time and frequency.
34. The apparatus of claim 32 , wherein the TFS computed is a time-varying power spectral density (PSD).
35. The apparatus of claim 32 , wherein the reconstruction unit is further configured to limit the TFS computed to a given frequency range.
36. The apparatus of claim 35 , wherein the given frequency range is 0.5 Hz to 3 Hz.
37. The apparatus of claim 32 , wherein:
the TFS is computed for each shift of a windowed data segment of the pre-processed heart-related signal; and
the reconstruction unit is further configured to reconstruct the heart related signal for each shift of the windowed data segment.
38. The apparatus of claim 37 , wherein, at each shift of the windowed data segment subsequent to an initial windowed data segment, the reconstruction unit reconstructs the representation by:
retaining up to a pre-determined number of frequency spectra to produce a subset of frequency spectra; and
selecting a frequency component to produce a selected frequency component for the reconstructing, the selected frequency component selected from amongst frequency components of the subset of frequency spectra and a previous frequency component, the previous frequency component having been selected for a previous shift of the windowed data segment to reconstruct the heart-related signal for the previous shift.
39. The apparatus of claim 25 , wherein the time-varying spectral analysis is a first time-varying spectral analysis and wherein the reconstruction unit is further configured to perform the reconstructing based on a second time-varying spectral analysis of a motion signal, the motion signal output by a motion sensor and representative of the motion artifacts in the heart-related signal.
40. The apparatus of claim 39 , wherein the reconstruction unit is further configured to employ the second time-varying spectral analysis of the motion signal to produce a movement classification of the movement, and wherein the reconstructing is further based on the classification of the movement.
41. The apparatus of claim 39 , wherein the heart-related signal and the motion signal are output by the biomedical sensor and the motion sensor, respectively, concurrently.
42. The apparatus of claim 39 , wherein the apparatus includes the biomedical sensor and the motion sensor.
43. The apparatus of claim 39 , wherein the motion sensor is an accelerometer.
44. The apparatus of claim 25 , further comprising a heart rate unit configured to employ the reconstructed representation to determine a heart rate estimate.
45. The apparatus of claim 25 , further comprising a heart rate variability (HRV) unit configured to employ the reconstructed representation to determine a heart rate variability (HRV) estimate.
46. The apparatus of claim 25 , further comprising an ailment unit configured to employ the reconstructed representation to detect or predict a heart-related ailment, the heart-related ailment including at least one of a heart rate variability (HRV) condition, atrial fibrillation condition, congestive heart failure condition, and tachycardia condition.
47. The apparatus of claim 25 , wherein the reconstruction unit, the output unit, and the pre-processing unit are configured to reconstruct, pre-process, and output, respectively, in real-time with respect to outputting of the heart-related signal by the biomedical sensor.
48. The apparatus of claim 25 , wherein the reconstruction unit, the output unit, and the pre-processing unit are configured to reconstruct, pre-process, and output, respectively, in non-real-time with respect to outputting of the heart-related signal by the biomedical sensor.
49. A non-transitory computer-readable medium having encoded thereon a sequence of instructions which, when loaded and executed by a processor, causes the processor to reconstruct a heart-related signal output by a biomedical sensor, the processor reconstructing the heart-related signal by:
pre-processing the heart-related signal to produce a pre-processed heart-related signal;
reconstructing a representation of the heart-related signal to produce a reconstructed representation of the heart-related signal, the reconstructing based on a time-varying spectral analysis of the pre-processed heart-related signal, the heart-related signal including motion artifacts, the motion artifacts being signal artifacts produced by movement of the biomedical sensor relative to a sensing location, the pre-processing reducing the motion artifacts in the pre-processed heart-related signal for the reconstructing, the heart-related signal being an electrocardiogram (ECG) signal; and
outputting the reconstructed representation of the heart-related signal.