Continuous cardiac pulse rate estimation from multi-channel source video data
View Patent ↗What is disclosed is a computationally efficient system and method for estimating a subject's cardiac pulse rate from multi-channel source video data. In one embodiment, A time-series signal is received. A sliding window is used to define overlapping segments of the time-series signal. Signal segments are processed by performing constrained independent component analysis (cICA) until convergence to obtain an estimated source signal. A frequency of each estimated source signal obtained by the cICA at convergence is determined to be the subject's estimated cardiac pulse rate for each signal segment. A seed reference signal used by the cICA is repeatedly updated. A sliding window is shifted to define a next time-series signal segment for processing. The method repeats for each signal segment until a termination criteria is met. In such a manner, the subject's cardiac pulse rate is estimated from a video of the subject on a continuous basis.
1. A method for continuous cardiac pulse rate estimation from video images captured of a subject of interest being monitored for cardiac function, the method comprising:
receiving a time-series signal of a subject of interest being monitored for cardiac function;
overlaying said received time-series signal with a sliding window, said sliding window defining segments of said received time-series signal for processing, each successive position of said sliding window defining a next segment of said time-series signal for processing;
defining a seed reference signal with a frequency range that approximates a frequency range of said subject's cardiac pulse; and
processing segments of said received time-series signal by repeatedly:
performing constrained source separation (cICA) on said signal segment using said seed reference signal to obtain an estimated source signal, said cICA converging on an occurrence of one of: an error between said estimated source signal and said seed reference signal being less than a threshold, and a pre-defined number of iterations having occurred;
A) in response to said error being less than said threshold:
determining a frequency of said estimated source signal as said subject's estimated cardiac pulse rate for this time-series signal segment;
using said estimated source signal as said seed reference signal for processing a next time-series signal segment on a next iteration; and
shifting said sliding window to define a next segment of said time-series signal, said next segment at least partially overlapping a previous time-series signal segment defined by said window on a last iteration; and
B) in response to said pre-defined number of iterations having occurred, performing one of:
updating said seed reference signal by changing any of: a frequency, an amplitude, a phase, and a waveform of said seed reference signal, and using said updated reference signal as said seed reference signal to re-process said time-series signal segment; and
selecting an estimated source signal obtained from having processed a previous time-series signal segment and using said selected estimated source signal as said seed reference signal to re-process said time-series signal segment.
2. The method of claim 1 , wherein said time-series signal is received on a continuous basis and processed as said time-series signal is being received.
3. The method of claim 1 , wherein said received time-series signal comprises any of: RGB signals, IR signals, RGB and IR signals, multi-spectral signals, and hyperspectral signals.
4. The method of claim 1 , wherein, in advance of using said estimated source signal as said seed reference signal, further comprising conditioning said estimated source signal to remove artifacts.
5. The method of claim 1 , wherein said waveform comprises any of: a sine wave, a square wave, a user-defined shape obtained from an ECG signal, and a cardiac pulse waveform derived from said subject.
6. The method of claim 1 , wherein said time-series signal comprises one of: stored values, and values generated from a streaming video.
7. The method of claim 1 , wherein at least two of said time-series signal segments are defined by windows with a different size.
8. The method of claim 1 , further comprising providing said subject's estimated cardiac pulse rate to a display device.
9. A system for continuous cardiac pulse rate estimation from video images captured of a subject of interest being monitored for cardiac function, the system comprising:
a memory; and
a processor in communication with said memory, said processor executing machine readable instructions for performing:
receiving a time-series signal of a subject of interest being monitored for cardiac function;
overlaying said received time-series signal with a sliding window, said sliding window defining segments of said received time-series signal for processing, each successive position of said sliding window defining a next segment of said time-series signal for processing;
defining a seed reference signal with a frequency range that approximates a frequency range of said subject's cardiac pulse; and
processing segments of said received time-series signal by repeatedly:
performing constrained source separation (cICA) on said signal segment using said seed reference signal to obtain an estimated source signal, said cICA converging on an occurrence of one of: an error between said estimated source signal and said seed reference signal being less than a threshold, and a pre-defined number of iterations having occurred;
A) in response to said error being less than said threshold:
determining a frequency of said estimated source signal as said subject's estimated cardiac pulse rate for this time-series signal segment;
using said estimated source signal as said seed reference signal for processing a next time-series signal segment on a next iteration; and
shifting said sliding window to define a next segment of said time-series signal, said next segment at least partially overlapping a previous time-series signal segment defined by said window on a last iteration; and
B) in response to said pre-defined number of iterations having occurred, performing one of:
updating said seed reference signal by changing any of: a frequency, an amplitude, a phase, and a waveform of said seed reference signal, and using said updated reference signal as said seed reference signal to re-process said time-series signal segment; and
selecting an estimated source signal obtained from having processed a previous time-series signal segment and using said selected estimated source signal as said seed reference signal to re-process said time-series signal segment.
10. The system of claim 9 , wherein said time-series signal is received on a continuous basis and processed as said time-series signal is being received.
11. The system of claim 9 , wherein said received time-series signal comprises any of: RGB signals, IR signals, RGB and IR signals, multi-spectral signals, and hyperspectral signals.
12. The system of claim 9 , wherein, in advance of using said estimated source signal as said seed reference signal, further comprising conditioning said estimated source signal to remove artifacts.
13. The system of claim 9 , wherein said waveform comprises any of: a sine wave, a square wave, a user-defined shape obtained from an ECG signal, and a cardiac pulse waveform derived from said subject.
14. The system of claim 9 , wherein at least two of said time-series signal segments are defined by windows with a different size.
15. The system of claim 9 , wherein said time-series signal comprises one of: stored values, and values generated from a streaming video.
16. The system of claim 9 , further comprising providing said subject's estimated cardiac pulse rate to a display device.
17. A non-transitory computer readable media containing machine readable program instructions for performing a method comprising: receiving a time-series signal of a subject of interest being monitored for cardiac function;
overlaying said received time-series signal with a sliding window, said sliding window defining segments of said received time-series signal for processing, each successive position of said sliding window defining a next segment of said time-series signal for processing;
defining a seed reference signal with a frequency range that approximates a frequency range of said subject's cardiac pulse; and
processing segments of said received time-series signal by repeatedly:
performing constrained source separation (cICA) on said signal segment using said seed reference signal to obtain an estimated source signal, said cICA converging on an occurrence of one of: an error between said estimated source signal and said seed reference signal being less than a threshold, and a pre-defined number of iterations having occurred;
A) in response to said error being less than said threshold:
determining a frequency of said estimated source signal as said subject's estimated cardiac pulse rate for this time-series signal segment;
using said estimated source signal as said seed reference signal for processing a next time-series signal segment on a next iteration; and
shifting said sliding window to define a next segment of said time-series signal, said next segment at least partially overlapping a previous time-series signal segment defined by said window on a last iteration; and
B) in response to said pre-defined number of iterations having occurred, performing one of:
updating said seed reference signal by changing any of: a frequency, an amplitude, a phase, and a waveform of said seed reference signal, and using said updated reference signal as said seed reference signal to re-process said time-series signal segment; and
selecting an estimated source signal obtained from having processed a previous time-series signal segment and using said selected estimated source signal as said seed reference signal to re-process said time-series signal segment.
18. The non-transitory computer readable media of claim 17 , wherein said time-series signal is received on a continuous basis and processed as said time-series signal is being received.
19. The non-transitory computer readable media of claim 17 , wherein said received time-series signal comprises any of: RGB signals, IR signals, RGB and IR signals, multi-spectral signals, and hyperspectral signals.
20. The non-transitory computer readable media of claim 17 , wherein, in advance of using said estimated source signal as said seed reference signal, further comprising conditioning said estimated source signal to remove artifacts.
21. The non-transitory computer readable media of claim 17 , wherein said waveform comprises any of: a sine wave, a square wave, a user-defined shape obtained from an ECG signal, and a cardiac pulse waveform derived from said subject.
22. The non-transitory computer readable media of claim 17 , wherein said time-series signal comprises one of: stored values, and values generated from a streaming video.
23. The non-transitory computer readable media of claim 17 , wherein at least two of said time-series signal segments are defined by windows with a different size.
24. The non-transitory computer readable media of claim 17 , further comprising providing said subject's estimated cardiac pulse rate to a display device.