Method and system for pulse waveform recognition

Abstract

A pulse waveform identification method includes: continuously applying pulse pressure to the wrist using a pressure application mechanism, and continuously measuring the hand pulse of the wrist using a pressure sensor to generate pulse pressure data; and performing the following operations using a processor: determining whether the pulse pressure data of the corresponding time window is a valid waveform according to a time sequence using a pulse valid waveform identification model, and if so, determining that the pulse pressure data has entered the valid measurement range; continuously measuring with a pressure sensor to generate pulse pressure data and determining whether the pulse pressure data of the corresponding time window is an invalid waveform according to a time sequence using the pulse valid waveform identification model; and when it is determined that the pulse pressure data of the corresponding time window is an invalid waveform, determining that the pulse pressure data has entered the invalid measurement range, and stopping the application of pulse pressure by the pressure application mechanism.

Claims

1. A pulse waveform identification method, comprising: continuously applying a pulse pressure to a wrist using a pressure applying mechanism, and continuously measuring a hand pulse of the wrist using a pressure sensor to generate pulse pressure data; and performing the following operations using a processor: determining, according to a time sequence, whether the pulse pressure data corresponding to a time window is a valid waveform using a pulse valid waveform identification model, and if so, determining that the pulse pressure data has entered a valid measurement interval; continuously measuring with the pressure sensor to generate the pulse pressure data and determining, according to the time sequence, whether the pulse pressure data corresponding to the time window is an invalid waveform using the pulse valid waveform identification model; and when it is determined that the pulse pressure data corresponding to the time window is the invalid waveform, determining that the pulse pressure data has entered an invalid measurement interval, and stopping the pressure applying the pulse pressure by the pressure applying mechanism.

2. The pulse waveform recognition method as described in claim 1, wherein the pressure-applying mechanism continuously applies the pulse-taking pressure to the wrist and the pressure sensor continuously measures the hand pulse of the wrist to generate pulse pressure training data, and the pulse waveform recognition method further includes: establishing a training dataset using the processor to train the effective pulse waveform recognition model, wherein establishing the training dataset includes the processor performing the following operations: receiving the pulse pressure training data corresponding to a complete measurement interval; dividing the pulse pressure training data corresponding to the complete measurement interval into an effective time-domain training interval and an invalid time-domain training interval using an effective range algorithm; and respectively dividing the effective time-domain training interval and the invalid time-domain training interval into a plurality of effective time-domain training data and a plurality of invalid time-domain training data according to time sequence using a training time window to include them in the training dataset.

3. The pulse waveform identification method as described in claim 2, wherein one pulse pressure waveform of the pulse pressure data includes a plurality of sequentially adjacent preceding and following peaks and a trough between the preceding and following peaks, and the establishment of the training dataset further includes the processor performing the following operations: using the effective range algorithm to divide the portion of the pulse pressure data in the pulse pressure waveform that is closer to the preceding peak into the invalid time-domain training interval, and dividing the portion of the pulse pressure data in the pulse pressure waveform that is closer to the following peak into the effective time-domain training interval.

4. The pulse waveform identification method as described in claim 2, wherein establishing the training dataset further comprises the processor performing the following operations: converting the plurality of valid time-domain training data and the plurality of invalid time-domain training data into a plurality of valid frequency-domain training data and a plurality of invalid frequency-domain training data by a time-frequency conversion for inclusion in the training dataset.

5. The pulse waveform identification method as described in claim 2, wherein the pressure applying mechanism applies the pulse pressure to generate a first pulse pressure training data corresponding to the pulse pressure training data, and establishing the training data set further includes the processor performing the following operations: receiving the first pulse pressure training data corresponding to the complete measurement interval; and dividing the first pulse pressure training data into a plurality of second pulse pressure training data in a time sequence according to the training time window to be included in the training data set.

6. The pulse waveform identification method as described in claim 1, wherein the time window segmented from the pulse pressure data is spaced one step apart, and the time window and the step size are determined based on a rapid pulse rate within a pulse rate interval detected by the pressure-applying mechanism for the hand pulse.

7. A pulse waveform recognition system, comprising: a pulse diagnostic instrument, including: a pressure application mechanism for continuously applying a pulse diagnosis pressure to a wrist; and a pressure sensor for continuously measuring a hand pulse of the wrist to generate pulse pressure data; and a processor connected to the pulse diagnostic instrument, the processor being configured to perform the following operations: determining, according to a pulse valid waveform recognition model, whether the pulse pressure data corresponding to a time window is a valid waveform, and if so, determining that the pulse pressure data has entered a valid measurement interval; continuously measuring with the pressure sensor to generate the pulse pressure data and determining, according to the pulse valid waveform recognition model, whether the pulse pressure data corresponding to the time window is an invalid waveform; and when determining that the pulse pressure data corresponding to the time window is the invalid waveform, determining that the pulse pressure data has entered an invalid measurement interval, and stopping the pressure application mechanism from applying the pulse diagnosis pressure.

8. The pulse waveform recognition system as described in claim 7, wherein the pressure-applying mechanism continuously applies the pulse pressure to the wrist and the pressure sensor continuously measures the hand pulse of the wrist to generate pulse pressure training data, and the processor is further configured to establish a training dataset to train the effective pulse waveform recognition model, wherein establishing the training dataset includes the processor performing the following operations: receiving the pulse pressure training data corresponding to a complete measurement interval; dividing the pulse pressure training data corresponding to the complete measurement interval into an effective time-domain training interval and an invalid time-domain training interval using an effective range algorithm; and respectively dividing the effective time-domain training interval and the invalid time-domain training interval into a plurality of effective time-domain training data and a plurality of invalid time-domain training data according to time sequence using a training time window to include them in the training dataset.

9. The pulse waveform recognition system as described in claim 8, wherein establishing the training dataset further comprises the processor performing the following operations: converting the plurality of valid time-domain training data and the plurality of invalid time-domain training data into a plurality of valid frequency-domain training data and a plurality of invalid frequency-domain training data by a time-frequency conversion for inclusion in the training dataset.

10. The pulse waveform recognition system as claimed in claim 8, wherein the pressure applying mechanism applies the pulse pressure to generate a first pulse pressure training data corresponding to the pulse pressure training data, and establishing the training data set further comprises the processor performing the following operations: receiving the first pulse pressure training data corresponding to the complete measurement interval; and dividing the first pulse pressure training data into a plurality of second pulse pressure training data in a time sequence according to the training time window to be included in the training data set.

Images