TWI635852B - Signal processing method for plural sensors detecting pulse in a palm - Google Patents

Abstract

A signal processing method for detecting a pulse of a palm of a multi-sensor includes: setting a complex sensor above the back shell of the mouse; using these sensors to sense pulse changes in different parts of the palm to generate a sense The test signal is implemented by a microprocessor executing an algorithm to process the sensing signals into a mixed waveform, and calculating a heart rate value of the user based on the mixed waveform. The above algorithm performs peak detection for each sensing signal. If the sensing signal is detected as a peak, increase its weight; otherwise, lower its weight. Then, based on the adjusted weights, weighted average calculations are performed on the sensing signals to obtain a mixed waveform. The method of the invention can effectively improve the probability of detecting the heart rate, and can effectively stabilize the quality of the signal and reduce the error rate of the heart rate value calculation.

Description

Signal processing method applied to multi-sensor detecting palm pulse

The invention relates to a method for processing a signal for detecting a pulse, in particular to a method for processing a light volume change description signal for detecting a pulse of a palm with a multi-sensor.

Traditionally, the medical community's method of assessing cardiac function is to measure the user's heart rate or cardiac activity status by electrocardiogram (ECG). Heart Rate Variability (HRV) is also used to estimate changes in the user's psychological stress. However, with the aging of the global population, health care medical equipment that is home to the future needs a simpler operation mode, and the size of the instrument needs to be lighter and more convenient. Therefore, the photoplethysmography signal (hereinafter referred to as PPG signal) has been applied to wearable devices such as wristbands and watches to replace the traditional electrocardiography measurement method. The PPG signal provides information on arteries and blood flow. It is non-invasive, easy to operate, and has no consumables. It can replace the shortage of ECG and is suitable for future telemedicine monitoring and home care to achieve early diagnosis of cardiovascular disease. prevention.

The process of using the PPG signal to obtain the heart rate value can be roughly divided into three stages: 1. Measurement of the PPG signal; 2. Optimization of the PPG signal (removing the noise, obtaining the basis for calculating the heart rate value); 3. Using the heart rate calculation The method calculates the heart rate value based on the optimized PPG signal (takes peaks and/or waves) Valley).

In the measurement of PPG signals, most wearable devices use a single sensor that contains a light emitting diode (LED) or a photodiode, such as a single sensor. The pulse wave and vasoconstriction of the finger are measured on the left side of the wired mouse. When measuring, the subject must adjust the posture of the hand so that the finger touches the sensor position for measurement. In addition, there are also literatures that use multi-sensors to measure PPG signals, such as: measuring PPG signals with four different wavelengths of four-channel sensors; or placing two array sensors on the bracelet. Above, used to measure the PPG signal at different positions of the wrist.

The PPG signal obtained by the sensor will generally be processed by the noise to optimize the signal, which can be used as the basis for the subsequent calculation of the heart rate value. For example, the integrated design of the three-axis accelerometer to reduce the impact of moving noise; or the use of particle filter to eliminate mobile noise. In addition, some documents use a specific algorithm to select several channels with better signal quality in the multi-channel signal as the basis for subsequent calculation.

In order to calculate the heart rate value in accordance with the PPG signals obtained by different detection and optimization methods, different literatures also propose their suitable heart rate algorithms. For example, an RPD (robust peak detection) algorithm is used to capture the peaks and troughs of the PPG signal to calculate the heart rate value.

In summary, the quality of the single channel PPG signal is relatively unstable and the measurement location is limited. For multi-channel PPG signals, the conventional technique selects one or several channels from a plurality of different channel signals as a basis for subsequent calculation of heart rate values. However, the signal quality of the same channel at different points in time may vary greatly, which will cause difficulties in selection. In view of this, there is still room for improvement in the processing method of multi-channel PPG signals.

An object of the present invention is to provide a signal processing method for detecting a pulse of a palm of a multi-sensor, which can effectively improve the probability of detecting a heart rate, and can effectively stabilize the quality of the signal and reduce the error rate of the heart rate calculation.

In order to achieve the above object, the present invention provides a signal processing method for detecting a pulse of a palm of a multi-sensor, comprising: setting a plurality of sensors above a back shell of a mouse for a user to cover the palms with a palm a sensor; using these sensors to sense pulse changes in different parts of the palm to generate a sensing signal; using a microprocessor to perform an algorithm to process the sensing signals into a mixed waveform, according to the mixed waveform Calculating one of the user's heart rate values; and the microprocessor outputs the mixed waveform and heart rate values to a display interface. The above algorithm includes: setting a weight for each sensing signal and performing a peak detecting process; if the sensing signal is detected to a peak, increasing its weight; if the sensing signal is not detected In the case of a peak, the weight is lowered; and the weighted average calculation of the sensing signals is performed according to the adjusted weight of each sensing signal to obtain a mixed waveform.

In an embodiment, the peak detection process includes a dynamic threshold setting method for determining whether the time difference between one of the sampling points of the sensing signal and the previous peak is reasonable, and the steps include: according to the first two peaks of the sampling point The time difference determines an upper time limit and a lower time limit; and if the time difference between the sampling point and the previous peak falls between the upper time limit and the lower time limit, the microprocessor calculates a slope value of the sampling point.

The foregoing dynamic threshold setting method further includes: providing a slope threshold and a flag; comparing the slope value with a slope threshold; if the slope value is greater than the slope threshold, setting the flag to true; and when the flag is true, Determine whether the slope value is negative or equal to zero; if the slope value is negative or equal to zero when the flag is true, record a peak position and multiply a maximum slope value in the buffer memory by one. The slope is judged constant to calculate a new slope threshold.

In an embodiment, the adjusted weight of each sensed signal has a minimum value of 1, and its maximum value is 20.

The sensing signals of different PPG channels are first mixed and processed by the above method to form a waveform for calculating the heart rate value, and the obtained mixed waveform has higher stability than the sensing signals of the individual PPG channels, and is dynamically set. The slope threshold is used for peak detection to improve the accuracy of the heart rate value.

10‧‧‧ Mouse

12‧‧‧The back shell of the mouse

100‧‧‧Signal Processing System

110a, 110b, 110c and 110d‧‧‧PPG sensors

120‧‧‧ analog circuit

130‧‧‧Microprocessor

140‧‧‧Wireless Transmission Module

S _A1 , S _A2 , S _A3 , S _A4 ‧ ‧ (analog) sensing signals

122a, 122b, 122c, 122d‧‧‧ with follower

124a, 124b, 124c, 124d‧‧‧ bandpass filter

126‧‧‧Programmable Gain Amplifier

130‧‧‧Microprocessor

140,240‧‧‧Wireless transmission module

S1, S2, S3, S4‧‧‧ (digital) sensing signals

W _f ‧‧‧ mixed waveform

H _R ‧‧‧ heart rate

200‧‧‧Display interface

S310~S350‧‧‧ Signal processing steps

S332~S336‧‧‧ algorithm steps

Steps of the S3321~S3330‧‧‧ wave detection process

X _n ‧‧ ‧ sampling point

S _n-1 ‧‧‧ Maximum slope value detected in a waveform before the sampling point

P _n-1 , P _n-2 ‧‧‧Crest

PPI _n-1 ‧‧‧ time difference between two peaks

Time difference between the PPI _n ‧‧ ‧ sampling point and its previous peak

FIG. 1A is a schematic structural diagram of a signal processing system according to an embodiment of the present invention.

FIG. 1B is a schematic view showing a multi-sensor disposed on a back shell of a mouse.

2 is a schematic diagram of a signal processing flow according to an embodiment of the present invention.

3 is a schematic diagram of a peak detection process according to an embodiment of the present invention.

FIG. 4 is a schematic diagram of a dynamic threshold setting method according to an embodiment of the present invention.

Figure 5 is a diagram showing the weight assignment of each sampling point in an embodiment of the present invention.

6 is a comparison result of various sensing signal waveforms when a mouse is manually moved horizontally by an embodiment of the present invention.

FIG. 7 is a comparison result of various sensing signal waveforms when a subject holds a mouse browsing and clicking action according to an embodiment of the present invention.

The above and other technical contents, features and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments. As mentioned in the following examples Directions to the direction, such as: up, down, left, right, front or back, etc., are only used to refer to the direction of the accompanying drawings. Therefore, the directional terms are used for illustration only and are not intended to limit the invention.

FIG. 1A is a schematic diagram showing the architecture of a signal processing system according to an embodiment of the present invention, which is suitable for performing the signal processing method of the present invention for detecting a pulse of a palm of a multi-sensor. The signal processing system 100 is disposed on a mouse 10 and includes a plurality of PPG sensors 110a, 110b, 110c and 110d, an analog circuit 120, a microprocessor 130 and a wireless transmission module 140.

As shown in FIG. 1B, the PPG sensors 110a, 110b, 110c, and 110d are disposed at different positions on the upper surface of the back shell 12 of the mouse 10. When a user covers the back shell 12 of the mouse 10 with his palm, the PPG sensors 110a, 110b, 110c, and 110d can detect a change in the pulse of the user's palm blood vessels, and each generates an analog sensing signal S. _A1 , S _A2 , S _A3 , S _A4 . The analog circuit 120 is disposed inside the mouse 10 and includes a plurality of followers 122a, 122b, 122c and 122d, a plurality of band pass filters 124a, 124b, 124c and 124d and a Programmable Gain Amplifier (PGA). 126. The analog sensing signal S _A1 (or S _A2 , S _A3 , S _A4 ) of each PPG sensor 110a (or 110b, 110c, 110d) passes through the follower 122a (or 122b, 122c, 122d) and band pass filtering The processor 124a (or 124b, 124c, 124d) is processed and input to the programmable gain amplifier 126. The programmable gain amplifier 126 amplifies the gains of all analog sensing signals S _A1 , S _A2 , S _A3 , S _A4 from the band pass filters 124a , 124b , 124c and 124d in a multiplexed manner, and then through a analogy to The digital converter 128 converts the gain-amplified analog sensing signals S _A1 , S _A2 , S _A3 , S _A4 into digital sensing signals S1 , S2 , S3 , S4 , and then transmits them to the microprocessor 130 to perform the operation as shown in FIG. 2 . The signal processing flow is shown, thereby performing waveform mixing processing.

The microprocessor 130 and the wireless transmission module 140 are also disposed inside the mouse 10. The microprocessor 130 performs waveform mixing processing on the digital sensing signals S1, S2, S3, and S4 to obtain a mixed waveform W _f , and calculates the heart rate value H _R based on the mixed waveform W _f . The mixed waveform W _f and the heart rate value H _{R are} transmitted to a display interface 200 by the wireless transmission modules 140 and 240 to display an associated picture. In practice, a serial Peripheral Interface (SPI) can be connected between the microprocessor 130 and the programmable gain amplifier 126; the microprocessor 130 and the plurality of PPG sensors 110a, 110b, 110c An integrated circuit (I ² C) can be connected between 110d and 110d.

2 is a schematic diagram of a signal processing flow in the microprocessor 130. The microprocessor 130 receives the digital sensing signals S1, S2, S3, and S4 processed from the analog-to-digital converter 128 (step S310), and performs digital filtering on the digital sensing signals S1, S2, S3, and S4 (steps). S320). Next, the microprocessor 130 executes an algorithm (step S330) to process the digitally filtered digital sensing signals S1, S2, S3, S4 into a stable and good quality mixed waveform _Wf . A heart rate value H _{R is} calculated based on the mixed waveform W _f (step S340), and the mixed waveform W _f and the heart rate value H _R are displayed on the display interface 200 (step S350).

Figures 3 through 5 illustrate the algorithm of the present invention in more detail (step S330). The digital sensing signals S1, S2, S3, and S4 collected by the four sensors 110a, 110b, 110c, and 110d are assigned different weights according to the signal quality, and the best quality signals will receive the maximum weight. The step of the weighted average algorithm (step S330) includes: performing a peak detection process on the digital sensing signals S1, S2, S3, and S4 in the same time interval (step S332); if the digital sensing signal S1, If S2, S3 or S4 is detected as a peak, the weight of the digital sensing signal is increased; if the digital sensing signal S1, S2, S3 or S4 is not detected as a peak, then the digital sensing is lowered. The weight of the signal (step S334); and the weighted average calculation of the digital sensing signals S1, S2, S3 or S4 according to the adjusted weights of each of the digital sensing signals S1, S2, S3 and S4 to obtain a mixed waveform W _f (step S336). This algorithm can amplify the best and second-best signal weight of the four channels of sensing signals, and reduce the signal weight of the signal with poor signal quality. The resulting mixed waveform is smaller than the signal waveform of a single sensor and four sensing signals. When the mixed waveforms with equal weights are stable, the mixed waveform obtained by the method of the present invention calculates the heart rate value, and a more stable and accurate calculation result can be obtained.

Figure 3 illustrates the peak detection process of one embodiment of the present invention in more detail (step S332). The microprocessor 130 uses this peak detection process to determine the quality of each of the digital sensing signals S1, S2, S3, S4. First, a target waveform to be measured is subjected to differential processing to obtain a slope value (step S3321). Peak detection is performed using this slope value and a dynamic thresholding method. When the peak detection starts, a flag is provided first, and the default value is false (False). It is first judged whether or not the preset value of this flag is true (step S3322). If the preset value of the flag is not true, the slope value is compared with the magnitude of a slope threshold (step S3323). If the slope value is greater than the slope threshold, the flag is updated to true (step S3324), and the next data is continuously determined (step S3330). If it is determined in step S3322 that the flag is true, it is judged whether the slope value is negative or equal to zero (step S3326). If the slope value is greater than zero, the slope value is stored in a buffer (step S3325). If the flag is set to true (True) and the slope value is negative or equal to zero, it means that the slope value is changed to zero or negative by the state greater than the slope threshold, indicating that a peak is detected, and the flag is changed to false (False). And the peak position is recorded (step S3327). Next, a new slope threshold is calculated based on a maximum slope value stored in the buffer memory (step S3328). Finally, the buffer is cleared (step S3329) and the peak detection is resumed, and the next data is continuously judged (step S3330).

In the peak detection process shown in FIG. 3, when the slope value is changed to zero or negative by a state larger than the slope threshold, it means that a sampling point in the target waveform may detect a new peak, if the sampling point This new peak is confirmed when the time difference from the previous peak is within a reasonable time range. In order to achieve the peak detection of the embodiment, it is necessary to first determine a time threshold and set a slope threshold. To this end, the present invention provides a method of dynamically setting this time threshold and slope threshold.

FIG. 4 is a schematic diagram of a dynamic setting method of a time threshold and a slope threshold. The horizontal axis of the diagram is time, and the 緃 axis is the amplitude of the waveform (unit: Volt). The time threshold includes an upper time limit and a lower time limit, and the determining method is expressed by the following formulas (1) and (2): PPI _L = PPI _{n -1} × ( C _L ) (1)

PPI _H = PPI _{n -1} ×( C _H ) (2)

In equations (1) and (2), PPI _n-1 is the time difference between the first two peaks P _n-1 and P _n-2 of the sampling point X _n , and C _L (%) and C _H (%) are respectively A constant ( C _H > C _L ) for adjusting the upper time limit PPI _H and a lower time limit PPI _L .

The slope threshold is obtained by multiplying the maximum slope value stored in the buffer memory by a slope determination constant, and the determination method is expressed by the following formula (3): T _n = S _{n -1} × ( C _T ) (3)

In equation (3), T _n is the nth slope threshold, S _n-1 is the maximum slope value detected in the previous waveform (ie, n-1), and C _T (%) is the slope judgment constant.

Thus, the microprocessor 130 in accordance with a sampling point in the first two peaks X _n P _n-1 and P _n-2 time difference between PPI _n-1, to determine the time limit and time limit PPI _H PPI _L; if the sampling point X _n When the time difference PPI _n from the previous peak P _n-1 falls between the upper time limit PPI _H and the lower time limit PPI _L , the peak detecting process of FIG. 3 is performed to calculate a slope value of the sampling point X _n .

In step S334, the microprocessor 130 adjusts the weight of each of the digital sensing signals S1, S2, S3, and S4 according to the signal quality determined by the previous peak detection. For the signal with detected peak, increase its weight W[n]; for the signal without detecting the peak, reduce its weight W[n]. The specific weight adjustment method is expressed by the following formula (4), where α is an arbitrary constant:

Fig. 5 shows the distribution of the weight W[n] of each sampling point n . In this embodiment, the sensing signals S1, S2, S3, and S4 of each sensor channel have an initial weight of 1, and α is set to 2. At the first sampling point (n=1), the first sensing signal S1 is detected to have a peak, so the weight is increased to 3. When the second sampling point (n=2) detects that the first sensing signal S1 has a second peak, the weight of the signal S1 is further increased to 5; and the second sensing signal S2 is detected to have a peak, so The weight of the second sensing signal S2 is increased to 3. At the third sampling point (n=3), the first sensing signal S1 has a third peak, so the weight of the first sensing signal S1 is increased to 7; but the second sensing signal S2 is not detected. The peak is detected, so the weight of the second sensing signal S2 is reduced to one. At each sampling point n , the weights of the sensing signals S1~S4 are analogous.

In step S336, the mixed waveform is finally obtained by the calculation formula (5) of the algorithm of the present invention, and the formula (5) is as follows:

Where W _i [n] is the weight assigned to each sampling point n of the i- th channel, and this weight is affected by the signal quality of the previous sampling point n-1 ; multiplied by the signal S _i [n] of the sampling point n In order to adjust the signal quality according to the signal quality, the better the signal quality, the greater the weight obtained to ensure the stability of the signal quality; the resulting S _mix [n] is the output of each sample point n through the algorithm of the present invention. result.

FIG. 6 is a graph showing the waveform of each channel (PPG Channel 1~Channel 4) signal after a rapid horizontal movement of the mouse 10 provided with the signal processing system 100 of FIG. 1A, through an arithmetic average algorithm (Avg. ) algorithm and the weighted average (weighted Avg.) according to the present invention, the result of the mixing signal waveform peak W _f electrocardiogram signal (Reference ECG) compared to R wave is obtained. The x-axis of these waveforms indicates that the ADC value is an analog-to-digital converted value. Compares the time interval between two adjacent peaks and the number of peaks, the highest level can be seen that the number of R-wave peaks of the electrocardiogram signal waveform obtained by the weighted average algorithm hybrid waveform W _f of the anastomosis. The mixed waveform obtained by the weighted average algorithm has higher signal stability than the mixed signal of each single channel and the arithmetic average calculation. Assisting peak detection by dynamically setting the time threshold and slope threshold improves the accuracy of heart rate calculations.

7 is a waveform of the signals of each channel (PPG Channel 1~Channel 4) after a mouse is held by the mouse 10 with the signal processing system 100 of FIG. 1A, and the average algorithm (Avg) .) and the results of the weighted average algorithm (Weighted Avg.) obtained by comparing the peak of the mixed signal waveform W _f with the R wave of the ECG waveform. At the bottom is the mouse click record. The signal from PPG Channel 1 to PPG Channel 4 finds that the browsing or clicking action will cause the signal to be unstable, and the waveform mixed by the weighted average algorithm can effectively stabilize the signal quality, thus reducing the peak detection error. rate.

The method of the invention can effectively improve the probability of detecting the heart rate, and can obtain a mixed waveform which is more stable than the signal quality of a single channel, can improve the usability and reliability of the signal, thereby reducing the movement generated when the mouse is used. The influence of noise can improve the accuracy of heart rate calculation.

The above is only the preferred embodiment of the present invention, and the scope of the invention is not limited thereto, that is, the simple equivalent changes and modifications made by the scope of the invention and the description of the invention are All remain within the scope of the invention patent. In addition, any of the objects or advantages or features of the present invention are not required to be achieved by any embodiment or application of the invention. In addition, the abstract sections and headings are only used to assist in the search of patent documents and are not intended to limit the scope of the invention.

Claims (5)

A signal processing method for detecting a pulse of a palm of a multi-sensor comprises: setting a plurality of sensors above a back shell of a mouse for a user to cover the sensors with a palm; using the same The sensor senses pulse changes in different parts of the palm to generate a sensing signal; providing a peak detecting process for performing peak detection on each of the sensing signals at a plurality of sampling points, wherein the sampling points are respectively Include a current sampling point and a previous sampling point; performing an algorithm using a microprocessor, wherein the algorithm includes: if the sensing signal is detected at the current sampling point, a peak is present The weight set by a sampling point is added by a value to obtain a new weight; if the sensing signal is not detected at the current sampling point, the weight set by the previous sampling point is reduced. Determining the value to obtain the new weight; and performing a weighted average calculation on the sensing signals according to the new weight of each of the sensing signals to process the sensing signals into a mixed waveform; Waveform The one user calculates heart rate value; and a microprocessor and an output waveform of the mixing of the heart rate to a display interface. The signal processing method for multi-sensor detecting palm pulse according to claim 1, wherein the peak detecting process includes a dynamic threshold setting method for determining the current sampling point of the sensing signal. Whether the time difference from the previous peak is reasonable, the steps include: Determining an upper time limit and a lower time limit according to a time difference between the first two peaks of the current sampling point; and if the time difference between the current sampling point and the previous peak falls between the upper limit of the time and the lower limit of the time, the The microprocessor calculates a slope value for the current sample point. The signal processing method for multi-sensor detecting palm pulse according to claim 2, wherein the dynamic threshold setting method further comprises: providing a slope threshold and a flag; comparing the slope value with the slope The size of the threshold; if the slope value is greater than the slope threshold, the flag is set to true; when the flag is true, the slope value is determined to be negative or equal to zero; and when the flag is true, if the flag is true If the slope value is negative or equal to zero, a peak position is recorded and a maximum slope value in the buffer memory is used to calculate a new slope threshold. The signal processing method for multi-sensor detecting palm pulse according to claim 3, wherein the step of calculating the new slope threshold comprises: multiplying the maximum slope value in the buffer memory by one The slope is judged constant. The signal processing method for multi-sensor detecting palm pulse according to claim 4, wherein the new weight of each of the sensing signals has a minimum value of 1, and a maximum value of 20.