PPG signal acquisition equipment stability evaluation method based on information similarity

Abstract

The invention relates to a PPG signal acquisition equipment stability evaluation method based on information similarity, and belongs to the technical field of PPG signal and acquisition equipment analysis. The method comprises the following steps: acquiring an original PPG signal in PPG signal acquisition equipment; preprocessing the original signal; feature point positioning and feature parameter calculation are carried out on the preprocessed signals, and downsampling processing is carried out on the preprocessed signals; pPG original signals acquired by different devices or different individuals are processed; and performing information similarity algorithm processing on the PPG signals subjected to downsampling processing and the characteristic parameters of the PPG signals in pairs. According to the method, PPG signals are directly processed, the method has the advantages of being simple, small in calculation amount and the like, on the basis of the method, the stability of the equipment for signal collection of the same individual and the specificity of signal collection of different individuals can be conveniently evaluated, and the stability and repeatability of the equipment are evaluated from the perspective of the collected signals.

Description

Technical Field

[0001] The present invention belongs to the technical field of PPG signal and its acquisition device analysis, and relates to a method for evaluating the stability of a PPG signal acquisition device based on information similarity. Background Art

[0002] Photoplethysmography (PPG) is a simple, low-cost, non-invasive optical technology that uses the absorption and scattering characteristics of light to detect changes in blood flow in subcutaneous microvascular tissues, and has been widely used in the fields of medical monitoring and wearable devices.

[0003] The types of PPG signal acquisition devices are becoming increasingly rich, and they play an important role in monitoring blood oxygen, pulse rate, non-invasive blood glucose, and hemodynamic status. However, due to the differences in their signal conditioning circuits, as well as the influence of component errors and external environments on the same device, the instability and inconsistency of the acquisition results may occur, which will affect the quality of the PPG signal and the utilization of subsequent data. Therefore, ensuring the stability and reliability of the acquired PPG signal is crucial for obtaining accurate physiological information, and the quality assessment of the PPG signal is an essential link.

[0004] At present, the evaluation of the stability of PPG signal acquisition devices mainly relies on material traceability, error testing, and whole machine testing at the hardware level; while the evaluation of the quality of PPG signals involves techniques such as time-frequency analysis, machine learning, and statistical analysis. However, some evaluation methods have some deficiencies, such as long test cycles, high computational complexity, and strict required conditions, etc., which need to be considered and improved in practical applications. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method for evaluating the stability of a PPG signal acquisition device based on information similarity, which directly processes the PPG signal itself and has the advantages of simple method and small computational amount.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] A method for evaluating the stability of a PPG signal acquisition device based on information similarity, specifically including the following steps:

[0008] S1: Data acquisition, obtaining the original PPG signal in the PPG acquisition device;

[0009] S2: Performing signal preprocessing on the original PPG signal obtained in step S1;

[0010] S3: Locate the feature points of the preprocessed PPG signal obtained in step S2;

[0011] S4: Calculate the feature parameters based on the PPG feature points obtained in step S3;

[0012] S5: Perform downsampling on the preprocessed PPG signal obtained in step S2;

[0013] S6: Repeat steps S1 - S5 to obtain another set of PPG signals and feature parameters of the same device or different devices, the same individual or different individuals;

[0014] S7: Perform information similarity algorithm processing on the two sets of PPG signals and feature parameters obtained in steps S1 - S6;

[0015] S8: Repeat steps S1 - S7 to obtain multiple sets of information similarity algorithm processing results;

[0016] S9: Analyze the results obtained in step S8 to draw a conclusion on the stability of the PPG signal acquisition device.

[0017] Further, step S2 specifically includes the following steps:

[0018] S21: Perform digital low - pass filtering on the original PPG signal to remove high - frequency noise interference. The frequency response of the n - order Butterworth low - pass filter is shown as follows:

[0019]

[0020] where ω is the angular frequency, ω c is the cut - off angular frequency, and n is the filter order;

[0021] In the digital low - pass filtering process, a Butterworth low - pass filter with 2 - 8 orders is used, and the cut - off frequency is set in the range of 10 - 20 Hz;

[0022] S22: Signal slicing processing, retaining the middle signal segment of the PPG signal;

[0023] S23: Remove the baseline drift to reduce the influence of breathing and movement during acquisition. Specifically, use the cubic spline interpolation algorithm to fit the baseline of the PPG signal to obtain the baseline of the PPG signal, and subtract the fitted baseline from the original signal to remove the baseline drift of the PPG signal. In each interval [x i , x i+1 , the cubic spline interpolation function is expressed as:

[0024] S i (x) = a i + b i (x - x i) + c i (x - x i ) 2 + d i (x - x i ) 3

[0025] where a i , b i , c i , d i are undetermined coefficients;

[0026] S24: Signal normalization processing to eliminate the deviation caused by the amplitude difference between different signals. The normalization formula is as follows:

[0027]

[0028] where PPGs are the PPG signal data points before normalization, mean(PPGs) is the mean of the PPG signal before normalization, and std(PPGs) is the standard deviation of the PPG signal before normalization.

[0029] Furthermore, step S3 specifically includes the following steps:

[0030] S31: Use the sliding window method to locate the main peak point (B) of the signal, obtain the peak value in each sliding window signal, and obtain the peak location of the actual PPG signal through interphase screening;

[0031] The width of the sliding window is 1 / 2 of the PPG signal sampling rate, and the sliding step is 1 / 4 of the PPG signal sampling rate;

[0032] S32: Locate the trough point (A) of the signal based on the main peak point (B) of the signal, and obtain the trough location of the PPG signal through forward search of the peak;

[0033] S33: Locate the dicrotic peak point (C) of the signal based on the main peak point (B), and obtain the dicrotic peak location of the PPG signal through backward search of the peak.

[0034] Furthermore, step S4 specifically includes the following steps:

[0035] S41: Divide the PPG signal into several pulse cycles according to the trough points;

[0036] S42: Extract the waveform features of each pulse cycle, including the main peak height h 1 , the dicrotic peak height h 2 , the rising branch time t 1 , the falling branch time t 2 , the main peak - dicrotic peak interval time t pp etc. Some of the feature calculation formulas are as follows:

[0037]

[0038] where x A is the coordinate position of the starting trough point within the pulse period, x A′ is the coordinate position of the ending trough point within the pulse period, x B is the coordinate position of the main peak point within the pulse period, x C is the coordinate position of the dicrotic wave peak point within the pulse period, and Sample_rate is the sampling rate of the PPG signal;

[0039] S43: Extract the heart rate variability features of two adjacent pulse periods, including the RR interval and the relative RR interval, etc. For n pulse periods, the feature calculation formula is as follows:

[0040]

[0041] where is the coordinate position of the main peak point within the i-th pulse period, RR i is the i-th RR interval value, rr i is the i-th relative RR interval value, and Sample_rate is the sampling rate of the PPG signal.

[0042] Furthermore, in step S5, the preprocessed PPG signal is downsampled, and the processing method is as follows:

[0043] z[n] = decimate(x[n], D, ftype);

[0044] where z[n] is the downsampled signal, x[n] is the original signal, D is the downsampling factor, ftype is the filter type, and decimate(a, b, c) represents the downsampling function.

[0045] Furthermore, step S7 specifically includes the following steps:

[0046] S71: For a continuous time interval sequence X i ={x 0 , x 1 , …, x n}, starting from x 0 , taking two consecutive interval sequences {x i , x i+1} as a group, one group of interval pairs will generate two change states of x i . Represent these two change states with 0 and 1 respectively, and perform binary conversion of the sequence. The conversion formula is as follows:

[0047]

[0048] S72: Convert the signal sequence of length n + 1 into a binary symbol sequence B of length n according to step S71 i ={I 0 , I 1 , …, I n}, take a window of length m, where m < n, starting from I 0 and moving 1 data element backward each time, divide B i into multiple binary symbol sequence sets of length m;

[0049] In step S72, the window length in the information similarity algorithm is set to 2 - 4 when the input data is PPG feature parameters, and set to 5 - 8 when the input data is the downsampled PPG signal.

[0050] S73: Each binary symbol sequence of length m represents a unique change pattern of the original data. For a set of data of length n, there are 2 m kinds of change patterns for the binary symbol sequence with window length m, and the combined set is denoted as W i ={M 0 , M 1 , …, M n-m}; Count the frequencies and probabilities of various change patterns occurring in set W, and sort them from largest to smallest according to the probability;

[0051] S74: Calculate the similarity distance between two segments of data according to the probability and ranking of each change pattern. The more similar the data, the smaller the similarity distance, and vice versa. The calculation formula for the similarity distance is as follows:

[0052]

[0053] where D m (S 1 , S 2 ) represents the similarity distance between data S 1 and S 2 , p 1 (w i ) and R 1 (w i ) respectively represent the probability and ranking of the binary symbol sequence w i appearing in data S 1 ; p 2 (w i ) and R 2 (w i ) respectively represent the probability and ranking of the binary symbol sequence w i appearing in data S 2 ;

[0054] S75: Plot the rankings of each change pattern of data S 1 against S 2 as a scatter plot. If the rankings under the same change pattern of two segments of data are similar, the scatter points will be closer to the vicinity of the diagonal line (y = x); the closer the scatter points on the entire scatter plot are to the diagonal line, the higher the similarity between the two signals, and vice versa; by calculating the average distance between each scatter point and the diagonal line, the similarity between the two segments of data can also be measured; the average distance calculation formula is:

[0055]

[0056] wherein, represents the average distance between each scatter point and the diagonal line.

[0057] Furthermore, step S9 specifically includes the following steps:

[0058] S91: Calculate the similarity distance D m between the two sets of data and the average distance between each scatter point and the diagonal line respectively;

[0059] S92: Compare the distance with the threshold value to obtain the similarity situation between the two sets of data. The comparison method is as follows:

[0060]

[0061] wherein, the similarity distance threshold D m_threshold ∈(min(D m ), max(D m )) and the average distance threshold between each scatter point and the diagonal line

[0062] The similarity distance threshold is set to 0.03 - 0.04, and the average distance threshold between each scatter point and the diagonal line is set to 3 - 6;

[0063] S93: For D m in step S92, judge the signal similarity. The smaller the value, the higher the similarity between the two sets of data. If the two sets of data are from different devices of the same individual, it indicates that the sampling consistency of the test device is good; if the two sets of data are from the same device of the same individual, it indicates that the sampling repeatability of the test device is good; if the two sets of data are from the same device of different individuals, it indicates that the sampling specificity of the test device is good; vice versa.

[0064] The beneficial effects of the present invention are as follows: The method of the present invention only needs to process the PPG signal itself. After preprocessing, feature point localization, feature parameter calculation, and downsampling of the obtained original PPG signal, the information similarity algorithm processing can be performed based on the feature parameter sequence or the downsampled PPG signal sequence to obtain the similarity information of the evaluated data, and further evaluate the stability of the PPG signal acquisition device. The present invention can be conveniently used to evaluate the stability of the device for signal acquisition of the same individual and the specificity of signal acquisition of different individuals; the processing method of the PPG signal is mature and reliable, the results are accurate and credible, and it has practicality. The information similarity algorithm used is simple and has a small amount of calculation, which is suitable for engineering applications.

[0065] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail with reference to the accompanying drawings, where:

[0067] Figure 1 It is a schematic flow chart of the method for evaluating the stability of a PPG signal acquisition device based on an information similarity algorithm proposed by the present invention;

[0068] Figure 2 It is a schematic flow chart of signal preprocessing, feature point localization, and feature parameter calculation in the method of the present invention;

[0069] Figure 3 It is a schematic flow chart of the information similarity algorithm in the method of the present invention;

[0070] Figure 4 It is a characteristic diagram of the waveform of a typical PPG signal;

[0071] Figure 5 It is the time-domain waveform of the original signal, the signal after preprocessing, the signal after feature point localization, and the signal after downsampling, as well as the calculation results of feature parameters in the method of the present invention;

[0072] Figure 6 It is a scatter plot drawn by the information similarity algorithm of the method of the present invention for the ranking of various change modes of different PPG signal data;

[0073] Figure 7 It is a heat map of the similarity distance between pairwise signals calculated by the information similarity algorithm of the method of the present invention for different PPG signal data;

[0074] Figure 8 This is a heat map of the average distance between each scatter point and the diagonal line between pairwise signals calculated by the information similarity algorithm in the method of the present invention for different PPG signal data. Detailed implementation manners

[0075] The following uses specific specific examples to illustrate the implementation manners of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0076] Please refer to Figures 1 to 8 , the present invention provides a method for evaluating the stability of a PPG signal acquisition device based on information similarity. Figure 1 As shown in the figure, which is the flow of this method. After obtaining the original PPG signal from the PPG signal acquisition device (two groups of signals are required), after the two groups of signals are processed by the same preprocessing method, Path 1: Continue with feature point positioning, and calculate relevant feature parameters (h1, h2, t1, t2, tpp, RRs, rrs, etc.) using the located feature points to obtain a sequence of feature parameters; Path 2: Downsample the preprocessed signal to obtain a downsampled PPG sequence. The purpose of downsampling is to remove the points with the same change trend in the PPG signal, so as to reduce the calculation amount in the information similarity algorithm. The information similarity algorithm can calculate for the sequence of feature parameters or the downsampled waveform sequence of two groups of PPG signals, and further obtain the signal similarity and the evaluation of the stability of the acquisition device.

[0077] Figure 2 As shown in the figure, which is the flow of signal preprocessing, feature point positioning, and feature parameter calculation. The specific flow is as follows:

[0078] (1) Perform signal preprocessing on the original PPG signal, which specifically includes the following steps:

[0079] S21: Perform digital low-pass filtering on the original PPG signal to remove high-frequency noise interference in the signal. The frequency response of the nth-order Butterworth low-pass filter is shown as follows:

[0080]

[0081] where ω is the angular frequency, ω c is the cut-off angular frequency, and n is the filter order;

[0082] In digital low-pass filtering, a Butterworth low-pass filter of order 2 to 8 is used, and the cut-off frequency is set in the range of 10 to 20 Hz;

[0083] S22: Signal slicing processing, retaining the middle signal segment of the PPG signal;

[0084] S23: Remove baseline drift and reduce the influence of breathing and movement during acquisition. Specifically, use the cubic spline interpolation algorithm to fit the baseline of the PPG signal to obtain the baseline of the PPG signal, and subtract the fitted baseline from the original signal to remove the baseline drift of the PPG signal; in each interval [x i , x i+1 , the cubic spline interpolation function is expressed as:

[0085] S i (x) = a i + b i (x - x i ) + c i (x - x i ) 2 + d i (x - x i ) 3

[0086] where a i , b i , c i , d i are undetermined coefficients;

[0087] S24: Signal normalization processing to eliminate the deviation caused by the amplitude difference between different signals. The normalization formula is as follows:

[0088]

[0089] where PPGs is the PPG signal data point before normalization, mean(PPGs) is the mean of the PPG signal before normalization, and std(PPGs) is the standard deviation of the PPG signal before normalization.

[0090] (2) Locate the feature points of the preprocessed PPG signal, which specifically includes the following steps:

[0091] S31: Use the sliding window method to locate the main peak point (B) of the signal, obtain the peak value in each segment of the sliding window signal, and obtain the peak location of the actual PPG signal through interphase screening;

[0092] The width of the sliding window is 1 / 2 of the PPG signal sampling rate, and the sliding step is 1 / 4 of the PPG signal sampling rate;

[0093] S32: Locate the signal trough point (A) based on the main signal peak point (B), and obtain the trough localization of the PPG signal through forward search of the peak.

[0094] S33: Locate the dicrotic wave peak point (C) of the signal based on the main peak point (B), and obtain the dicrotic wave peak localization of the PPG signal through backward search of the peak.

[0095] (3) Calculate characteristic parameters based on PPG characteristic points, specifically including the following steps:

[0096] S41: Divide the PPG signal into several pulse cycles according to the trough points.

[0097] S42: Extract the waveform characteristics of each pulse cycle, including the height h of the main peak 1 , the height h of the dicrotic wave peak 2 , the rising branch time t 1 , the falling branch time t 2 , the interval time t between the main peak and the dicrotic wave peak pp etc. The calculation formulas for some of the characteristics are as follows:

[0098]

[0099] Among them, x A is the coordinate position of the starting trough point within the pulse cycle, x A′ is the coordinate position of the ending trough point within the pulse cycle, x B is the coordinate position of the main peak point within the pulse cycle, x C is the coordinate position of the dicrotic wave peak point within the pulse cycle, and Sample_rate is the sampling rate of the PPG signal;

[0100] S43: Extract the heart rate variability characteristics of two adjacent pulse cycles, including the RR interval and the relative RR interval, etc. For n pulse cycles, the calculation formulas for their characteristics are as follows:

[0101]

[0102]

[0103] Among them, is the coordinate position of the main peak point within the i-th pulse cycle, RR i is the value of the i-th RR interval, rr i is the value of the i-th relative RR interval, and Sample_rate is the sampling rate of the PPG signal.

[0104] Figure 3 The information similarity algorithm process is shown as follows, specifically including the following steps:

[0105] S71: For a continuous sequence of time intervals X i ={x 0 , x 1 , …, x n}, starting from x 0 , take two consecutive interval sequences {x i , x i+1} as a group. One group of interval pairs will generate two change states of x i . Represent these two change states with 0 and 1 respectively to perform binary conversion of the sequence. The conversion formula is as follows:

[0106]

[0107] S72: Convert the signal sequence of length n + 1 into a binary symbol sequence B i ={I 0 , I 1 , …, I n} according to step S71. Take a window of length m, where m < n, and starting from I 0 , move 1 data element backward each time to divide B i into multiple binary symbol sequence sets of length m;

[0108] In step S72, the window length in the information similarity algorithm is set to 2 - 4 when the input data is PPG feature parameters, and set to 5 - 8 when the input data is the downsampled PPG signal.

[0109] S73: Each binary symbol sequence of length m represents a unique change pattern of the original data. For a set of data of length n, there are 2 m change patterns of binary symbol sequences with a window length of m. The combined set is denoted as W i ={M 0 , M 1 , …, M n-m}; Count the frequencies and probabilities of various change patterns in the set W i and sort them in descending order according to the probability;

[0110] S74: Calculate the similarity distance between two segments of data according to the probability and sorting of each change pattern. The more similar the data, the smaller the similarity distance, and vice versa. The calculation formula of the similarity distance is as follows:

[0111]

[0112] Among them, D m (S 1 , S 2 ) represents the data S 1The similarity distance with S 2 is p 1 (w i ) and R 1 (w i ) respectively represent the probability and ranking of the binary symbol sequence w i appearing in the data S 1 ; p 2 (w i ) and R 2 (w i ) respectively represent the probability and ranking of the binary symbol sequence w i appearing in the data S 2 ;

[0113] S75: Plot the ranking of each change pattern of the data S 1 and S 2 as a scatter plot. If the rankings of the same change pattern of the two segments of data are similar, the scatter points are closer to the vicinity of the diagonal line (y = x); the closer the scatter points on the entire scatter plot are to the diagonal line, the higher the similarity of the two signals, and vice versa; by calculating the average distance between each scatter point and the diagonal line, the similarity of the two segments of data can also be measured; the average distance calculation formula is:

[0114]

[0115] wherein, represents the average distance between each scatter point and the diagonal line.

[0116] Figure 4 As shown in the figure is the typical PPG signal waveform feature diagram, Figure 4 in which, point A is the trough point of the PPG signal, point B is the main peak point of the PPG signal, point C is the dicrotic wave peak point of the PPG signal, point D is the maximum slope point of the rising branch of the PPG signal, point E is the maximum slope point (absolute value) of the falling branch of the PPG signal, point F is the dicrotic notch of the PPG signal; h1 is the height of the main peak of the PPG signal, h2 is the height of the dicrotic wave peak of the PPG signal; t1 is the rising branch time of the PPG signal, t2 is the falling branch time of the PPG signal, and tpp is the time interval from the main peak to the dicrotic wave peak of the PPG signal.

[0117] Figure 5 Shows the time domain waveforms of the original signal, the signal after preprocessing, the signal after feature point positioning, and the signal after downsampling, as well as the calculation results of the characteristic parameters, Figure 5Among them, preprocessing, downsampling, feature point localization, feature parameter calculation and other operations are performed on an original PPG signal with a sampling frequency of 200 Hz and a length of 3000 points. In preprocessing: digital low-pass filtering is performed using a 6th-order Butterworth low-pass filter with a cut-off frequency of 15 Hz; in slicing, the data of the middle 2000 points of the signal segment is retained; the baseline drift removal and normalization formulas are described in detail in the specification. In downsampling: the set downsampling factor is 10, and the filter type is "iir". In feature point localization: the window width and sliding step of the sliding window used are described in detail in the specification, and the trough point A, main peak point B, and dicrotic peak point C of the preprocessed PPG signal are located. It can be seen from the figure that the localization is accurate. In feature parameter calculation: 7 characteristic parameter sequences of each complete pulse beat are calculated. The meanings and calculation methods of the 7 characteristic parameters are described in detail in the specification. In particular: before calculating the characteristic parameters, it is necessary to ensure that the number of main peak points is equal to the number of dicrotic peak points, and 1 less than the number of trough points, so as to ensure that each main peak point and dicrotic peak point are in a complete pulse beat.

[0118] Figure 6 , Figure 7 , Figure 8 Among them, data of 3 people are collected using the same PPG signal acquisition device, a total of 7 groups (S1-S7). Among them, S1-S5 are collected from the same person within 20 minutes, and S6 and S7 come from the other 2 people respectively. The information similarity algorithm is used to analyze each pair of signals in each group. The window length m = 6 is taken, and the change patterns and their probabilities and rankings appearing in each group of PPG signals are counted. Visual analysis is performed on the ranking of each signal, and a scatter plot between each signal is drawn. Each point represents a change pattern ( Figure 5 ). When the similarity between two signals is higher, the distribution of their points will be closer to the diagonal line, and vice versa, it will be farther away from the diagonal line. Calculate the similarity distance of each signal and the average distance of each change pattern from the diagonal line, and draw a heat map of the similarity results of each signal to visualize the data ( Figure 6 , Figure 7 ). The darker the color of the square in the heat map, the higher the similarity between the two signals corresponding to this square, and vice versa. Analysis Figures 5 to 7 shows that signals S1-S5 can be considered to have high similarity with each other, and the similarity between S1-S5 and S6, S7 is low. The method can reflect the stability and specificity of the evaluated device for data of the same device for the same individual and different individuals of the same device.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the present technical solution, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A PPG signal acquisition device stability evaluation method based on information similarity, characterized in that: The method specifically comprises the following steps: S1: Obtain the original PPG signal from the PPG acquisition device; S2: performing signal preprocessing on the original PPG signal obtained in step S1; S3: Positioning feature points of the preprocessed PPG signal obtained in step S2; S4: Calculating feature parameters based on the PPG feature points obtained in step S3; S5: downsampling the pre-processed PPG signal obtained in step S2; S6: Repeat steps S1 to S5 to obtain another set of PPG signals and characteristic parameters of the same device or different devices, the same individual or different individuals; S7: Processing the two groups of PPG signals and characteristic parameters obtained in steps S1 to S6 using an information similarity algorithm; S8: Repeat steps S1 to S7 to obtain multiple sets of information similarity algorithm processing results; S9: Analyze the results obtained in step S8 to draw a conclusion on the stability of the PPG signal acquisition device.

2. The PPG signal acquisition device stability evaluation method according to claim 1, characterized in that: Step S2 specifically includes the following steps: S21: Perform digital low-pass filtering on the original PPG signal to remove high-frequency noise interference in the signal; S22: signal slicing processing, retaining the intermediate signal segments of the PPG signal; S23: removing the baseline drift, specifically using a cubic spline interpolation algorithm to fit the baseline to obtain the baseline of the PPG signal, and subtracting the fitted baseline from the original signal to remove the baseline drift of the PPG signal; S24: Signal normalization processing to eliminate the deviation caused by the amplitude difference between different signals.

3. The PPG signal acquisition device stability evaluation method according to claim 1, characterized in that: Step S3 specifically includes the following steps: S31: using a sliding window method to locate the main peak point B of the signal, obtaining the peak value in each sliding window signal, and obtaining the peak location of the actual PPG signal through interval screening; S32: Locate the signal trough point A based on the main wave peak point B of the signal, and obtain the trough location of the PPG signal by forward searching the wave peak; S33: Locate the dicrotic wave peak point C of the signal based on the main wave peak point B, and obtain the dicrotic wave peak location of the PPG signal by backward peak search.

4. The PPG signal acquisition device stability evaluation method according to claim 1, characterized in that: Step S4 specifically includes the following steps: S41: Divide the PPG signal into a number of pulse cycles according to the trough points; S42: Extract the waveform features of each pulse cycle, including the main wave peak height h1, the dicrotic wave peak height h2, the rising branch time t1, the falling branch time t2, the main wave peak-dicrotic wave peak interval time t pp ; S43: extracting heart rate variability features of two pulse cycles, including RR intervals and relative RR intervals.

5. The PPG signal acquisition device stability evaluation method according to claim 1, characterized in that: In step S5, the pre-processed PPG signal is downsampled, and the processing method is as follows: z[n]=decimate(x[n], D, ftype); Wherein, z[n] is the signal after downsampling, x[n] is the original signal, D is the downsampling factor, ftype is the filter type, and decimate(a, b, c) represents the downsampling function.

6. The PPG signal acquisition device stability evaluation method according to claim 1, characterized in that: Step S7 specifically includes the following steps: S71: For a continuous time interval sequence X i ={x0, x1, ..., x n }, starting from x0, two consecutive interval sequences {x i , x i+1 } As a group, a set of interval pairs will produce x i The two change states are represented by 0 and 1 respectively, and the sequence is converted into binary form. The conversion formula is as follows: S72: According to step S71, the signal sequence with a length of n+1 is converted into a binary symbol sequence B with a length of n. i ={I0, I1, ..., I n }, take a window of length m, m<n, start from I0 and move back one data element at a time, and replace B i Divide into multiple sets of binary symbol sequences with a length of m; S73: Each binary symbol sequence of length m represents a unique variation pattern of the original data. For a set of data of length n, there are 2 binary symbol sequences of window length m. m The combination of these patterns is denoted as W i ={M0, M1, ..., M n-m }; Statistical set W i The frequency and probability of various change modes appearing in the data, and sort them from large to small according to the probability; S74: Calculate the similarity distance between two segments of data based on the probability and ranking of each change pattern. The more similar the data is, the smaller the similarity distance is, and vice versa. The calculation formula of the similarity distance is as follows: Among them, D m (S1, S2) represents the similarity distance between data S1 and S2, p1(w i ) and R1(w i ) represent the binary symbol sequence w i The probability and ranking of appearing in data S1; p2(w i ) and R2(w i ) represent the binary symbol sequence w i Probability and ranking of appearance in data S2; S75: Plot the rankings of the change modes of data S1 and S2 into a scatter plot. If the rankings of the two data segments under the same change mode are similar, the closer the scatter points are to the diagonal line; the closer the scatter points on the entire scatter plot are to the diagonal line, the higher the similarity of the two signals, and vice versa; by calculating the average distance between each scatter point and the diagonal line, the similarity of the two data segments can also be measured; the average distance calculation formula is: in, Represents the average distance between each scattered point and the diagonal line.

7. The PPG signal acquisition device stability evaluation method according to claim 6, characterized in that: In step S72, the window length in the information similarity algorithm is set to 2-4 when the input data is a PPG feature parameter, and is set to 5-8 when the input data is a downsampled PPG signal.

8. The PPG signal acquisition device stability evaluation method according to claim 6, characterized in that: Step S9 specifically includes the following steps: S91: Calculate the similarity distance D between the two sets of data respectively m and the average distance between each scattered point and the diagonal line S92: Compare the distance with the threshold to obtain the similarity between the two sets of data. The comparison method is as follows: Among them, the similarity distance threshold D m_threshold ∈(min(D m ),max(D m )), the average distance threshold between each scattered point and the diagonal line S93: For D in step S92 m and To judge the signal similarity, the smaller the value, the higher the similarity between the two sets of data. If the two sets of data come from the same individual but different devices, it means that the sampling consistency of the test device is good; if the two sets of data come from the same individual but the same device, it means that the sampling repeatability of the test device is good; if the two sets of data come from different individuals but the same device, it means that the sampling specificity of the test device is good; and vice versa.

Images