Pulse wave signal processing method, device and system

Abstract

The invention provides a pulse wave signal processing method and device. The method comprises the following steps: acquiring and storing an original pulse wave signal; intercepting the original pulse wave signal by using a sliding time window function, constructing a fitting function, and adjusting parameters of the sliding time window function and / or parameters of the fitting function, so that fitting signals corresponding to the intercepted signal segments approach to the signal segments; generating a pulse wave signal set according to the signal segments or the corresponding fitting signals; splicing signal segments in the pulse wave signal set into signal segments with certain signal characteristics to obtain a target periodic signal set; detecting signal segments in the target periodic signal set, and extracting heart rhythm signal features and signal features between heart rhythm beats as secondary signal features; and determining to output a normal pulse wave signal or determining motion artifacts or noise according to the secondary signal characteristics. By means of the scheme, extraction of high-quality pulse wave signals in a complex interference state and a large motion state can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of biomedical signal processing, and in particular to a pulse wave signal processing method, device and system. Background Art

[0002] The pulse wave is a signal formed by the pulsation generated by the beating of the human heart, which propagates along the arteries and blood flow to the periphery of the body. The pulse wave can be collected by signal sensors such as pressure sensors, photoelectric sensors, and ultrasonic sensors. The pressure sensor can be placed directly in the artery or on the skin surface close to the artery to measure the pressure changes of the pulse wave and convert them into electrical signals. The photoelectric sensor records the changes in blood vessel volume during the cardiac cycle by detecting the difference in the intensity of transmitted or reflected light absorbed by the human blood and tissues. This change can be converted into an electrical signal synchronized with the pulse wave. The ultrasonic sensor is placed on the skin surface close to the artery, sending ultrasonic signals through the skin to reach the artery and its surrounding tissues and generate reflected echoes. The sensor detects these echo changes and converts them into electrical signals synchronized with the pulse wave.

[0003] Regardless of which sensor is used, the collected pulse wave signal is easily affected by multiple factors such as the measured human body's own activities, the movement or position change of the collection device itself, and changes in the surrounding environment, generating interference signals of various patterns and characteristics. These interference signals are superimposed on the desired biological signal, resulting in baseline drift, waveform distortion, signal distortion, etc., causing the quality of the biological signal to deteriorate, and having a negative impact on subsequent signal processing analysis and extraction of physiological information components.

[0004] In addition, since the contribution of the motion component in the pulse wave signal often exceeds the contribution of the human body signs by more than a certain amount when the human body is in motion, motion artifacts will be generated in the pulse wave signal, which will cause erroneous interpretation and deteriorate the accuracy and reliability of the estimation of cardiovascular parameters.

[0005] Therefore, how to eliminate interference and motion artifacts in pulse wave signals and improve the quality of pulse wave signals is an important issue that the industry needs to solve. Summary of the invention

[0006] The present invention provides a pulse wave signal processing method, device and system to effectively remove interference noise and motion artifacts in the pulse wave signal and realize the extraction of high-quality pulse wave signals in complex interference states and large motion states.

[0007] To this end, the present invention provides the following technical solutions:

[0008] The present invention provides a pulse wave signal processing method, the method comprising:

[0009] Acquire and store original pulse wave signals;

[0010] The original pulse wave signal is intercepted by using a sliding time window function, and a fitting function is constructed, and the sliding time window function parameters and / or fitting function parameters are adjusted so that the fitting signals corresponding to the intercepted signal segments are close to the signal segments, and a fitting signal segment corresponding to the signal segment is obtained; a pulse wave signal set is generated according to the signal segment or the corresponding fitting signal segment;

[0011] splicing the signal segments in the pulse wave signal set into signal segments with certain signal characteristics to obtain a target periodic signal set;

[0012] Detecting the signal segments in the target periodic signal set, extracting the heart rhythm signal features and the signal features between heart rhythm beats as secondary signal features;

[0013] According to the secondary signal characteristics of the signal segments in the target periodic signal set, it is judged whether these signal segments meet the decision conditions; if the decision conditions are met, these signal segments are spliced ​​and output as normal pulse wave signals; if the decision conditions are not met, these signal segments are regarded as motion artifacts or noise.

[0014] Optionally, generating a pulse wave signal set according to the signal segment includes: recording the signal segment into the pulse wave signal set.

[0015] Optionally, generating a pulse wave signal set according to the signal segments includes:

[0016] generating a fitting signal corresponding to each signal segment;

[0017] The fitting signal is recorded in the pulse wave signal set.

[0018] Optionally, the fitting function is a monotonic function or a periodic function.

[0019] Optionally, constructing a fitting function and adjusting the sliding time window function parameters and / or fitting function parameters so that the fitting signal corresponding to each intercepted signal segment approaches the signal segment includes:

[0020] Constructing a fitting function, and using the fitting function to generate a fitting signal corresponding to the signal segment;

[0021] Calculating signal characteristics of the fitting signal and corresponding signal segments;

[0022] If the signal characteristics of the fitting signal and the corresponding signal segment meet the set conditions, the fitting signal is used as the fitting signal segment corresponding to the signal segment;

[0023] Otherwise, the time window function parameters and / or fitting function parameters are adjusted until the signal characteristics of the fitting signal and the corresponding signal segment meet the set conditions.

[0024] Optionally, the step of splicing the signal segments in the pulse wave signal set into signal segments with certain signal characteristics comprises: sequentially splicing a plurality of signal segments in the pulse wave signal set that are adjacent in time sequence and have signal characteristics that satisfy set conditions into a signal segment with certain signal characteristics.

[0025] Optionally, detecting whether the signal segment in the target periodic signal set meets the decision conditions based on the secondary signal characteristics of the signal segment includes: judging whether the own signal characteristics of two adjacent signal segments in the target periodic signal set, the similarity and correlation between the two signal segments, the variability between the signal segments, and the continuity of the signal segments meet the decision conditions based on the secondary signal characteristics.

[0026] Optionally, any one of the following methods is used to determine whether the signal segments meet the decision condition: voting method, stacking method, ensemble learning method, Bayesian fusion method.

[0027] Optionally, the signal characteristics include any one or more of the following: obvious monotonicity or periodicity, strong similarity or correlation between cycles, relatively slow variation between cycles, and certain continuity.

[0028] Optionally, the monotonicity or periodicity of the signal is calculated using any of the following methods: a time domain analysis method, a frequency domain analysis method, or a statistical method.

[0029] Optionally, any of the following methods is used to calculate the similarity or correlation between signals: distance measurement method, statistical feature comparison method, correlation analysis method, time-frequency analysis method, machine learning or deep learning method.

[0030] Optionally, any one of the following methods is used to calculate the variability between signals: time domain analysis method, autocorrelation or cross-correlation analysis method, frequency domain analysis method, statistical method.

[0031] Optionally, the method further includes: before intercepting the original pulse wave signal, performing noise reduction and filtering processing on the original pulse wave signal.

[0032] Optionally, the filtering processing of the original pulse wave signal includes: using any one of the following filtering methods to perform noise reduction and filtering processing on the original pulse wave signal: low-pass filtering, high-pass filtering, band-pass filtering, median filtering, sliding window filtering, wavelet transform, Kalman filtering, empirical mode decomposition EMD, variational mode decomposition VMD, local mean decomposition LMD.

[0033] The present invention also provides a pulse wave signal processing device, the device comprising:

[0034] A signal acquisition module, used for acquiring and storing the original pulse wave signal;

[0035] A segmentation module is used to intercept the original pulse wave signal using a sliding time window function, construct a fitting function, adjust the sliding time window function parameters and / or fitting function parameters, so that the fitting signal corresponding to each intercepted signal segment approaches the signal segment; and generate a pulse wave signal set according to the signal segment or the corresponding fitting signal;

[0036] A splicing module, used for splicing the signal segments in the pulse wave signal set into signal segments with certain signal characteristics to obtain a target periodic signal set;

[0037] A signal detection module is used to detect the signal segments in the target periodic signal set, extract the secondary signal features of the signal segments and the signal segments; detect whether the signal segments meet the decision conditions according to the secondary signal features of the signal segments in the target periodic signal set; if the decision conditions are met, these signal segments are spliced ​​and output as normal pulse wave signals; if the decision conditions are not met, these signal segments are regarded as motion artifacts or noise.

[0038] Optionally, the device further includes: a preprocessing module, configured to perform noise reduction and filtering processing on the original pulse wave signal before intercepting the original pulse wave signal.

[0039] The present invention also provides a pulse wave signal processing system, the system comprising: a signal acquisition device, and the pulse wave signal processing device;

[0040] The acquisition device is used to acquire human pulse wave signals and transmit the acquired pulse wave signals of a certain time length to the pulse wave signal processing device;

[0041] The pulse wave signal processing device is used to process the human body pulse wave signal and output a normal pulse wave signal.

[0042] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the pulse wave signal processing method are executed.

[0043] The present invention also provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes the steps of the pulse wave signal processing method when running the computer program.

[0044] The pulse wave signal processing method, device and system provided by the present invention perform noise reduction and filtering processing on the original pulse wave signal, then perform segmentation processing on the original pulse wave signal using a sliding time window function, fit the segmented signal segments by artificially generating a fitting signal, adjust the segmentation parameters and the fitting signal parameters, so that the error between the fitting signal and the signal segment is less than a preset value, generate a pulse wave signal set from the signal segment or the corresponding fitting signal segment, splice the signal segments in the pulse wave signal set into signal segments with certain signal characteristics, then extract the secondary characteristics of the signal segments, and splice the signal segments that meet the decision rules into a complete signal according to the pulse wave signal characteristics, the signal characteristics between adjacent signals, and the continuity of the signal segments through a decision rule fusion method, so that the signal interference and motion artifacts in the pulse wave signal can be adaptively filtered out, and the extraction of high-quality pulse wave signals in complex interference states and large motion states is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 is a flow chart of the pulse wave signal processing method provided by the present invention;

[0047] Figure 2 is a flow chart of generating a fitting signal corresponding to a signal segment in an embodiment of the present invention;

[0048] Figure 3 It is a structural schematic diagram of a pulse wave signal processing device provided by the present invention;

[0049] Figure 4 It is a structural schematic diagram of the pulse wave signal processing provided by the present invention. DETAILED DESCRIPTION

[0050] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0051] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0052] Using pulse waves to analyze and calculate blood pressure, blood oxygen saturation, and other physiological characteristics usually requires high-quality pulse wave signals as basic data. High-resolution signals can more accurately reflect the details of the pulse wave and the dynamic changes of the cardiovascular system; continuous signals are helpful for data analysis and result output during long-term monitoring, ensuring the continuity and comparability of data; stable signals help improve the signal-to-noise ratio and reduce measurement errors caused by interference or motion artifacts. Therefore, removing signal interference and motion artifacts from the pulse wave signal collected by the sensor can directly improve the accuracy and reliability of blood pressure, blood oxygen saturation, and other physiological characteristics analysis.

[0053] To this end, the present invention provides a pulse wave signal processing method and device, which performs noise reduction and filtering processing on the original pulse wave signal, and then uses a sliding time window function to segment it. The segmented signal segments are first spliced ​​into signal segments with certain signal characteristics, and then these signal segments are spliced ​​into a complete signal based on the pulse wave signal characteristics, the correlation between adjacent signals, and the continuity of the signal segments.

[0054] like Figure 1 FIG. 1 is a flow chart of a pulse wave signal processing method provided by the present invention, comprising the following steps:

[0055] In step 101, an original pulse wave signal is acquired and stored.

[0056] The original pulse wave signal may be an original pulse wave signal collected by a sensor.

[0057] In a non-limiting embodiment, the original pulse wave signal may be subjected to noise reduction and filtering to filter out baseline drift and obtain a preprocessed signal.

[0058] The purpose of filtering the original pulse wave signal is to filter out baseline drift. Specifically, the original pulse wave signal can be denoised and filtered by any of the following filtering methods, but not limited to: low-pass filtering, high-pass filtering, band-pass filtering, median filtering, sliding window filtering, wavelet transform, Kalman filtering, empirical mode decomposition (EMD) and its extension method, variational mode decomposition (VMD) and its extension method, local mean decomposition (LMD) and its extension method.

[0059] For example, the original analog signal of the pulse wave is obtained from the sensor with a sampling frequency of 200Hz. The collected analog signal is converted into a decimal original digital signal through 32-bit analog-to-digital conversion. This signal may contain interference signals from various sources, such as DC components, baseline drift, motion artifacts, or electromagnetic interference.

[0060] A Butterworth digital bandpass filter is used, and the passband frequency of the filter is set to 0.5Hz to 20Hz, the passband ripple is better than 3dB, the stopband frequency is less than 0.25Hz and greater than 30Hz, and the stopband attenuation is better than 30dB. The original digital signal is filtered to obtain a preprocessed signal. Compared with the original signal, the preprocessed signal has removed unnecessary DC components, baseline drift, low-frequency and high-frequency components, and retained the main pulse wave information.

[0061] In step 102, the original pulse wave signal is intercepted by using a sliding time window function, and a fitting function is constructed, and the sliding time window function parameters and / or fitting function parameters are adjusted so that the fitting signals corresponding to each intercepted signal segment are close to the signal segment; and a pulse wave signal set is generated according to the signal segment or the corresponding fitting signal.

[0062] When intercepting the preprocessed signal, a time window function with a length of w may be preset, and a signal with a length of w may be intercepted from the preprocessed signal according to the time window function to obtain a plurality of signal segments.

[0063] In some embodiments, each signal segment finally intercepted may be recorded in the pulse wave signal set, and motion artifacts therein may be adaptively eliminated by splicing these signal segments to obtain a high-quality pulse wave signal.

[0064] In other embodiments, the fitting signals corresponding to the finally intercepted signal segments may be recorded in the pulse wave signal set, and the motion artifacts therein may be subsequently eliminated by splicing these fitting signal segments adaptively to obtain a high-quality pulse wave signal.

[0065] The constructed fitting function may be a monotonic function or a periodic function, which is not limited in the embodiment of the present invention.

[0066] Corresponding to each signal segment, a fitting signal with a length of w is generated using the fitting function. By adjusting the time window function parameters and / or the fitting function parameters, the fitting signal can be made to approach the corresponding signal segment.

[0067] like Figure 2 FIG. 1 is a flowchart of generating a fitting signal corresponding to a signal segment in an embodiment of the present invention, comprising the following steps:

[0068] Step 201, construct a fitting function.

[0069] Step 202: Generate a fitting signal corresponding to the intercepted signal segment using a fitting function.

[0070] Step 203: Calculate signal features of the fitting signal and the corresponding signal segments.

[0071] The signal characteristics may include, but are not limited to, any one or more of the following: obvious monotonicity or periodicity (such as temporal repeatability of typical feature points or segments, periodic time-frequency domain characteristics, etc.), strong similarity or correlation between periods (such as a small Euclidean distance between signals in periods, a large correlation coefficient, etc.), slower variability of signal characteristics between periods (such as the amplitude change rate, rising edge slope change rate, peak-to-peak value change rate, etc. that are consistent with a normal pulse wave), etc.

[0072] The calculation of the signal periodicity characteristics may be performed by, but not limited to, any of the following methods:

[0073] (1) Time domain analysis method

[0074] For example, calculate whether the signal segments with time domain characteristics such as peaks, valleys, and the same slope in the time domain waveform are repeated; calculate the autocorrelation function of the signal to observe whether it has significant peaks under certain specific delays.

[0075] (2) Frequency domain analysis method

[0076] For example, when performing Fourier transform on a signal, by performing discrete Fourier transform (DFT) calculation on the signal, a periodic signal will usually show discrete frequency components in the spectrum.

[0077] (3) Statistical methods

[0078] For example, we can calculate the periodicity index of the signal, such as the periodic intensity (PPI) or the periodic factor (PF); and analyze the energy distribution of the signal at different frequencies. Periodic signals usually have a higher energy concentration at a specific frequency.

[0079] The calculation of the similarity or correlation of the signals may be performed by, but not limited to, any of the following methods:

[0080] (1) Distance measurement method

[0081] For example, the Euclidean distance, Manhattan distance, etc. between the intercepted signal segment and the corresponding fitted signal can be calculated to evaluate the similarity or correlation between the two signals.

[0082] (2) Statistical feature comparison method

[0083] For example, the statistical characteristics of the intercepted signal segment and the fitted signal, such as mean, variance, cycle length, time length, slope, amplitude, kurtosis, skewness, etc., are compared to evaluate the similarity or correlation between the two signals.

[0084] (3) Similarity or correlation analysis method

[0085] For example, the Pearson correlation coefficient, cross-correlation function, etc. of the intercepted signal segment and the fitted signal are calculated to evaluate the similarity or correlation between the two signals.

[0086] (4) Time-frequency analysis method

[0087] For example, through Fourier transform and wavelet transform, the intercepted signal fragment and the fitting signal are decomposed into wavelet components of different frequencies, and then the energy or characteristics of these wavelet components are compared to evaluate the similarity or correlation between the two signals; or, the frequency domain characteristics and time domain characteristics of the target signal and the fitting signal are calculated and compared to evaluate the similarity or correlation between the two signals.

[0088] (5) Machine learning methods.

[0089] For example, machine learning techniques such as support vector machine (SVM), neural network, etc. are used for pattern recognition and classification to evaluate the similarity or correlation between the intercepted signal fragment and the fitted signal.

[0090] Step 204 , determining whether the signal characteristics of the fitting signal and the corresponding signal segment meet the set conditions; if yes, executing step 205 ; otherwise, executing step 206 .

[0091] Step 205: Record the fitting signal into the pulse wave signal set.

[0092] Step 206 , adjusting the time window function parameters and / or the fitting function parameters, and then returning to step 202 .

[0093] It should be noted that if the time window function is adjusted, the signal segment corresponding to the fitting signal will also change, that is, the starting point and the end point of the signal segment will be re-determined.

[0094] For each captured signal segment, follow the above Figure 2 The process shown generates the corresponding fitting signal and the corresponding target signal.

[0095] Each target signal is recorded as each finally intercepted signal segment into the pulse wave signal set, or each fitting signal is recorded into the pulse wave signal set.

[0096] The following uses the distance measurement method as an example to further illustrate the process of generating a pulse wave signal set in step 102.

[0097] A time window function with a length of w is preset. The initial value of w can be set to 0.25s, for example. The value of w can be optimized and determined based on the sampling frequency of the pulse wave signal, the signal-to-noise ratio of the pulse wave signal, the heart rhythm cycle characteristics contained in the pulse wave signal, etc., and the time window function is optimized based on experience.

[0098] Parameters can improve computational efficiency and the resolution of fragment decomposition.

[0099] A maximum allowable fitting error ε is preset. The initial value of ε can be set to 8, for example. Optimizing ε based on experience can optimize the computational efficiency and the resolution of fragment decomposition.

[0100] The starting point A0 of the preprocessed signal is taken as the starting point, a signal with a length of w is intercepted, the end point is B0, and the intercepted signal segment S0 is taken as the target signal.

[0101] According to the characteristics of the target signal S0, a quadratic polynomial is constructed as the fitting function. The quadratic polynomial function formula is as follows:

[0102] f(x)=ax 2 +bx+c (1)

[0103] According to the fitting function, a fitting signal C0 with a length of w is generated, and the fitting signal C0 is made as close to the target signal S0 as possible by adjusting the parameters a, b, and c of the fitting function.

[0104] The distance d between two points is calculated as follows:

[0105]

[0106] Select point A on the target signal S0 i , calculate point Ai The distance d to each point on the fitting signal C0, select the minimum value and record it as A i The distance d from the point to the fitting signal C0.

[0107] Adjust the time window function parameters and the fitting signal parameters a, b, c. When the distance d from all points on the target signal S0 to the fitting signal C0 is not greater than the maximum allowable fitting error ε, record the starting point A0 and the end point B0 of the current target signal S0, and record the target signal S0 in the pulse wave signal set; or record the starting point A0′ of the current fitting signal C0, and record the fitting signal C0 in the pulse wave signal set.

[0108] The above steps are repeated until the preprocessed signal is completely processed, the preprocessed signal is divided into a number of target signal segments, and a fitting signal corresponding to each target signal segment is generated.

[0109] Continue to refer to Figure 1 In step 103, the signal segments in the pulse wave signal set are spliced ​​into signal segments with certain signal characteristics to obtain a target periodic signal set.

[0110] Specifically, a plurality of signal segments in the pulse wave signal set that are adjacent in time sequence and have similar or related pulse wave signal characteristics are sequentially spliced ​​into a signal segment with certain signal characteristics.

[0111] For example, n1 signal segments that are adjacent in time sequence in the pulse wave signal set are spliced ​​together to form a new signal segment, the new signal segment has similar or related pulse wave signal features, the new signal segment represents m1 pulse wave cycles, and these spliced ​​signal segments representing m1 pulse wave cycles are recorded in the target periodic signal set. Repeat the above process until all signal segments in the pulse wave signal set are processed, and these processed signal segments are recorded in the target periodic signal set.

[0112] The signal characteristics mainly refer to the signal characteristics of the original pulse wave signal or the fitted signal within a certain window function. The signal characteristics may include, but are not limited to: 1. obvious monotonicity or periodicity, 2. similarity or correlation between adjacent signal segments, 3. slow change between adjacent signal segments, 4. certain repeatability, etc.

[0113] The signal segments in the pulse wave signal set may be spliced ​​using, but not limited to, any of the following methods.

[0114] (1) Slope characteristic method

[0115] Calculate the slope value of each signal segment in the pulse wave signal set; splice n1 adjacent signal segments with the same slope together to merge into a new signal segment, which represents the rising or falling segment of m1 complete pulse wave cycles; and record the new signal segment in the target cycle signal set.

[0116] For example, the slope values ​​of two adjacent signal segments in the pulse wave signal set are calculated, and the two slope values ​​are multiplied. If the product of the slope values ​​of the current signal segment and the next adjacent signal segment is greater than or equal to 0, the two signal segments are merged into the current signal segment.

[0117] Repeat the above process to calculate the slope values ​​of the current signal segment and the next signal segment, and multiply the two slope values ​​until the product of the slopes of two adjacent segments is less than 0. At this time, the combined current signal segment is recorded in the target periodic signal set.

[0118] The above process is repeated until all signal segments in the pulse wave signal set are processed and recorded in the target periodic signal set, and all signal segments in the target periodic signal set have similar signal characteristics.

[0119] (2) Peak period characteristic method

[0120] Calculate the peak-to-peak value (the amplitude difference between the maximum positive peak and the maximum negative peak of the signal waveform in a pulse wave cycle, that is, the total amplitude difference between the peak and valley of the waveform) and the distance between peaks (the distance difference between the maximum positive peak and the maximum negative peak of the signal waveform in a pulse wave cycle) of each signal segment in the pulse wave signal set; splice n1 signal segments with adjacent signal segments and similar peak-to-peak values ​​together to form a new signal segment, and this new signal segment represents several complete pulse wave cycles; or splice n1 signal segments with adjacent signal segments and similar distance between peak-to-peak values ​​together to form a new signal segment, and this new signal segment represents m1 complete pulse wave cycles. Then, record the above new signal segment in the target period signal set.

[0121] Continue to refer to Figure 1 In step 104, the signal segments in the target periodic signal set are detected, and the heart rhythm signal features and the signal features between heart rhythm beats are extracted as secondary signal features.

[0122] The secondary signal features are mainly for signal segments in the target periodic signal set or signal features after the signal segments are combined. The heart rhythm refers to the rhythm and regularity of the heartbeat, which can be one heartbeat cycle or multiple heartbeat cycles. When extracting the secondary signal features, it is necessary to first identify each heartbeat beat of the first spliced ​​signal segments, and then extract the secondary signal features.

[0123] It should be noted that the secondary signal features are similar to the signal features in step 103, and may also include but are not limited to: 1. Obvious monotonicity or periodicity, 2. Similarity or correlation between adjacent signal segments, 3. Slow variation between adjacent signal segments, and 4. Certain repeatability. However, each of them has the following characteristics:

[0124] The signal features described in the above step 103 are used to perform the first splicing of the original signal decomposed by the time window function to obtain a signal segment with monotonic characteristics or certain periodic characteristics. The signal features include: obvious periodicity (such as the time repeatability of typical feature points of the beat cycle that conforms to the physiological characteristics of the human body, the periodic time-frequency domain characteristics, etc.), strong similarity or correlation between cycles (such as the signal between cycles has a small Euclidean distance, a large correlation coefficient, etc.), and slower signal feature variability between cycles (such as the beat change rate that conforms to the physiological characteristics of the human body, the possible amplitude change rate of a normal pulse wave, the rising edge slope change rate, the peak-to-peak value change rate, etc.), etc.

[0125] The signal features described in step 103 have the following characteristics: 1. Since a fitting function is constructed, the characteristics of the fitting function itself limit the many characteristics of the pulse wave signal that can be fitted with it. Therefore, the signal features in step 103 pay more attention to the similarity and consistency between the fitting signal and the pulse wave signal, whether the signal itself has periodicity, and whether different signal segments have similarities and slow changes. 2. The judgment rule does not emphasize the characteristics related to the physiological characteristics of the heart rhythm, but only emphasizes the characteristics of the signal itself.

[0126] The secondary signal features described in the above step 104: 1. First, the rhythm of the pulse wave is identified, and one or more heartbeat cycles are used as segments for extracting features; 2. The judgment rule emphasizes whether it is similar to a reasonable physiological signal.

[0127] A normal pulse wave, where each beat corresponds to a heart beat, should have the following characteristics:

[0128] 1. The beat has obvious periodicity (the periodicity emphasized here includes the periodic characteristics between each beat and the periodic characteristics between n beats and n beats);

[0129] 2. Strong similarity and correlation between beat cycles (single beats are similar, and n beats are also similar to n beats);

[0130] 3. The relatively slow variation between beat cycles (the amplitude, length, time domain characteristic value, frequency domain characteristic value, etc. between each single beat, or between n beats, should be within a certain range. If it exceeds the limit, it is considered that at least one of the two pulse wave periods is an abnormal pulse wave);

[0131] 4. The beat has a certain continuity (a high-quality, stable pulse wave signal should have a certain period length).

[0132] The signal segment periodicity calculation may be performed by, but is not limited to, any of the following methods:

[0133] (1) Time domain analysis method

[0134] For example, calculate whether the signal segments with time domain characteristics such as peaks, valleys, and the same slope in the time domain waveform are repeated; calculate the autocorrelation function of the signal to observe whether it has significant peaks under certain specific delays.

[0135] (2) Frequency domain analysis method

[0136] For example, by performing a Fourier transform on a signal and performing a discrete Fourier transform (DFT) on the signal, a periodic signal will usually exhibit discrete frequency components in the spectrum.

[0137] (3) Statistical methods

[0138] For example, we can calculate the periodicity index of the signal, such as the periodic intensity (PPI) or the periodic factor (PF); and analyze the energy distribution of the signal at different frequencies. Periodic signals usually have a higher energy concentration at a specific frequency.

[0139] The similarity or correlation calculation between two adjacent signal segments may be performed by, but is not limited to, any of the following methods:

[0140] (1) Distance measurement method

[0141] For example, the Euclidean distance, Manhattan distance, etc. between two adjacent signal segments are calculated to evaluate the similarity or correlation between the two adjacent signal segments.

[0142] (2) Statistical feature comparison method

[0143] For example, the statistical characteristics of two adjacent signal segments, such as mean, variance, cycle length, time length, slope, amplitude, kurtosis, skewness, etc., are compared to evaluate the similarity or correlation between the two adjacent signal segments.

[0144] (3) Correlation analysis method

[0145] For example, the Pearson correlation coefficient, cross-correlation function, etc. of two adjacent signal segments are calculated to evaluate the similarity or correlation between the two adjacent signal segments.

[0146] (4) Time-frequency analysis method

[0147] For example, two adjacent signal segments are decomposed into wavelet components of different frequencies through Fourier transform or wavelet transform, and then the energy or characteristics of these wavelet components are compared to evaluate the similarity or correlation between the two adjacent signal segments. Alternatively, the frequency domain characteristics and time domain characteristics of two adjacent signal segments are calculated and compared to evaluate the similarity or correlation between the two adjacent signal segments.

[0148] (5) Machine Learning Methods

[0149] Machine learning techniques such as support vector machines and neural networks are used for pattern recognition and classification to evaluate the similarity or correlation between two adjacent signal segments.

[0150] The calculation of the relatively slow variability between cycles can be performed by, but is not limited to, any of the following methods:

[0151] 1. Time domain analysis

[0152] For example, calculating the fluctuation amplitude of the signal and calculating the difference between the maximum and minimum values ​​of the signal based on the fluctuation amplitude can reflect the degree of change of the signal. Alternatively, calculating the standard deviation of the signal can measure the variability of the signal amplitude.

[0153] 2. Autocorrelation analysis

[0154] For example, by calculating the autocorrelation function, we can observe the similarity of the signal under different time delays. If the autocorrelation function drops rapidly, it means that the signal is more variable; otherwise, the variability is weaker.

[0155] 3. Frequency domain analysis

[0156] For example, we can perform Fourier transform on a signal and observe the energy distribution of the signal in the frequency domain. The wider the spectrum is, the more variable the signal is, because it contains more frequency components. Alternatively, we can more accurately evaluate the frequency domain variability of the signal by calculating the power spectral density (PSD).

[0157] 4. Statistical properties

[0158] For example, calculating the entropy of a signal can measure its uncertainty. The higher the entropy value, the greater the variability and complexity of the signal. Alternatively, calculating the coefficient of variation of a signal (the ratio of the standard deviation to the mean) can help assess the relative variability of the signal and is suitable for comparing signals of different dimensions.

[0159] The following uses the statistical feature comparison method as an example to illustrate the comparison of the cycle length, time length, amplitude, slope and other signal features of two adjacent signal segments in an embodiment of the present invention. The inherent characteristics of the pulse wave and the correlation conditions are set: the cycle length confidence interval [p0, p1], the pulse wave time length confidence interval [t0, t1], the pulse wave amplitude change rate confidence interval [k0, k1], the pulse wave cycle length change law confidence interval [q0, q1], and the pulse wave slope change rate confidence interval [s0, s1].

[0160] Set the continuous condition of the number of pulse wave signal segments: the confidence value r of the number of continuous wave segments.

[0161] Furthermore, adjacent signal segments P0 and P1 are selected from the target periodic signal set, and the period lengths L of the two signal segments are calculated. p0 and L p1 (or time length T p0 and T p1 ), calculate the rising amplitude A of the pulse wave cycle of these two signal segments u0 and A u1 , and the decrease rate A d0 and A d1 , calculate the pulse wave signal slope S of these two signal segments p0 and S p1 If L p0 and L p1 The pulse wave cycle length confidence interval [p0, p1] range is met, and the pulse wave cycle rise amplitude A u0 and A u1 , and the decrease rate A d0 and A d1 The ratio satisfies the pulse wave amplitude change rate confidence interval [k0, k1] range, and the pulse wave signal slope is S p0 and S p1 .

[0162] Continue to refer to Figure 1 In step 105, it is determined whether these signal segments in the target periodic signal set meet the decision conditions according to the secondary signal characteristics of the signal segments; if the decision conditions are met, these signal segments are spliced ​​and output as normal pulse wave signals; if the decision conditions are not met, these signal segments are regarded as motion artifacts or noise.

[0163] Specifically, it is determined based on the secondary signal characteristics whether the self-signal characteristics of two adjacent signal segments in the target periodic signal set, the similarity and correlation between the two signal segments, the variability between the signal segments, and the continuity of the signal segments meet the decision conditions. If the decision conditions are met, these adjacent signal segments are spliced ​​together to form a complete signal output. If the decision conditions are not met, these signals are regarded as motion artifacts or interference signals.

[0164] The above process is repeated until all signal segments in the target periodic signal set are processed, and the obtained spliced ​​signal is the final processed pulse wave signal with signal interference and motion artifacts removed.

[0165] For example, n1 signal segments that are adjacent in time sequence in the pulse wave signal set are spliced ​​together to form a new signal segment, the new signal segment has similar or related pulse wave signal features, the new signal segment represents m1 pulse wave cycles, and these spliced ​​signal segments representing m1 pulse wave cycles are recorded in the target periodic signal set. Repeat the above process until all signal segments in the pulse wave signal set are processed, and these processed signal segments are recorded in the target periodic signal set.

[0166] In order to better improve the accuracy of the system's recognition of pulse waves, increase the probability of the system recognizing the correct signal fragments, and reduce the probability of the system missing the correct signal fragments. Enhance the system's robustness, so that the recognition of signal fragments is not easily affected by individual signal feature errors or uncertainties, and improve the system's adaptability to abnormal situations. After calculating the periodic characteristics, inter-cycle similarity and correlation characteristics, inter-cycle variability characteristics, and periodic continuity characteristics of the signal fragments, the decision fusion method is used to fuse the above feature results for analysis, and finally output reasonable and reliable results.

[0167] When judging whether the signal segments in the target periodic signal set meet the decision conditions, a decision fusion method may be adopted, such as but not limited to any of the following methods:

[0168] 1. Voting Law

[0169] For example, a simple voting method is used to determine whether all signal features meet the preset conditions. If all or most of the signal segments meet the preset conditions, it is determined that the decision rule is met. Alternatively, a weighted voting method is used to assign different weights according to different signal features. If the signal segments meet the preset conditions after weighting, it is determined that these signal segments meet the decision rule.

[0170] 2. Stacking method

[0171] For example, the feature values ​​of multiple signal segments are used to train and generate a higher-level judgment condition. If the signal segments meet the judgment condition, it is judged that these signal segments meet the decision rule.

[0172] 3. Ensemble Learning

[0173] For example, using the bagging method, multiple models are trained by resampling the signal segment set through random forests, and their prediction results are combined to generate a higher-level judgment rule. If the signal segments meet the preset conditions, then these signal segments are judged to meet the decision rule. Alternatively, using the boosting method, multiple models are trained by gradually adjusting the sample weights to generate a higher-level judgment rule. If the signal segments meet the preset conditions, then these signal segments are judged to meet the decision rule.

[0174] 4. Bayesian Fusion

[0175] The Bayesian theorem is used to combine the judgment results of different signal segment feature values ​​to generate a higher-level judgment rule. If the signal segments meet the preset conditions, then these signal segments are judged to meet the decision rule.

[0176] These methods can be selected and combined according to specific pulse wave signal attributes and features to achieve more effective decision fusion.

[0177] The following takes a simple voting method as an example to illustrate how, in an example of the present invention, decision fusion is performed based on the secondary signal features of the target signal segment set to ultimately output a pulse wave signal with signal interference and motion artifacts removed.

[0178] Set the following preset conditions to determine whether a signal segment and its adjacent segments in the target signal set satisfy the following relationship:

[0179]

[0180] or

[0181]

[0182] If the above preset conditions are met, the two signal segments P0 and P1 meet the inherent signal characteristics of the pulse wave. The two signal segments are spliced ​​to form a pulse wave period signal S0. Otherwise, the two signal segments P0 and P1 are recorded as noise.

[0183] Repeat the above steps to obtain the combined pulse wave period signal S0...S n .

[0184] Further, the time length L of the pulse wave period signals S0 and S1 is calculated s0 and Ls1 , and pulse wave amplitude A s0 and A s1 If the time length L s0 and L s1 The ratio satisfies the confidence interval [q0,q1] range of the pulse wave cycle length variation law, and the pulse wave amplitude A s0 and A s1 The ratio satisfies the confidence interval of the pulse wave cycle length variation law [s0,s1], and its calculation relationship is as follows:

[0185]

[0186] If the above conditions are met, the two combined pulse waves meet the correlation condition, and the periodic signals S0 and S1 are recorded as normal pulse wave signals. Otherwise, the periodic signals S0 and S1 are recorded as noise.

[0187] Repeat the above steps. If conditions (3) or (4) are met, and the number of continuous normal pulse wave cycles in (5) is greater than or equal to the confidence value r of the number of continuous segments (for example, r = 10, that is, a pulse wave with more than 10 heart beats is considered to be a stable and reliable segment), then all of this pulse wave cycle will be counted as a normal pulse wave signal. Otherwise, this pulse wave cycle will be discarded.

[0188] It should be noted that the simple voting decision rule (3) or (4) or (5) given in this embodiment is a set of priori set conditions given according to the characteristics of the pulse wave signal. The above set conditions are only an example listed in the present invention. The present invention cannot object to enumerating other possible set conditions. Other conditions set by those skilled in the art are also within the protection scope of the present invention.

[0189] In the embodiment of the present invention, the original pulse wave is decomposed into different signal segments according to a sliding time window function (the sliding time windows may partially overlap). Since the signal segments may contain interference, motion artifacts, noise and other signals that are inconvenient to process, the segmented signal segments are spliced ​​into signal segments with certain signal characteristics, and then the pulse wave signal characteristics, the signal characteristics between adjacent signals, the rate of change between adjacent signals and the continuity of the signal segments are calculated, and the decision conditions are set, and the signal segments that meet the decision rules are spliced ​​into a complete signal, so that the signal interference and motion artifacts in the pulse wave signal can be adaptively filtered out, and the extraction of high-quality pulse wave signals in complex interference states and large motion states is achieved.

[0190] In addition, the resolution of the signal fragments can be reduced through fitting technology, and the signal fragments can be blurred into fitting signals. The fitting signal highlights the macroscopic features and removes the microscopic features, which is not only easier to analyze and process, but also makes the final decision more accurate.

[0191] Accordingly, the present invention also provides a pulse wave signal processing device, such as Figure 3 The figure is a schematic diagram of the structure of the pulse wave signal processing device provided by the present invention.

[0192] The pulse wave signal processing device 300 includes the following modules:

[0193] The signal acquisition module 301 is used to acquire the original pulse wave signal;

[0194] The segmentation module 302 is used to intercept the original pulse wave signal using a sliding time window function, construct a fitting function, adjust the sliding time window function parameters and / or fitting function parameters, so that the fitting signal corresponding to each intercepted signal segment approaches the signal segment; and generate a pulse wave signal set according to the signal segment or the corresponding fitting signal;

[0195] A splicing module 303 is used to splice the signal segments in the pulse wave signal set into signal segments with certain signal characteristics to obtain a target periodic signal set;

[0196] The signal detection module 304 is used to detect the signal segments in the target periodic signal set, extract the heart rhythm signal features and the signal features between heart rhythm beats as secondary signal features; judge whether these signal segments meet the decision conditions according to the secondary signal features of the signal segments in the target periodic signal set; if the decision conditions are met, splice these signal segments and output them as normal pulse wave signals; if the decision conditions are not met, regard these signal segments as motion artifacts or noise.

[0197] In another non-limiting embodiment, the pulse wave signal processing device 300 may further include: a pre-processing module (not shown) for performing noise reduction and filtering processing on the original pulse wave signal.

[0198] The specific implementation of each of the above modules can be found in the description of the above method embodiments of the present invention, which will not be repeated here.

[0199] Accordingly, the present invention also provides a pulse wave signal processing system, such as Figure 4 The figure shows a structural diagram of the system.

[0200] The pulse wave signal processing system comprises: a signal acquisition device 400 and the pulse wave signal processing device 300 . The pulse wave signal processing device 300 in this embodiment further comprises a storage module 30 .

[0201] The signal acquisition device 400 is used for collecting human pulse wave data, and transmits the collected pulse wave signal of a certain time length to the pulse wave signal processing device 300 .

[0202] The signal acquisition device 400 may include a variety of sensors, such as, but not limited to, photoelectric sensors, pressure sensors, ultrasonic sensors, etc., which can be used for pulse wave signal acquisition, and are used to acquire the original pulse wave signal of the subject, and the acquisition locations include, but are not limited to, wrists, fingertips, finger bases, arms, ankles, necks, ears, etc., where pulse wave signals can be acquired. The signal acquisition device 400 can acquire pulse wave signals of a certain length of time, and then send them to the pulse wave signal processing device 300. After receiving the signals, the signal acquisition module 301 in the pulse wave signal processing device 300 stores them in the storage module 30 as an original pulse wave signal.

[0203] In some embodiments, the signal acquisition device 400 may further perform noise reduction and filtering on the collected pulse wave signal before sending it to the pulse wave signal processing device 300. For example, a corresponding filtering module is provided in the signal acquisition device 400, and the filtering module may adopt a Butterworth bandpass filter, with a passband frequency of 0.5 Hz to 20 Hz and a filter order of 4.

[0204] In other embodiments, the preprocessing module in the pulse wave signal processing device 300 may first perform noise reduction and filtering on the received original pulse wave signal, and then perform subsequent processing on the filtered pulse wave signal, which is not limited in this embodiment of the present invention.

[0205] When the pulse wave signals in the storage module 30 reach a certain number, the segmentation module 302, the splicing module 303 and the signal detection module 304 perform corresponding calculations and processing on them to remove motion artifacts and noise and output high-quality pulse wave signals.

[0206] The following example illustrates the process of processing the received pulse wave signal.

[0207] A time window function with a length of w is preset. The segmentation module 302 extracts a pulse wave signal from the signal acquisition module 301, and cuts a signal with a length of w from it as the target signal y i :

[0208]

[0209] The fitting function is constructed using the following quadratic polynomial as the basis function:

[0210] f(x j )=a2x j 2 +a1x j +a0,j∈[0,w] (7)

[0211] The time window function selects the rectangular window function, and the initial value of the length w is set to 0.25s.

[0212] Calculate the fitting function that meets the conditions. For example, use the least squares method to transform f(x) to the objective function Perform fitting to obtain a set of basis function parameters a0, a1, a2, and calculate the fitting function f(x) and the objective function The sum of squares of the errors between ε and ε is calculated. The maximum allowable error ε is preset (for example, ε is 8), and the time window length w is cyclically updated. When the calculated sum of squares of the errors is less than or equal to ε, the cyclic update ends. As shown in the following formula:

[0213]

[0214] At this time, record the pulse wave signal corresponding to this time window length w To pulse wave signal collection Or record the calculated fitting signal f1(x) to the pulse wave signal set

[0215] Repeat the above steps until all the pulse wave data has been processed. The resulting pulse wave signal set is as follows:

[0216]

[0217] The segmentation module 302 stores the above calculation results in the storage module 30 .

[0218] In the following processing, for the pulse wave signal set The intercepted signal fragment or The processing of the fitting signal segments in is the same. Therefore, the following is a pulse wave signal set The following is an example of the intercepted signal fragment in FIG.

[0219] Then, the splicing module 303 extracts the pulse wave signal set stored in the storage module 30 Calculate the features of the signal segments and compare them with the set reasonable feature range conditions. If the features of the periodic signal segments meet the set conditions, these periodic signals are concatenated into one periodic signal and recorded in the target periodic signal set.

[0220] Repeat the above steps until the pulse wave signal is collected. All signal segments in are processed completely.

[0221] Specifically, the target signal Find the derivative, calculate the monotonicity of the signal, and splice together adjacent signal segments with the same monotonic property to form a new signal segment. That is, splice together adjacent signal segments with the same increasing monotonic property, and splice together adjacent signal segments with the same decreasing monotonic property.

[0222] After completing the above splicing, if the monotonic properties of two adjacent signal segments are opposite, then continue to splice these two signal segments with opposite monotonic properties together to form a periodic signal with complete fluctuations.

[0223] Furthermore, the confidence interval of the change rate of the length of adjacent signal segments is set to [l0,l1], the confidence interval of the change rate of the amplitude of adjacent signal segments is set to [k0,k1], and the confidence interval of the change rate of the slope of adjacent signal segments is set to [s0,s1]. In , select the current signal segment and the next adjacent signal segment. If the length change rate of the above two signal segments satisfies the range [l0,l1], the amplitude change rate of the above two adjacent signal segments satisfies the range [k0,k1], and the slope change rate of the above two adjacent signal segments satisfies the range [s0,s1], then these periodic signal segments are recorded as a reasonable period, otherwise the current signal segment is recorded as noise.

[0224] Repeat the above steps and record the spliced ​​periodic signal fragments into the target periodic signal set

[0225] The values ​​of the above parameters are, for example: [l0,l1] is [0.5,1.5], [k0,k1] is [0.5,1.5], and [s0,s1] is [-1.4,-0.6].

[0226] The splicing module 303 stores the splicing result in the storage module 30 .

[0227] Then, the signal detection module 304 extracts the target periodic signal set stored in the storage module 30 The self-features, related features, and continuous features of these periodic signal segments are calculated, and these features are compared with the set reasonable feature range conditions. If the features of the periodic signal segments meet the set conditions, these periodic signal segments are counted as reasonable pulse waves and spliced ​​into a continuous pulse wave. Otherwise, this periodic signal or several periodic signals are counted as noise.

[0228] Specifically, the confidence interval of the periodic signal length is set [p0, p1]. If the length of one or several periodic signals meets the range [p0, p1], then this periodic signal period is counted as a reasonable pulse wave period, or several periodic signals are spliced ​​together to be counted as a reasonable pulse wave period. Otherwise, this periodic signal or several periodic signals are recorded as noise.

[0229] Furthermore, the confidence interval of the change rate of the length of adjacent pulse wave cycles is set to [q0, q1], and the confidence interval of the change rate of the amplitude is set to [k0, k1]. The current pulse wave cycle signal and the next adjacent pulse wave cycle signal are selected. If the length change rate of the above two signals satisfies the range of [q0, q1], and if the amplitude change rate of the above two signals satisfies the range of [k0, k1], then these two pulse wave cycle signals are counted as reasonable pulse wave cycles. Otherwise, the current pulse wave cycle signal is recorded as noise.

[0230] Furthermore, the minimum value of the correlation coefficient of adjacent periodic signals is set to ∈, and the correlation coefficient ρ between the current periodic signal and the next adjacent periodic signal is calculated. The calculation method is as follows:

[0231]

[0232] If the correlation coefficient of the above two periodic signals is ρ≥∈, the two pulse wave periodic signals are recorded as reasonable pulse wave periods. Otherwise, the current pulse wave periodic signal is recorded as noise.

[0233] Specifically in this embodiment, under normal circumstances, the heart rate interval of a normal person is 40 beats / minute to 220 beats / minute, so the confidence interval of the periodic signal length [p0, p1] is set to [fs·60 / 220, fs·60 / 40], the confidence interval of the change rate of adjacent pulse wave period lengths [q0, q1] is set to [0.5, 1.5], the confidence interval of the change rate of adjacent pulse wave amplitudes [k0, k1] is set to [0.4, 1.6], and the minimum value ∈ of the correlation coefficient of two periodic signals is set to 0.7.

[0234] Repeat the above steps and record the spliced ​​reasonable pulse wave target signals into the set

[0235] For collections Calculate the number of reasonable pulse wave cycles that appear in a continuous time. Set the minimum value of continuous cycles r (for example, r is 5). If the number of reasonable pulse wave cycles that appear in a continuous time is greater than or equal to r, splice these continuous pulse wave cycles into a continuous pulse wave signal. Repeat the above steps to finally generate several continuous pulse wave signals as the final result output of removing motion artifacts and interference.

[0236] Through the above process, the pulse wave signal processing device 300 can adaptively remove pulse wave signal interference and motion artifacts.

[0237] It should be noted that, in a specific application, the segmentation module 302 , the splicing module 303 and the signal detection module 304 may be independent modules or integrated into one module, which is not limited in the embodiment of the present invention.

[0238] Compared with the prior art, the pulse wave signal processing method, device and system provided by the present invention have the following advantages:

[0239] (1) The motion artifacts and noise in the pulse wave signal are removed without the help of an accelerometer or other sensors, which overcomes the reliance of the motion artifact removal process on the accelerometer, reduces the complexity of the data in the denoising process, makes the implementation process more real-time, and improves the efficiency of removing the pulse wave motion artifact.

[0240] (2) The motion artifacts and noise in the pulse wave signal are removed without the help of multi-channel pulse wave acquisition, which overcomes the dependence of the motion artifact removal process on multi-channel pulse wave sampling, greatly reduces the amount of data collection, and can reduce the implementation cost and computing cost of hardware equipment, making the implementation process more real-time and improving the efficiency of removing pulse wave motion artifacts.

[0241] (3) The motion artifacts and noise in the pulse wave signal are removed without the help of pre-collected signals for model training, thus overcoming the dependence on pre-collected data and the defect that the pre-trained model is unusable due to the different parameters and signal detail characteristics brought by different devices and sensors collecting pulse wave signals. Simple parameters are set only according to the basic characteristics of the pulse wave signal itself, and the data characteristics of the collected pulse wave signal itself are used to identify the effective pulse wave cycle, and the correlation between cycles is used to identify and remove interference signals and motion artifacts, and retain the normal pulse wave signal. The adaptability of the method and system is greatly improved, and the implementation cost and computing cost are reduced.

[0242] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

[0243] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0244] In several embodiments provided by the present invention, it should be understood that the disclosed device can be implemented in other ways.

[0245] The present invention also provides a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and the computer program can be executed when it is run. Figure 1 or Figure 2 The storage medium may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc. The storage medium may also include a non-volatile memory (non-volatile) or a non-transitory memory, etc.

[0246] Accordingly, the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor executes Figure 1 or Figure 2 All or part of the steps in the pulse wave signal processing method.

[0247] It should be understood that the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article indicates that the associated objects before and after are in an "or" relationship.

[0248] The "plurality" appearing in the embodiments of the present application refers to two or more.

[0249] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless means.

[0250] The embodiments of the present invention are described in detail above. The present invention is described in detail using specific implementation methods herein. The description of the above embodiments is only used to help understand the method and system of the present invention. It is only a part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should belong to the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A pulse wave signal processing method, characterized in that: The method comprises: Acquire and store original pulse wave signals; The original pulse wave signal is intercepted by using a sliding time window function, and a fitting function is constructed, and the sliding time window function parameters and / or fitting function parameters are adjusted so that the fitting signals corresponding to the intercepted signal segments are close to the signal segments, and a fitting signal segment corresponding to the signal segment is obtained; a pulse wave signal set is generated according to the signal segment or the corresponding fitting signal segment; splicing the signal segments in the pulse wave signal set into signal segments with certain signal characteristics to obtain a target periodic signal set; Detecting the signal segments in the target periodic signal set, extracting the heart rhythm signal features and the signal features between heart rhythm beats as secondary signal features; According to the secondary signal characteristics of the signal segments in the target periodic signal set, it is judged whether these signal segments meet the decision conditions; if the decision conditions are met, these signal segments are spliced ​​and output as normal pulse wave signals; if the decision conditions are not met, these signal segments are regarded as motion artifacts or noise.

2. The pulse wave signal processing method according to claim 1, characterized in that: Generating a pulse wave signal set according to the signal segments comprises: The signal segment is recorded into the pulse wave signal set.

3. The pulse wave signal processing method according to claim 1, characterized in that: Generating a pulse wave signal set according to the signal segments comprises: generating a fitting signal corresponding to each signal segment; The fitting signal is recorded in the pulse wave signal set.

4. The pulse wave signal processing method according to claim 3, characterized in that: The fitting function is a monotonic function or a periodic function.

5. The pulse wave signal processing method according to claim 3, characterized in that: The step of constructing a fitting function and adjusting the sliding time window function parameters and / or fitting function parameters so that the fitting signal corresponding to each intercepted signal segment approaches the signal segment includes: Constructing a fitting function, and using the fitting function to generate a fitting signal corresponding to the signal segment; Calculating signal characteristics of the fitting signal and corresponding signal segments; If the signal characteristics of the fitting signal and the corresponding signal segment meet the set conditions, the fitting signal is used as the fitting signal segment corresponding to the signal segment; Otherwise, the time window function parameters and / or fitting function parameters are adjusted until the signal characteristics of the fitting signal and the corresponding signal segment meet the set conditions.

6. The pulse wave signal processing method according to claim 1, characterized in that: The step of splicing the signal segments in the pulse wave signal set into signal segments having certain signal characteristics comprises: Multiple signal segments in the pulse wave signal set that are adjacent in time sequence and have signal characteristics that meet set conditions are sequentially spliced ​​into a signal segment with certain signal characteristics.

7. The pulse wave signal processing method according to claim 1, characterized in that: The detecting, according to the secondary signal features of the signal segments in the target periodic signal set, whether the signal segments meet the decision condition comprises: According to the secondary signal characteristics, it is determined whether the self signal characteristics of two adjacent signal segments in the target periodic signal set, the similarity and correlation between the two signal segments, the variability between the signal segments, and the continuity of the signal segments meet the decision conditions.

8. The pulse wave signal processing method according to claim 7, characterized in that: Use any of the following methods to determine whether these signal fragments meet the decision conditions: voting method, stacking method, ensemble learning method, Bayesian fusion method.

9. The pulse wave signal processing method according to any one of claims 1 to 8, characterized in that: The signal characteristics include any one or more of the following: obvious monotonicity or periodicity, strong similarity or correlation between cycles, relatively slow variation between cycles, and certain continuity.

10. The pulse wave signal processing method according to claim 9, characterized in that: Use any of the following methods to calculate the monotonicity or periodicity of the signal: time domain analysis method, frequency domain analysis method, statistical method.

11. The pulse wave signal processing method according to claim 9, characterized in that: The similarity or correlation between signals is calculated using any of the following methods: distance measurement method, statistical feature comparison method, correlation analysis method, time-frequency analysis method, machine learning or deep learning method.

12. The pulse wave signal processing method according to claim 9, characterized in that: The variability between signals is calculated using any of the following methods: time domain analysis method, autocorrelation or cross-correlation analysis method, frequency domain analysis method, and statistical method.

13. The pulse wave signal processing method according to claim 1, characterized in that: The method further comprises: Before intercepting the original pulse wave signal, the original pulse wave signal is subjected to noise reduction and filtering processing.

14. The pulse wave signal processing method according to claim 13, characterized in that: The filtering process of the original pulse wave signal comprises: The original pulse wave signal is subjected to denoising and filtering by any of the following filtering methods: low-pass filtering, high-pass filtering, band-pass filtering, median filtering, sliding window filtering, wavelet transform, Kalman filtering, empirical mode decomposition (EMD), variational mode decomposition (VMD), and local mean decomposition (LMD).

15. A pulse wave signal processing device, characterized in that: The device comprises: A signal acquisition module, used for acquiring and storing the original pulse wave signal; A segmentation module is used to intercept the original pulse wave signal using a sliding time window function, construct a fitting function, adjust the sliding time window function parameters and / or fitting function parameters, so that the fitting signal corresponding to each intercepted signal segment approaches the signal segment; and generate a pulse wave signal set according to the signal segment or the corresponding fitting signal; A splicing module, used for splicing the signal segments in the pulse wave signal set into signal segments with certain signal characteristics to obtain a target periodic signal set; A signal detection module is used to detect the signal segments in the target periodic signal set, extract the heart rhythm signal features and the signal features between heart rhythm beats as secondary signal features; judge whether these signal segments meet the decision conditions according to the secondary signal features of the signal segments in the target periodic signal set; if the decision conditions are met, splice these signal segments and output them as normal pulse wave signals; if the decision conditions are not met, regard these signal segments as motion artifacts or noise.

16. The pulse wave signal processing device according to claim 15, characterized in that: The device also includes: The preprocessing module is used to perform noise reduction and filtering on the original pulse wave signal before intercepting the original pulse wave signal.

17. A pulse wave signal processing system, characterized in that: The system comprises: a signal acquisition device, and a pulse wave signal processing device as claimed in claim 15 or 16; The acquisition device is used to acquire human pulse wave signals and transmit the acquired pulse wave signals of a certain time length to the pulse wave signal processing device; The pulse wave signal processing device is used to process the human body pulse wave signal and output a normal pulse wave signal.

18. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the pulse wave signal processing method according to any one of claims 1 to 14 are executed.

19. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor runs the computer program, the processor performs the steps of the pulse wave signal processing method according to any one of claims 1 to 14.

Images