PPG signal denoising and feature extraction method and device

Abstract

The present invention discloses a method and device for denoising and extracting features of a PPG signal, wherein the method includes: obtaining PPG data; preprocessing the PPG signal to remove noise such as baseline drift in the PPG signal; determining the optimal minimum interval to obtain a more accurate heart rate and peak point location; removing noise in the signal based on kurtosis denoising; locating the PPG signal dicrotic wave and calculating morphological features; and calculating the PPI feature based on the peak point information. The present invention uses a method based on variation index threshold and kurtosis for denoising, an adaptive method for locating the peak, and a morphological classification method for locating the dicrotic wave, and can obtain accurate and detailed PPG signal features.

Description

Technical Field

[0001] The present invention relates to PPG signal processing technology, and in particular to a PPG signal denoising and feature extraction method and device. Background Art

[0002] Photoplethysmography (PPG) uses an LED light source to emit light to the skin, and a photodiode absorbs the reflected light. With each heartbeat, the contraction and expansion of blood vessels will affect the transmission of light, so blood flow information can be obtained through the PPG waveform. The peak detection of PPG is generally detected by detecting the local maximum value. The peak detection method of PPG is greatly affected by noise, and it is necessary to judge and remove noise from the data itself before peak detection.

[0003] At present, feature extraction based on heart rate interval PPI is mainly achieved by analyzing the pulse rate variability (PRV, an alternative indicator of heart rate variability) of the PPG signal. Since PRV has certain requirements for the time window, it is difficult to achieve real-time monitoring of the indicator. In addition, the algorithm currently under study is not accurate in extracting the peak value of the pulse wave, which has an important impact on the extraction of PRV features.

[0004] At present, the morphological feature extraction of PPG is mainly carried out through time domain methods, using differential and differential methods to extract the features of PPG signals with special physiological significance due to time changes, such as contraction peak height, dicrotic wave height, length ratio between special points, area ratio under the curve and period, etc. Among these methods, it is very important to find the main wave peak and dicrotic wave. Current research includes: analyzing the first-order derivative and second-order derivative features of the pulse wave; using the threshold method to judge; using the wavelet transform algorithm to identify, etc. These methods are either unable to effectively locate certain hidden dicrotic waves, or are very sensitive to noise and have requirements on the time length of the PPG signal. Summary of the invention

[0005] Purpose of the invention: In view of the problems existing in the prior art, the present invention provides a method and device for denoising and extracting features of PPG signals with higher effectiveness, stronger applicability and higher accuracy.

[0006] Technical solution: The PPG signal denoising method of the present invention comprises:

[0007] S1, obtaining an original PPG signal, and performing preprocessing to obtain a second PPG signal;

[0008] S2. Locate the peak point of the second PPG signal to obtain a first peak point sequence and a first PPI sequence;

[0009] S3, calculating the variation index of the first PPI sequence, and determining the variation index judgment threshold according to the variation index of the first PPI sequence;

[0010] S4, calculating the variation index of each point in the first PPI sequence, and identifying and deleting the points exceeding the variation index judgment threshold as noise, to obtain a second PPI sequence and a second peak point sequence;

[0011] S5. Based on a preset minimum peak interval range, search the second PPG signal for an optimal minimum peak interval, and segment the second PPG signal according to the optimal minimum peak interval to obtain a third peak point sequence and a third PPI sequence;

[0012] S6. Calculate the kurtosis of the third PPI sequence to obtain a first kurtosis sequence;

[0013] S7. Calculate the change speed of the first kurtosis sequence, determine the data whose change speed exceeds a preset change speed threshold as abnormal noise and delete it, so as to obtain a denoised fourth peak point sequence and a third PPG signal, wherein the third PPG signal is the desired denoised signal.

[0014] Furthermore, step S3 specifically includes:

[0015] S301, calculating the variation index of the first PPI sequence according to the following formula:

[0016]

[0017] In the formula, VI represents the variation index, x represents the first PPI sequence, dif represents the first-order difference of sequence x, abs represents the absolute value of the value in the sequence, std represents the standard deviation of the sequence, and mean represents the mean value of the sequence;

[0018] S302: Determine the variation index judgment threshold according to the following formula based on the variation index of the first PPI sequence:

[0019]

[0020] Among them, VI threshold represents the variation index judgment threshold, VI max Represents the upper limit of the variation index, VI min Represents the lower limit of the variation index.

[0021] Furthermore, step S4 specifically includes:

[0022] S401, setting the calculation window and moving step size;

[0023] S402, moving the calculation window according to the moving step length, and calculating the variation index of the sequence segment located in the current calculation window in the first PPI sequence;

[0024] S403: for each point in the first PPI sequence, obtain the variation index of the sequence segments of all calculation windows containing the point, and use the mean of the obtained variation index as the variation index of the corresponding point;

[0025] S404, determining whether the variation index of each point in the first PPI sequence exceeds a determination threshold, wherein if the variation index of a sequence segment in a calculation window exceeds the determination threshold, the variation indexes of all points in the calculation window are determined to exceed the determination threshold;

[0026] S405: Delete points exceeding the judgment threshold in the first PPI sequence, and the deleted points form a second PPI sequence. The peak points remaining after the deletion form a second peak point sequence.

[0027] Furthermore, step S5 specifically includes:

[0028] S501, setting the minimum peak interval range to a first preset value and a second preset value times the average value of the second PPI sequence;

[0029] S502, setting the value of the minimum interval of the peak points to the minimum value of the minimum peak interval range;

[0030] S503, finding the peak value of the second PPG signal according to the minimum interval between peak points;

[0031] S504, determining a peak point interval sequence according to the peak value found in step S503, and calculating the variation degree of each peak point interval;

[0032] S505, determining whether the minimum value in the peak point interval sequence is equal to the maximum value of the minimum peak interval range, if not, executing step S506, if yes, executing step S507;

[0033] S506, adding 1 to the value of the minimum interval of the peak points, and executing step S503;

[0034] S507: taking the peak point interval with the smallest variation as the optimal minimum peak interval, and dividing the second PPG signal periodically by the optimal minimum peak interval to obtain a third peak point sequence, and taking the intervals between all peak points as a third PPI sequence.

[0035] Further, step S6 specifically includes:

[0036] S601. For the third PPI sequence, the window is moved according to a preset step size, and the kurtosis of the sequence segment within the window is calculated using the following formula:

[0037]

[0038] In the formula, kurtosisk It represents the kurtosis calculated when the window is moved for the kth time according to the preset step size. Refers to the PPI average value of the current window, x i Indicates the value of the i-th PPI point in the current window, and n is the size of the window;

[0039] S602, when the window moves to the end of the third PPI sequence, the calculation ends, and all kurtosis component sequences are calculated to form a first kurtosis sequence.

[0040] Furthermore, step S7 specifically includes:

[0041] S701, calculating the change speed of each data in the first kurtosis sequence;

[0042] S702: Delete data whose changing speed exceeds a preset changing speed threshold as noise, and obtain a fourth PPI sequence, a fourth peak point sequence, and a third PPG signal with kurtosis noise deleted.

[0043] The PPG signal feature extraction method of the present invention comprises:

[0044] S8, using the PPG signal denoising method to denoise the original PPG signal, and using the obtained fourth peak point sequence to segment the denoised third PPG signal, with each segment being a PPG pulse, thereby obtaining a PPG pulse sequence;

[0045] S9, classifying the PPG pulses according to the morphological characteristics and multiple order derivative characteristics of the PPG pulse sequence, and performing dicrotic wave positioning on the PPG pulses using a dicrotic wave positioning method corresponding to the PPG pulse classification;

[0046] S10. Calculate morphological parameter characteristics and PPI signal characteristics according to the systolic peak position and dicrotic wave position of the PPG pulse sequence.

[0047] Furthermore, step S9 specifically includes:

[0048] S901, invert each PPG pulse and locate the Jiangzhongxia, where the Jiangzhongxia is the trough closest to the midpoint of a signal segment;

[0049] S902, obtaining a PPG pulse from a PPG pulse sequence, determining whether the current PPG pulse is an abnormal signal that cannot locate a dicrotic wave, if determined to be an abnormal signal, discarding the pulse segment, and returning to step S902, if not an abnormal signal, executing step S903; wherein the determination method is: finding the first zero-crossing point after the third-order derivative signal of the PPG pulse falls in the middle gorge, if there is no zero-crossing point, determining it to be an abnormal signal, otherwise determining it to be a non-abnormal signal;

[0050] S903, determine whether the current PPG pulse is a first type of hidden dicrotic wave, if so, execute step S905, if not, execute step S904; wherein the determination method is: locate the first trough value of the fourth-order derivative of the PPG pulse after the gorge of the third-order derivative, if there is a trough value, it does not belong to the first type of hidden dicrotic wave, otherwise it is determined to belong to the first type of hidden dicrotic wave;

[0051] S904, determining whether the current PPG pulse is a second-type hidden dicrotic wave, if so, executing step S906, if not, executing step S907; wherein the determination method is: locating the first trough value of the fourth-order derivative of the PPG pulse after the gorge of the third-order derivative, if the trough value is greater than 0, determining it as a standard dicrotic wave or a non-hidden dicrotic wave, otherwise determining it as a second-type hidden dicrotic wave;

[0052] S905, locating the first type of hidden dicrotic wave, the locating method is: taking the middle value of the PPG pulse third-order derivative curve falling in the middle gorge and the first zero-crossing point position after the third-order derivative curve falls in the middle gorge as the position of the dicrotic wave;

[0053] S906, locating the second type of hidden dicrotic wave, the locating method is: taking the first peak of the third-order derivative curve of the PPG pulse after descending the middle gorge as the position of the dicrotic wave;

[0054] S907, locating the standard dicrotic wave or the non-hidden dicrotic wave, the method of locating is: taking the zero-crossing point of the PPG third-order derivative curve after the middle gorge as the position of the dicrotic wave.

[0055] The PPG denoising device of the present invention comprises:

[0056] one or more processors;

[0057] A memory for storing one or more programs;

[0058] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned denoising method.

[0059] The PPG feature extraction device of the present invention comprises:

[0060] one or more processors;

[0061] A memory for storing one or more programs;

[0062] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned feature extraction method.

[0063] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages: the denoising method of the present invention can effectively remove the noise of the PPG signal, providing a basis for subsequent feature extraction; the feature extraction method of the present invention can effectively locate various dicrotic waves, including some hidden dicrotic waves, and has no requirements on the time length of the PPG signal, and has higher calculation accuracy and stronger applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 is a flow chart of the PPG signal denoising method provided by the present invention;

[0065] Figure 2 It is a flow chart of the PPG signal feature extraction method provided by the present invention;

[0066] Figure 3 It is a waveform diagram of various dicrotic waves. DETAILED DESCRIPTION

[0067] In order to clearly illustrate the technical features, solutions and effects of the present invention, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0068] Embodiment 1

[0069] This embodiment provides a PPG signal denoising method. The execution subject of the method can be any portable device, including but not limited to a device with PPG acquisition capability. The terminal can be a terminal device such as a mobile phone, a tablet computer, a wearable device, or a device with data processing capability such as a laptop computer or a desktop computer. Figure 1 As shown, the denoising method of this embodiment includes:

[0070] S1. Obtain an original PPG signal and perform preprocessing to obtain a second PPG signal.

[0071] The user's original pulse wave data can be collected through a PPG photoelectric pulse wave sensor. The PPG photoelectric pulse wave sensor can be set in the above-mentioned portable device or other devices with PPG collection capabilities. The PPG photoelectric sensor can include visible light such as green light and red light, and can also include invisible light of multiple wavelengths such as infrared light.

[0072] Due to the sensor itself, the wearing of the sensor and the design of the acquisition scheme, there is a lot of noise such as baseline drift in the process of collecting PPG signals. These signals will greatly cover up the information of the pulse wave. Therefore, the collected original PPG signal is preprocessed.

[0073] The preprocessing may be filtering, and the filtering may be bandpass filtering. The filtering may use a zero-phase-shift digital filter to perform bandpass filtering on the signal. The signal may be processed by bandpass filtering to obtain a second PPG signal with baseline drift and high-frequency noise removed. The preprocessing may also be wavelet transform and reconstruction. After wavelet transform, the second PPG signal with baseline drift and high-frequency noise removed is obtained by removing the wavelet decomposition components in a specific range and then reconstructing. Similarly, the preprocessing may also be empirical mode decomposition (EMD) and reconstruction. After EMD decomposition, the intrinsic mode functions under some decomposition times are removed, and the data is reconstructed to obtain a PPG signal with baseline drift and high-frequency noise removed, i.e., a second PPG signal.

[0074] S2. Locate the peak point of the second PPG signal to obtain a first peak point sequence and a first PPI sequence.

[0075] The peak point positioning method can be the peak detection function findpeaks provided by the MATLAB software, which is used to find the peak in the PPG signal to obtain the first peak point sequence and the first PPI sequence.

[0076] S3. Calculate the variation index of the first PPI sequence, and determine the variation index judgment threshold according to the variation index of the first PPI sequence.

[0077] This step specifically includes:

[0078] S301, calculating the variation index of the first PPI sequence according to the following formula:

[0079]

[0080] In the formula, VI represents the variation index, x represents the first PPI sequence, dif represents the first-order difference of sequence x, abs represents the absolute value of the value in the sequence, std represents the standard deviation of the sequence, and mean represents the mean value of the sequence;

[0081] S302. Determine the variation index judgment threshold according to the variation index of the first PPI sequence according to the following formula. As shown in the following formula, an upper limit of the variation index and a lower limit of the variation index are preset. When the variation index exceeds the upper limit of the variation index, the threshold is defined as the upper limit of the first variation degree. When the variation index is lower than the lower limit of the variation index, the threshold is defined as the lower limit of the variation index. When the variation index is between the upper limit and the lower limit, the threshold is defined to be equal to the variation index.

[0082]

[0083] Among them, VI threshold represents the variation index judgment threshold, VI max Represents the upper limit of the variation index, VImin Represents the lower limit of the variation index.

[0084] S4. Calculate the variation index of each point in the first PPI sequence, and identify the points exceeding the variation index judgment threshold as noise and delete them, so as to obtain a second PPI sequence and a second peak point sequence.

[0085] This step specifically includes:

[0086] S401, set the calculation window and moving step size; the calculation window size can be 20 and the moving step size can be 10;

[0087] S402, moving the calculation window according to the moving step length, and calculating the variation index of the sequence segment located in the current calculation window in the first PPI sequence;

[0088] The formula for calculating the variation index is:

[0089]

[0090] Where, VI j represents the variation index calculated when the j-th moving calculation window is calculated, x j represents the sequence segment of the first PPI sequence in the j-th moving calculation window, dif represents the first-order difference, abs represents the absolute value of the value in the sequence, std represents the standard deviation of the sequence, and mean represents the mean of the sequence;

[0091] S403: for each point in the first PPI sequence, obtain the variation index of the sequence segments of all calculation windows containing the point, and use the mean of the obtained variation index as the variation index of the corresponding point;

[0092] S404, determining whether the variation index of each point in the first PPI sequence exceeds a determination threshold, wherein if the variation index of a sequence segment in a calculation window exceeds the determination threshold, the variation indexes of all points in the calculation window are determined to exceed the determination threshold;

[0093] S405: Delete points exceeding the judgment threshold in the first PPI sequence, and the deleted points form a second PPI sequence. The peak points remaining after the deletion form a second peak point sequence.

[0094] S5. Based on a preset minimum peak interval range, search the second PPG signal for an optimal minimum peak interval, and segment the second PPG signal according to the optimal minimum peak interval to obtain a third peak point sequence and a third PPI sequence.

[0095] It is not difficult to understand that in the process of finding the peak point of the signal, the search can be carried out by specifying constraints such as the minimum distance between peak points and the minimum value of peak points. The minimum interval basically corresponds to the minimum heart rate of the subject within a period of time, which should be equal to or slightly lower than the minimum heart rate. Since the heart rate of the subject is in dynamic change, the optimal minimum interval in different time ranges is different. Therefore, in this embodiment, when the PPG data is long or the calculated heart rate fluctuates greatly, the calculation can be divided into segments according to the actual situation to find the optimal minimum interval segment by segment.

[0096] Step S5 specifically includes:

[0097] S501, setting the minimum peak interval range to be from a first preset value times to a second preset value times the average value of the second PPI sequence; in this embodiment, the first preset value is selected to be 0.5 and the second preset value is selected to be 0.6, that is, the minimum peak interval range is set to 0.5-0.6 times the average value of the second PPI sequence;

[0098] S502, setting the value of the minimum interval of the peak points to the minimum value of the minimum peak interval range;

[0099] S503. Find the peak of the second PPG signal according to the minimum interval between peak points. The finding method may be the findpeaks function built into the MATLAB software. When using this function, specify the minimum peak interval parameter of the function as the minimum interval between peak points, and then search for the peak point of the data. The finding method may also be: find the maximum value within the minimum peak interval range as the peak value, and then move point by point to find the next maximum value point. If there is a point that is larger than the peak value marked within the range, unmark the peak and mark the peak at the new maximum value point.

[0100] S504. Determine the peak point interval sequence according to the peak value found in step S503, and calculate the degree of variation of each peak point interval; wherein, the degree of variation may be calculated by the standard deviation, and the degree of variation of the peak point interval is quantitatively evaluated by calculating the standard deviation of the peak point interval, or may be calculated by the variation index calculation method in step 301.

[0101] S505, determining whether the minimum value in the peak point interval sequence is equal to the maximum value of the minimum peak interval range, if not, executing step S506, if yes, executing step S507;

[0102] S506, adding 1 to the value of the minimum interval of the peak points, and executing step S503;

[0103] S507: taking the peak point interval with the smallest variation as the optimal minimum peak interval, and dividing the second PPG signal periodically by the optimal minimum peak interval to obtain a third peak point sequence, and taking the intervals between all peak points as a third PPI sequence.

[0104] In this embodiment, the variation index of all possible values ​​within the defined range is directly searched. In other embodiments, the amount of search calculation can be reduced by pruning.

[0105] S6. Calculate the kurtosis of the third PPI sequence to obtain a first kurtosis sequence.

[0106] This step specifically includes:

[0107] S601. For the third PPI sequence, the window is moved according to a preset step size, and the kurtosis of the sequence segment within the window is calculated using the following formula:

[0108]

[0109] In the formula, kurtosis k It represents the kurtosis calculated when the window is moved for the kth time according to the preset step size. Refers to the PPI average value of the current window, x i Indicates the value of the i-th PPI point in the current window, n is the size of the window; the number of points in 0.5-1.5s can be used as the window size, and the step size is greater than or equal to 1 and less than or equal to the window length;

[0110] S602, when the window moves to the end of the third PPI sequence, the calculation ends, and all kurtosis component sequences are calculated to form a first kurtosis sequence.

[0111] In experimental research, points in the PPG waveform with a faster kurtosis change rate often correspond to areas where noise is located. Therefore, the location of the noise point can be determined by setting a kurtosis change rate threshold.

[0112] S7. Calculate the change speed of the first kurtosis sequence, determine the data whose change speed exceeds a preset change speed threshold as abnormal noise and delete it, so as to obtain a denoised fourth peak point sequence and a third PPG signal, wherein the third PPG signal is the desired denoised signal.

[0113] This step specifically includes:

[0114] S701, calculating the first-order difference of each data of the first kurtosis sequence as the change speed of the corresponding data;

[0115] S702: Delete data whose changing speed exceeds a preset changing speed threshold as noise, and obtain a fourth PPI sequence, a fourth peak point sequence, and a third PPG signal with kurtosis noise deleted.

[0116] It should be understood that the value of the change speed threshold should be determined based on experiments and should be determined in conjunction with certain data. In this embodiment, the selected threshold is 0.1.

[0117] Embodiment 2

[0118] This embodiment provides a PPG signal feature extraction method, such as Figure 2 As shown, steps S1 to S7 are the same as those in the first embodiment, and will not be repeated here. After step S7, the following steps are also included:

[0119] S8. Use the obtained fourth peak point sequence to segment the denoised third PPG signal, with each segment serving as a PPG pulse, thereby obtaining a PPG pulse sequence.

[0120] S9. Classify the PPG pulses according to the morphological characteristics and multiple order derivative characteristics of the PPG pulse sequence, and use the dicrotic wave localization method corresponding to the PPG pulse classification to locate the dicrotic wave of the PPG pulse.

[0121] Since the PPG waveform will produce different forms with the movement of the body, different heart beats and other factors, it is difficult to locate the dicrotic wave in various forms of PPG waveforms with a single method. Therefore, according to the summary and induction, the PPG pulse signal can be divided into four types according to the different characteristics of the PPG waveform: the first type of hidden dicrotic wave, the second type of hidden dicrotic wave, non-hidden dicrotic wave and standard dicrotic wave. The waveforms of standard dicrotic wave, non-hidden dicrotic wave and hidden dicrotic wave are as follows: Figure 3 As shown, Figure 3 (1) is the standard dicrotic wave, Figure 3 (2) is the first type of hidden dicrotic wave, Figure 3 (3) is the second type of hidden dicrotic wave. Since the non-hidden dicrotic wave and the standard dicrotic wave have similar characteristics, the difference between its morphology and the standard dicrotic wave is that the second peak of the waveform is higher than the first peak. Therefore, the same positioning method is used. The corresponding dicrotic wave positions are located by three different positioning methods for the four types of dicrotic waves.

[0122] Specifically include:

[0123] S901, invert each PPG pulse and locate the Jiangzhongxia, where the Jiangzhongxia is the trough closest to the midpoint of a signal segment;

[0124] S902, obtaining a PPG pulse from a PPG pulse sequence, determining whether the current PPG pulse is an abnormal signal that cannot locate a dicrotic wave, if determined to be an abnormal signal, discarding the pulse segment, and returning to step S902, if not an abnormal signal, executing step S903; wherein the determination method is: finding the first zero-crossing point after the third-order derivative signal of the PPG pulse falls in the middle gorge, if there is no zero-crossing point, determining it to be an abnormal signal, otherwise determining it to be a non-abnormal signal;

[0125] S903, determine whether the current PPG pulse is a first type of hidden dicrotic wave, if so, execute step S905, if not, execute step S904; wherein the determination method is: locate the first trough value of the fourth-order derivative of the PPG pulse after the gorge of the third-order derivative, if there is a trough value, it does not belong to the first type of hidden dicrotic wave, otherwise it is determined to belong to the first type of hidden dicrotic wave;

[0126] S904, determining whether the current PPG pulse is a second-type hidden dicrotic wave, if so, executing step S906, if not, executing step S907; wherein the determination method is: locating the first trough value of the fourth-order derivative of the PPG pulse after the gorge of the third-order derivative, if the trough value is greater than 0, determining it as a standard dicrotic wave or a non-hidden dicrotic wave, otherwise determining it as a second-type hidden dicrotic wave;

[0127] S905, locating the first type of hidden dicrotic wave, the locating method is: taking the middle value of the PPG pulse third-order derivative curve falling in the middle gorge and the first zero-crossing point position after the third-order derivative curve falls in the middle gorge as the position of the dicrotic wave;

[0128] S906, locating the second type of hidden dicrotic wave, the locating method is: taking the first peak of the third-order derivative curve of the PPG pulse after descending the middle gorge as the position of the dicrotic wave;

[0129] S907, locating the standard dicrotic wave or the non-hidden dicrotic wave, the method of locating is: taking the zero-crossing point of the PPG third-order derivative curve after the middle gorge as the position of the dicrotic wave.

[0130] S10. Calculate morphological parameter characteristics and PPI signal characteristics according to the systolic peak position and dicrotic wave position of the PPG pulse sequence.

[0131] Among them, the systolic peak height, pulse width, ratio of systolic peak width to overall width and other characteristics of a single PPG pulse signal can be calculated as the morphological characteristics of a single PPG pulse. To calculate the PPI characteristics, the characteristics of the fourth PPI sequence can be calculated by time domain method, frequency domain method and nonlinear method. The PPI mean, total power (TP), low frequency power (LF), high frequency power (HF), low frequency / high frequency (LF / HF), approximate entropy and other indicators of the PPI sequence can be calculated as the characteristics of the PPI sequence.

[0132] Embodiment 3

[0133] This embodiment provides a PPG denoising device. The embodiment of the present invention provides services for the implementation of the method of embodiment 1, including one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method of embodiment 1.

[0134] The system memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media. The memory may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of the present invention. A program / utility having a set (e.g., at least one) of program modules may be stored in, for example, a memory, such program modules including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination of these examples may include the implementation of a network environment. The program module generally performs the functions and / or methods in the embodiments described in the present invention. The processing unit executes various functional applications and data processing by running the program stored in the system memory, such as implementing the method of embodiment one.

[0135] Embodiment 4

[0136] This embodiment provides a PPG feature extraction device. The embodiment of the present invention provides services for the implementation of the method of embodiment 2, including one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method of embodiment 2.

[0137] The system memory may include computer system readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The device may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system may be used to read and write non-removable, non-volatile magnetic media. The memory may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of the present invention. A program / utility having a set (e.g., at least one) of program modules may be stored in, for example, a memory, such program modules including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment. The program module generally performs the functions and / or methods in the embodiments described in the present invention. The processing unit executes various functional applications and data processing by running the program stored in the system memory, such as implementing the method of embodiment 2.

Claims

1. A PPG signal denoising method, characterized in that: include: S1, obtaining an original PPG signal, and performing preprocessing to obtain a second PPG signal; S2. Locate the peak point of the second PPG signal to obtain a first peak point sequence and a first PPI sequence; S3, calculating the variation index of the first PPI sequence, and determining the variation index judgment threshold according to the variation index of the first PPI sequence; S4, calculating the variation index of each point in the first PPI sequence, and identifying and deleting the points exceeding the variation index judgment threshold as noise, to obtain a second PPI sequence and a second peak point sequence; S5. Based on a preset minimum peak interval range, search the second PPG signal for an optimal minimum peak interval, and segment the second PPG signal according to the optimal minimum peak interval to obtain a third peak point sequence and a third PPI sequence; S6. Calculate the kurtosis of the third PPI sequence to obtain a first kurtosis sequence; S7, calculating the change speed of the first kurtosis sequence, determining the data whose change speed exceeds a preset change speed threshold as abnormal noise and deleting it, to obtain a denoised fourth peak point sequence and a third PPG signal, wherein the third PPG signal is the desired denoised signal; Step S3 specifically includes: S301, calculating the variation index of the first PPI sequence according to the following formula: In the formula, VI represents the variation index, x represents the first PPI sequence, dif represents the first-order difference of the sequence x, abs represents the absolute value of the value in the sequence, std represents the standard deviation of the sequence, and mean represents the mean value of the sequence; S302: Determine the variation index judgment threshold according to the following formula based on the variation index of the first PPI sequence: Among them, VI threshold represents the variation index judgment threshold, VI max Represents the upper limit of the variation index, VI min Represents the lower limit of the variation index; Step S4 specifically includes: S401, setting the calculation window and moving step size; S402, moving the calculation window according to the moving step length, and calculating the variation index of the sequence segment located in the current calculation window in the first PPI sequence; S403: for each point in the first PPI sequence, obtain the variation index of the sequence segments of all calculation windows containing the point, and use the mean of the obtained variation index as the variation index of the corresponding point; S404, determining whether the variation index of each point in the first PPI sequence exceeds a determination threshold, wherein if the variation index of a sequence segment in a calculation window exceeds the determination threshold, the variation indexes of all points in the calculation window are determined to exceed the determination threshold; S405, deleting points exceeding the judgment threshold in the first PPI sequence, the deleted points forming a second PPI sequence, and the peak points remaining after the deletion forming a second peak point sequence; Step S5 specifically includes: S501, setting the minimum peak interval range to be from a first preset value times to a second preset value times the average value of the second PPI sequence; S502, setting the value of the minimum interval of the peak points to the minimum value of the minimum peak interval range; S503, finding the peak value of the second PPG signal according to the minimum interval between peak points; S504, determining a peak point interval sequence according to the peak value found in step S503, and calculating the variation degree of each peak point interval; S505, determining whether the minimum value in the peak point interval sequence is equal to the maximum value of the minimum peak interval range, if not, executing step S506, if yes, executing step S507; S506, adding 1 to the value of the minimum interval of the peak points, and executing step S503; S507: taking the peak point interval with the smallest variation as the optimal minimum peak interval, and dividing the second PPG signal periodically by the optimal minimum peak interval to obtain a third peak point sequence, and taking the intervals between all peak points as a third PPI sequence.

2. The PPG signal denoising method according to claim 1, characterized in that: Step S6 specifically includes: S601. For the third PPI sequence, the window is moved according to a preset step size, and the kurtosis of the sequence segment within the window is calculated using the following formula: In the formula, kurtosis k It represents the kurtosis calculated when the window is moved for the kth time according to the preset step size. Refers to the PPI average value of the current window, x i Indicates the value of the i-th PPI point in the current window, and n is the size of the window; S602: When the window moves to the end of the third PPI sequence, the calculation ends, and all the calculated kurtosis components are sequenced to form a first kurtosis sequence.

3. The PPG signal denoising method according to claim 1, characterized in that: Step S7 specifically includes: S701, calculating the change speed of each data in the first kurtosis sequence; S702: Delete data whose changing speed exceeds a preset changing speed threshold as noise, and obtain a fourth PPI sequence, a fourth peak point sequence, and a third PPG signal with kurtosis noise deleted.

4. A PPG signal feature extraction method, characterized in that: include: S8. De-noise the original PPG signal using the method of claim 1, and segment the de-noised third PPG signal using the obtained fourth peak point sequence, with each segment being a PPG pulse, thereby obtaining a PPG pulse sequence; S9, classifying the PPG pulses according to the morphological characteristics and multiple order derivative characteristics of the PPG pulse sequence, and performing dicrotic wave positioning on the PPG pulses using a dicrotic wave positioning method corresponding to the PPG pulse classification; S10. Calculate morphological parameter characteristics and PPI signal characteristics according to the systolic peak position and dicrotic wave position of the PPG pulse sequence.

5. The PPG signal feature extraction method according to claim 4, characterized in that: Step S9 specifically includes: S901, invert each PPG pulse and locate the Jiangzhongxia, where the Jiangzhongxia is the trough closest to the midpoint of a signal segment; S902, obtaining a PPG pulse from a PPG pulse sequence, determining whether the current PPG pulse is an abnormal signal that cannot locate a dicrotic wave, if determined to be an abnormal signal, discarding the pulse segment, and returning to step S902, if not an abnormal signal, executing step S903; wherein the determination method is: finding the first zero-crossing point after the third-order derivative signal of the PPG pulse falls in the middle gorge, if there is no zero-crossing point, determining it to be an abnormal signal, otherwise determining it to be a non-abnormal signal; S903, determine whether the current PPG pulse is a first type of hidden dicrotic wave, if so, execute step S905, if not, execute step S904; wherein the determination method is: locate the first trough value of the fourth-order derivative of the PPG pulse after the gorge of the third-order derivative, if there is a trough value, it does not belong to the first type of hidden dicrotic wave, otherwise it is determined to belong to the first type of hidden dicrotic wave; S904, determining whether the current PPG pulse is a second-type hidden dicrotic wave, if so, executing step S906, if not, executing step S907; wherein the determination method is: locating the first trough value of the fourth-order derivative of the PPG pulse after the gorge of the third-order derivative, if the trough value is greater than 0, determining it as a standard dicrotic wave or a non-hidden dicrotic wave, otherwise determining it as a second-type hidden dicrotic wave; S905, locating the first type of hidden dicrotic wave, the locating method is: taking the middle value of the PPG pulse third-order derivative curve falling in the middle gorge and the first zero-crossing point position after the third-order derivative curve falls in the middle gorge as the position of the dicrotic wave; S906, locating the second type of hidden dicrotic wave, the locating method is: taking the first peak of the third-order derivative curve of the PPG pulse after descending the middle gorge as the position of the dicrotic wave; S907, locating the standard dicrotic wave or the non-hidden dicrotic wave, the method of locating is: taking the zero-crossing point of the PPG third-order derivative curve after the middle gorge as the position of the dicrotic wave.

6. A PPG denoising device, characterized in that include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 3.

7. A PPG feature extraction device, characterized in that include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 4 and 5.

Images