A method and system for motion artifact removal based on differential PPG signals

Abstract

The application belongs to the technical field of optical signal processing, and provides a motion artifact removal method and system based on differential PPG signals. The motion artifact removal method based on differential PPG signals comprises dynamically adjusting the light source intensity of each photoelectric sensor in a wearable device according to different subjects, acquiring PPG signals received by a plurality of groups of photoelectric sensors in real time, calculating a corresponding differential signal sequence according to each group of PPG signals, and constructing a differential matrix by attributing each column of signals to a group of photoelectric sensors; calculating the mean value of each column of signals in the differential matrix to obtain a mean signal set; determining the motion artifact signal interference degree of each group of photoelectric sensors according to the size of each element in the mean signal set, and then obtaining a preliminary motion artifact removed pulse wave signal, from which the pulse wave signal interfered by the motion artifact signal is adaptively removed again to obtain a final motion artifact removed pulse wave signal.

Description

Technical Field

[0001] This invention belongs to the field of optical signal processing technology, and in particular relates to a method and system for removing motion artifacts based on differential PPG signals. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Photoplethysmography (PPG) is a non-invasive biosignal that obtains cardiovascular pulsation information by measuring changes in light absorption on the skin surface. Its advantages lie in its simplicity, non-invasiveness, and low cost, making it widely used in heart rate monitoring, blood oxygenation measurement, and emotion recognition. PPG signals can be detected using optical sensors. However, in practical applications, the accuracy and reliability of PPG signals are often affected by motion artifacts (MA), especially during exercise or physical activity. MA manifests as distortion of the PPG signal due to muscle movement, making it difficult to obtain accurate physiological parameters.

[0004] Currently, accelerometers are often used in exercise heart rate monitoring technology to help remove motion sickness (MA). However, there is no strong correlation between acceleration and MA. For example, the acceleration and MA collected when the device is well-fitted to the skin and under strong restraint differ significantly from those collected when worn under normal pressure. The acceleration signal detected by the accelerometer includes not only limb and trunk motion information but also the effect of gravitational acceleration, which cannot completely and accurately characterize the relative motion between the PPG sensor and the skin.

[0005] In patent CN 109044315 A, "A Device and Method for Detecting Exercise Heart Rate," the reference light intensity is set below a preset minimum threshold for human blood absorption rate. Motion artifacts (MA) are filtered out by subtracting the reference light from the signal light containing the pulse wave signal. However, since the reference light and signal light are not simply additive or subtractive in amplitude, subtraction alone cannot effectively eliminate motion artifacts. In patent CN101632588A, "A Method for Measuring Blood Oxygen Saturation to Eliminate Motion Interference," an adaptive filter is constructed using the difference between two standardized PPG signals of different wavelengths as an ideal signal. This filter reduces some interference, and then reference signal series are constructed according to blood oxygen saturation levels from 0% to 100%. Different wavelengths of light have different penetration depths and carry different information. Using different wavelengths for subtraction is inflexible in its use of light intensity and wavelength, resulting in inaccurate reflection of pulse wave information. Neither of these patents considers the impact of individual differences on the signal, hindering the algorithm's versatility.

[0006] In summary, existing technologies for addressing motion artifact interference with PPG signals still suffer from the following problems: using reference signals to compensate for PPG is unreasonable; reference signals have poor adaptability to different motion interferences; and motion artifacts cannot be eliminated in real time based on individual differences. Summary of the Invention

[0007] To address the technical problems mentioned above, this invention provides a method and system for removing motion artifacts based on differential PPG signals, which improves the accuracy and versatility of biosignal measurements.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] The first aspect of the present invention provides a method for motion artifact removal based on differential PPG signals.

[0010] A method for motion artifact removal based on differential PPG signals, comprising:

[0011] The light intensity of each photoelectric sensor in the wearable device is dynamically adjusted according to different subjects. The PPG signals received by several groups of photoelectric sensors are acquired in real time. The corresponding differential signal sequence is obtained by subtracting the PPG signals of each group. The differential matrix is ​​constructed by assigning each signal to a group of photoelectric sensors.

[0012] The mean value of each column of the signal in the difference matrix is ​​calculated to obtain the mean signal set. The degree of motion artifact interference received by each group of photoelectric sensors is determined based on the magnitude of each element in the mean signal set.

[0013] The PPG signal is filtered according to the degree of interference from motion artifacts to obtain a pulse wave signal with preliminary motion artifact removal.

[0014] Based on the set physiological indicators and motion artifact signal characteristics, the pulse wave signal that is interfered with by the motion artifact signal is adaptively removed again from the pulse wave signal that has been initially de-motion artifacted, so as to obtain the final pulse wave signal that has been de-motion artifacted.

[0015] As one implementation method, the process of filtering the PPG signal according to the degree of motion artifact interference is as follows:

[0016] Remove all signals detected by the group of photoelectric sensors with the highest level of interference;

[0017] The average value of the PPG signals of the remaining photoelectric sensors is then calculated to obtain the pulse wave signal with preliminary motion artifact removal.

[0018] As one implementation method, the process of filtering the PPG signal according to the degree of motion artifact interference is as follows:

[0019] Based on the degree of motion artifact signal interference received by each group of photoelectric sensors, the weighting coefficient of the average PPG signal of each group of photoelectric sensors is obtained.

[0020] Then, the PPG signal mean of all groups of photoelectric sensors is weighted and summed to obtain the initial motion artifact-free pulse wave signal.

[0021] In one implementation, the mean PPG signal of each group of photoelectric sensors is negatively correlated with its weighting coefficient.

[0022] As one implementation, the sum of the weighting coefficients of the average PPG signals of all groups of photoelectric sensors is equal to 1.

[0023] As one implementation method, the process of dynamically adjusting the light source intensity of each photoelectric sensor in the wearable device according to different subjects is as follows:

[0024] A pulse wave signal was acquired from the subject under resting conditions, and after preprocessing, a reference signal PWS free of motion artifacts was obtained. ref ;

[0025] When the subject begins to move, the light intensity of the photoelectric sensor is set to the preset maximum light intensity I. MAX Record the current PPG signal and calculate its differential signal DS. The current light intensity is denoted as I. d I d =I MAX ;

[0026] Calculate DS and PWS ref Correlation coefficient r1: If r1>R _th1 Then adjust the light intensity to I. d_next =1 / 2*I d_pre Calculate the corresponding differential signal DS, and then continue to calculate the current DS and PWS. ref The correlation coefficient is calculated until r1 is satisfied. <R _th1 At that time, the light intensity adjustment process stops; among which, R _th1 The correlation threshold; I d_next I is the light intensity for the next PPG signal acquisition. d_pre The light intensity is the light intensity of the PPG signal collected last time.

[0027] As one implementation method, the process of dynamically adjusting the light source intensity of each photoelectric sensor in the wearable device according to different subjects is as follows:

[0028] A pulse wave signal was acquired from the subject under resting conditions, and after preprocessing, a reference signal PWS free of motion artifacts was obtained. ref ;

[0029] As the subject begins to move, the light intensity of the photoelectric sensor is gradually adjusted from zero to a preset maximum light intensity I. MAX The adjustment step size is ΔI, and the adjustment time interval is T. Within the time interval T, the PPG signal under the current light intensity is recorded and its differential signal DS is calculated.

[0030] After the light intensity adjustment is completed, a set of PWS is obtained. _mat {PWS1,PWS2,…,PWS n} and a set of DS _mat1 {DS1,DS2,…,DS n The subscripts correspond one-to-one with the light intensity.

[0031] Calculate DS _mat1 Each column of DS in i Signals and PWS ref The correlation coefficient 'a', and all those satisfying 'a' <R _th1 The differential signals DS form a set DS _mat2 Among them, R _th1 This is the correlation threshold;

[0032] Calculate DS _mat2 Each column of DS i Signals and PWS _mat The corresponding subscript PWS i The correlation coefficient b of the signal, when b is at its maximum, represents the corresponding DS. m With PWS m The correlation between motion artifacts and interference is the highest, so the light intensity corresponding to the current index is the optimal value.

[0033] A second aspect of the present invention provides a motion artifact removal system based on differential PPG signals.

[0034] The differential matrix construction module is used to dynamically adjust the light intensity of each photoelectric sensor in the wearable device according to different subjects, acquire PPG signals received by several groups of photoelectric sensors in real time, and then obtain the corresponding differential signal sequence by subtracting each group of PPG signals, and construct the differential matrix by assigning each column of signals to a group of photoelectric sensors.

[0035] The motion artifact interference level determination module is used to calculate the mean value of each column of the signal in the difference matrix, obtain the mean signal set, and determine the degree of motion artifact signal interference received by each group of photoelectric sensors based on the magnitude of each element in the mean signal set.

[0036] The preliminary motion artifact removal module is used to filter the PPG signal according to the degree of interference of the motion artifact signal to obtain the pulse wave signal with preliminary motion artifact removal.

[0037] The final motion artifact removal module is used to adaptively remove the pulse wave signal that is interfered with by the motion artifact signal from the pulse wave signal obtained by the initial motion artifact removal based on the set physiological index information and the characteristics of the motion artifact signal, so as to obtain the final motion artifact removed pulse wave signal.

[0038] A third aspect of the present invention provides a computer-readable storage medium.

[0039] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the motion artifact removal method based on differential PPG signals as described above.

[0040] A fourth aspect of the present invention provides an electronic device.

[0041] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the motion artifact removal method based on differential PPG signals as described above.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] (1) In the case of measuring the relative movement between the wearable device and the skin surface using a PPG sensor, the present invention can obtain an optical signal strongly correlated with motion artifacts without additional sensors (such as accelerometers, piezoelectric sensors, etc.) and use it as a means to remove motion artifacts in the PPG. Unlike accelerometers that rigidly collect information on device acceleration, the present invention measures the relative movement direction based on optical methods and calculates it according to the changes in PPG waveforms during actual movement. This can more accurately characterize the relative movement between the watch and the skin during movement and improve adaptability to different motion interferences.

[0044] (2) This invention sets the positional distribution of several groups of PD (Photodiode) sensors and the relative positional relationship of the differential signal (DS) reference light source, so that motion artifacts can better reflect motion artifacts. In addition to taking into account the influence of spatial position on DS, the light intensity of the original PPG signal reference light source of the differential signal DS is set to a low level. Under this low light intensity, only a small amount of pulse wave signal is contained. Furthermore, the pulse wave component in the signal is removed through the differential algorithm, so that it only contains components related to motion artifacts, which is more suitable as a noise reference signal for adaptive filters. Motion artifacts in PPG signals are eliminated by relying on the strong correlation between motion artifacts, adaptive filtering algorithms and statistical methods.

[0045] (3) By dynamically adjusting the light intensity for different subjects, the present invention makes the differential signal contain only a small number of pulse wave components and has a high correlation with motion artifacts, making it better able to characterize motion artifacts and avoiding the influence of individual differences.

[0046] (4) This invention proposes a motion artifact removal method based on differential signals, which can eliminate the influence of individual differences on the signal and make it a reference signal for extracting superior signals from pulse wave signals. This is essential for the calculation of relevant physiological parameters that rely on feature point recognition, such as blood pressure. Different signal construction methods are used to further eliminate the influence of individual differences on device measurements.

[0047] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0048] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0049] Figure 1 This is a system structure diagram of the PPG sensor for measuring relative movement according to an embodiment of the present invention;

[0050] Figure 2(a) is a waveform diagram of the motion interference time-differential signal along the first straight line direction according to an embodiment of the present invention;

[0051] Figure 2(b) is a waveform diagram of the motion interference time-differential signal along the second straight line direction according to an embodiment of the present invention;

[0052] Figure 3(a) is a waveform diagram of PPG and MA signals in a static state according to an embodiment of the present invention;

[0053] Figure 3(b) is a waveform diagram of PPG and MA signals under the perturbation of the swinging arm according to an embodiment of the present invention;

[0054] Figure 3(c) is a waveform diagram of PPG and MA signals under the disturbance of the rotating arm in an embodiment of the present invention;

[0055] Figure 4 This is a timing diagram of the PDs and LEDs in an embodiment of the present invention;

[0056] Figure 5 This is a flowchart of heart rate calculation according to an embodiment of the present invention;

[0057] Figure 6 This is a flowchart of blood pressure calculation according to an embodiment of the present invention;

[0058] Figure 7This is a flowchart of the motion artifact removal method based on differential PPG signals according to an embodiment of the present invention. Detailed Implementation

[0059] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0060] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0061] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0062] Example 1

[0063] like Figure 7 As shown, this embodiment provides a method for motion artifact removal based on differential PPG signals, which includes:

[0064] Step 1: Dynamically adjust the light source intensity of each photoelectric sensor in the wearable device according to different subjects, acquire PPG signals received by several groups of photoelectric sensors in real time, and then obtain the corresponding differential signal sequence by subtracting each group of PPG signals. Construct a differential matrix by assigning each signal to a group of photoelectric sensors.

[0065] Previously, it was generally believed that the wrist was the optimal position for data collection, and that wearable devices should be in close contact with the skin. However, in daily use, wearable devices are not always in a single plane, and may even form a spatial angle with the skin due to movement. Current solutions involve using sensors such as accelerometers and gyroscopes to collect multi-axis acceleration within the space of the smart wearable device, using acceleration as a selection or compensation criterion between optical sensor signals. However, because wearable devices are not rigidly connected to the skin, the acceleration of the wearable device and its dynamic range (MA) are not strongly correlated.

[0066] Compared to the third derivative of displacement, such as acceleration, there is an urgent need for a signal acquisition and representation method to measure the relative displacement between wearable devices and skin. This variable needs to be acquired and correlated with the PPG signal time series. By removing the interval with relative movement (MA), the effective information (pulse wave component) in the PPG signal can be retained. Therefore, a method for measuring the relative movement between a wristwatch and the skin surface using a PPG sensor is proposed, such as... Figure 1 As shown.

[0067] The PPG signals received by PD1, PD2, PD3, PD4, PD5, and PD6 under the light intensities of LED1, LED2, LED3, LED4, LED5, and LED6 are named PPG1, PPG2, PPG3, PPG4, PPG5, and PPG6, respectively. The incident light at this intensity is sufficient to penetrate muscle tissue and reach blood vessels, so that the signal contains useful pulse wave signals. The intensity range can be determined experimentally. The signal sequence DS1 is obtained by subtracting the PPG signals (PPGD1, PPGD2) received by PD1 and PD2 under the light intensities of LED1 and LED2; the signal sequence DS2 is obtained by subtracting the PPG signals (PPGD3, PPGD4) received by PD3 and PD4 under the light intensities of LED3 and LED4; the signal sequence DS3 is obtained by subtracting the PPG signals (PPGD5, PPGD6) received by PD5 and PD6 under the light intensities of LED5 and LED6; DS1, DS2, and DS3 are denoted as the difference matrix D1{DS1; DS2; DS3}.

[0068] Perform the following operations on D1:

[0069] Step 1: Calculate the mean of each column in D1 to obtain a set of means A: {a1, a2, a3};

[0070] Step 2: Based on the values ​​of the three variables in A, determine the relative movement direction between the device and the skin:

[0071] Specifically, ① when a1>a2>a3, the relative movement of the watch is mainly concentrated in the direction of the first straight line; ② when a2>a1>a3, the relative movement of the watch is mainly concentrated in the direction of the second straight line; ③ when a3>a1≈a2, the relative movement of the watch is mainly concentrated in the direction of the third straight line; ④ when a1≈a2>a3, the relative movement of the watch is mainly concentrated in the direction of the superposition vector of the first and second straight lines and is perpendicular to the direction of the third straight line; ⑤ to determine a more detailed orientation, it can be compared with the previously established experience database. As shown in Figures 2(a) and 2(b).

[0072] Based on the above method, a DS signal sequence with strong correlation to MA is obtained, which can be used as a reference signal for removing MA from the PPG signal.

[0073] Two implementation methods are provided below: PWS (pulse wave signal) and DS (differential signal) which are strongly correlated with MA (motion artifact) are obtained based on the above method.

[0074] PPGP1 = (PPG1 + PPG2) / 2;

[0075] PPGP2 = (PPG3 + PPG4) / 2;

[0076] PPGP3 = (PPG5 + PPG6) / 2;

[0077] For obtaining PWS:

[0078] 1. Since the acquisition of the DS1 signal and PPGP1 are on the same straight line, when a1 is at its maximum value, the PPGP1 signal is most susceptible to interference, resulting in less useful signal. If post-processing of PPGP1 continues, its accuracy will decrease. Therefore, PPGP1 can be deleted. At this point, PWS = (PPG... P2 +PPG P3 ) / 2 (where PWS is the pulse wave signal). Similarly, in other cases, the PPGP signal in the direction of motion is removed, and the remaining PPGP signal is averaged and then processed to obtain PWS.

[0079] 2. Example 1 described above describes filtering the three PPGP signals along three straight lines based on the magnitudes of the three variables {a1, a2, a3} in D3, removing the worst PPGP signal, and averaging the remaining PPGP signals as the final PWS signal. However, in actual use of a wristwatch, motion is often complex; it is not a single straight line but a vector superposition of two lines with different proportions. Therefore, if a reference line PGP is added for each additional vector direction... P Signal acquisition is impractical due to limitations in equipment size and optimization; increasing the number of signal acquisition channels is not feasible. Therefore, a weighted assignment method is proposed. Based on the difference matrix D1, three weight coefficients M1, M2, and M3 are obtained. The PPGP signals are then weighted to obtain the final PWS, i.e., PWS = M1 * PPG. P1 +M2*PPG P2 +M3*PPG P3 .

[0080] For the acquisition of DS signals:

[0081] (1) Find the maximum value a in the mean set A: {a1, a2, a3}. i If the motion disturbance is greatest in that direction, then DSi is selected as DS.

[0082] (2) Based on the difference matrix D1, three weight coefficients N1, N2 and N3 are obtained. After weighting the DS1, DS2 and DS3 signals, the final DS is obtained, that is, DS = N1*DS1 + N2*DS2 + N3*DS3.

[0083] During DS acquisition, the reference light source is set to a relatively low intensity. At this point, the incident light is almost completely absorbed by muscle tissue and melanin before it reaches the blood vessels deep within the skin tissue. Consequently, the PPG signal at this time will contain less pulse wave signal. The PPG signals acquired in PD at the same time should have the same pulse wave component. Therefore, to further eliminate the pulse wave component carried in the reference signal, differential processing is performed on the two signals.

[0084] Taking the first straight line as an example, in terms of positional distribution, PD1 and LED1 are closer together, so the reference light source for PD1 comes more from LED1. Similarly, the reference light source for PD2 comes more from LED2. When MA occurs, i.e., when the device and skin undergo relative displacement along the first straight line, due to the greater distance between LED1 and LED2, the changes in the skin condition and the magnitude of the relative displacement are different, resulting in differences in the photoelectric signals received by PD1 and PD2. In this case, PPGD1 ≠ PPGD2.

[0085] Because it is differential, DS does not contain pulse wave signals; therefore, the magnitude of DS is only related to the degree of MA interference. Figures 3(a)-3(c) It can be seen that DS and PWS have a strong correlation. By utilizing this strong correlation, MA in PWS can be quickly removed, which helps to improve the real-time performance of the device's data acquisition.

[0086] like Figure 4 As shown in the figure, when the device is working, the working timing of PDs and LEDs is as shown in the figure. The pulse wave signal and PPG signal are alternately acquired and recorded in a short period of time. The acquisition interval is 100us, which is much smaller than the sampling period of 10ms (a difference of 100 times). Therefore, it can be considered that the two signals are acquired at the same time.

[0087] When muscle activity (MA) is generated and the intensity of the reference light for the digital subtraction (DS) signal is low, it is almost completely absorbed by muscle tissue and melanin before reaching the blood vessels deep within the skin tissue. In this case, the pulse wave component in the pulsed light (PPG) signal will be less, while the component related to MA will be more prominent. When the incident light intensity is too low, the correlation between DS and MA will also be very weak; however, when the incident light intensity is too high, DS will carry a large pulse wave component, which is very disadvantageous as a reference signal for MA. Due to individual differences (different muscle tissue thickness and skin melanin content among individuals), the amount of pulse wave component in the DS signal acquired under the same light intensity will vary. Therefore, it is necessary to find a suitable light intensity for a specific individual, so that the DS signal contains a small amount of pulse wave component while ensuring a strong correlation with MA.

[0088] In some embodiments, the process of dynamically adjusting the light source intensity of each photoelectric sensor in the wearable device according to different subjects is as follows:

[0089] A pulse wave (PWS) signal was acquired from the subject under resting conditions. After preprocessing, a reference PWS signal free from motion artifact interference (MA) was obtained. ref ;

[0090] When the subject begins to move, the light intensity of the photoelectric sensor is set to the preset maximum light intensity I. MAX Record the PWS and PPG signals and calculate the differential (DS) signal of the PPG signal. The current light intensity is denoted as I. d I d =I MAX ;

[0091] Comparison of DS and PWS ref The correlation between DS and PWS was obtained. ref The correlation coefficient r1; if r1>R _th1 Then adjust the light intensity to I. d_next =1 / 2*I d_pre Calculate the corresponding DS signal, and then continue to compare the current DS and PWS. ref The correlation until r1 is satisfied. <R _th1 At that time, the light intensity adjustment process stops; among which, R _th1 The correlation threshold; I d_next I is the light intensity for the next PPG signal acquisition. d_pre This is the light intensity from the previous time.

[0092] Among them, the correlation threshold R _th1 The maximum light intensity I can be determined experimentally. MAX It can be determined by experiment.

[0093] The above process ensures that the DS contains a small number of pulse wave signals under the highest possible light intensity.

[0094] In some alternative embodiments, the process of dynamically adjusting the light source intensity of each photoelectric sensor in the wearable device according to different subjects is as follows:

[0095] A pulse wave (PWS) signal was acquired from the subject under resting conditions. After preprocessing, a reference PWS signal free from motion artifact interference (MA) was obtained. ref ;

[0096] As the subject begins to move, the light intensity of the photoelectric sensor is gradually adjusted from zero to a preset maximum light intensity I. MAX The adjustment step size is ΔI, and the adjustment time interval is T. Within the time interval T, the PWS signal and PPG signal under the current light intensity are recorded, and the differential (DS) signal of the PPG signal is obtained.

[0097] After the light intensity adjustment is completed, a set of PWS is obtained. _mat {PWS1,PWS2,…,PWS n} and a set of DS _mat1 {DS1,DS2,…,DS n The subscripts correspond one-to-one with the light intensity.

[0098] Calculate DS _mat1 Each column of DS i Signals and PWS ref The correlation coefficient α (calculated using methods such as correlation analysis, spectral analysis, adaptive filtering, independent component analysis, and time-domain analysis) is used to determine the correlation coefficient α. All parameters satisfying α are then considered. <R _th1 The DS signal constitutes a set of DS signals. _mat2 Among them, R _th1 This is the correlation threshold;

[0099] Calculate DS _mat2 Each column of DS i Signals and PWS _mat The corresponding subscript PWS i The correlation coefficient b of a signal (calculated using methods such as correlation analysis, spectral analysis, adaptive filtering, independent component analysis, and time-domain analysis) represents the maximum value of the corresponding DS. m With PWS m If the correlation between the MA and the current index is the highest, then the light intensity corresponding to the current index is the optimal value.

[0100] ΔI and T can be set by the user, for example, adjusting the light intensity from 0 to 20mA current with an adjustment step of 1mA and a time interval of 2s.

[0101] Step 2: Calculate the mean value of each column of the signal in the difference matrix to obtain the mean signal set, and determine the degree of motion artifact signal interference to each group of photoelectric sensors based on the magnitude of each element in the mean signal set.

[0102] Step 3: Filter the PPG signal according to the degree of motion artifact interference to obtain a preliminary motion artifact-free pulse wave signal.

[0103] In one or more embodiments, the average PPG signal of the photoelectric sensor group that continues to participate in the calculation is obtained by removing all signals detected by the group of photoelectric sensors with the highest interference.

[0104] Specifically, after removing all signals detected by the group of photoelectric sensors with the greatest interference, the average value of the PPG signals of the remaining groups of photoelectric sensors is calculated to obtain the pulse wave signal with preliminary motion artifact removal.

[0105] For example, since the source signal acquisition of DS1 is in the same straight line as PPGP1, when a1 is at its maximum value, the PPGP1 signal is most affected by interference, resulting in less useful signal. If PPGP1 is further post-processed, its accuracy will decrease. Therefore, PPGP1 can be deleted. In this case, PWS = (PPGP2 + PPGP3) / 2 (pulse wave signal: PWS). The same principle applies to other cases. The PPGP signals in the motion direction are removed, and the remaining PPGP signals are averaged and post-processed to obtain PWS. For the DS signal, the DS signals of the three straight lines correspond to the three PPGP signals respectively: the DS signal serves as the reference signal for removing MA from PPGP. When a1 is at its maximum, DS1 is removed, and DS2 and DS3 serve as the reference signals for PPGP2 and PPGP3 respectively. The same principle applies to other cases. When removing the PPGP signals in the motion direction, the corresponding DS signals are also deleted.

[0106] In other embodiments, a weighting coefficient is obtained for the average PPG signal of each group of photoelectric sensors based on the degree of motion artifact signal interference received by each group of photoelectric sensors. The initial motion artifact-free pulse wave signal is obtained by weighted summation of the average PPG signals of all groups of photoelectric sensors.

[0107] Among them, the mean PPG signal of each group of photoelectric sensors is negatively correlated with its weighting coefficient.

[0108] For example, based on the magnitudes of the three variables {a1, a2, a3} in D3, the three PPGP signals along the three straight lines are filtered, removing the worst PPGP signal, and the average of the remaining PPGP signals is taken as the final PWS signal. However, in actual use of a wristwatch, the motion is often complex, not a single straight line, but a vector superposition of two lines with different proportions. Adding a reference line PPGP signal for each additional vector direction is impractical due to limitations in device size and design. Therefore, a weighted assignment method is proposed. Based on the difference matrix D1, three weight coefficients M1, M2, and M3 are obtained. The PPGP signals are weighted to obtain the final PWS, i.e., PWS = M1*PPGP1 + M2*PPGP2 + M3*PPGP3.

[0109] Rules for setting weight coefficients (conventional weighting method):

[0110] Based on the three variables {a1, a2, a3} in the difference matrix D1, determine M1, M2, and M3 respectively. Sort a1, a2, a3, assuming a1 > a2 > a3, then set M1 = a3 / (a1 + a2 + a3), M2 = a2 / (a1 + a2 + a3), and M3 = a1 / (a1 + a2 + a3). That is, the larger the variable a is, the smaller the corresponding weight coefficient is assigned. The construction of DS is similar, obtained by weighting the three variables.

[0111] Step 4: Based on the set physiological index information and motion artifact signal characteristics, adaptively remove the pulse wave signal that is interfered with by the motion artifact signal from the initially motion artifact-free pulse wave signal to obtain the final motion artifact-free pulse wave signal.

[0112] Let the pulse wave signal for initial motion artifact removal be PWS0;

[0113] PWS0 contains rich information on human physiological indicators, but it is susceptible to interference from MA (Modulation-Modulation). Extracting physiological information from PWS0 presents challenges, and the accuracy of physiological indicators is negatively correlated with the degree of MA interference. To eliminate the influence of MA, baseline drift removal and filtering are first applied to the PWS0 and DS signals. Filtering algorithms include mean filtering, low-pass filtering, and SG filtering.

[0114] Heart rate calculation:

[0115] like Figure 5As shown, an FFT (Fast Fourier Transform Algorithm) is performed on PWS0 (the horizontal axis of the spectrum has been converted from frequency to heart rate, the conversion formula is HR = 60 * f, where f is the frequency and HR (Heartrate) represents the heart rate), to obtain the coordinates of the first peak and the second peak in the spectrum: (x1, y1) and (x2, y2), where x1 is tentatively set to the measured heart rate HR. mea1 The criteria for determining whether spectral energy is concentrated are as follows: (H th and L th If the above two equations are satisfied, then the energy is considered concentrated. (The constant is determined by experiments.)

[0116] 1. Perform curve fitting on historical heart rates. The number of points required for fitting can be set by the user; for example, select 10 historical data points for fitting {HR}. n-10 HR n-9 ,...,HR i ,...,HR n-1 Using the least squares method, we obtain the equation: f(i)=θ0+θ1i+θ2i 2 +…+θ k i k , where θ0, θ1, ..., θ k These are the coefficients of the polynomial. Substituting i = 11, we calculate the current heart rate f(11), denoted as HR. pre .

[0117] 2. If | HR mea1 -HR pre │<ΔHR th (ΔHR th (where is a constant representing the maximum permissible variation between two consecutive heart rate measurements) and if the spectral energy is concentrated as determined in step 1, then HR is considered to be... mea1 Reliable, representing the true heart rate value (HR) true Otherwise, it is assumed that PWS is significantly affected by MA interference, requiring further processing of PWS.

[0118] a) Use adaptive filtering algorithms (such as least squares (LMS), recursive least squares (RLS), Kalman filter, etc.) to remove MA. The input signal is the differential signal DS of PWS0 and PPG signals, where DS is the noise reference signal and the error signal is the output signal.

[0119] b) Perform an FFT on the adaptively filtered output signal (the horizontal axis of the spectrum has been converted from frequency to heart rate), obtain the horizontal axis of the first peak, and denote it as the measured heart rate HR. mea2 .

[0120] c) If |HR mea2 -HR pre │<ΔHR th Then they believe HR mea2 Reliable, representing the true heart rate value (HR) true If the difference is still large, then HR will be adjusted. pre As the true heart rate value HR true .

[0121] Blood pressure calculation:

[0122] Blood pressure calculation relies heavily on the identification of feature points in the PPG waveform. Therefore, it is essential to extract the signal segments in the PPG waveform that are not affected by MA interference and extract the feature points therein.

[0123] Heart rate is the signal period in PWS. When PWS is not interfered with by MA, due to the similarity of signal segments in each period, its statistical parameters, such as standard deviation, kurtosis, and skewness, are close. However, in segments interfered with by MA, these values ​​will deviate from the statistical parameter values ​​of undisturbed segments. Therefore, the statistical parameters of each signal segment can be used to screen for good periods in the signal.

[0124] The magnitude of the statistical parameters of each cycle of PWS is affected by individual differences, wearing position, and method; the statistical parameter values ​​of the interfered segments are related to the MA degree, and their deviation from the normal range may not be obvious. All of these factors make it difficult to rely on the statistical parameters of PWS for signal screening.

[0125] To overcome the above difficulties, DS is used instead of PWS as the reference signal for signal screening. The magnitude of DS is only related to MA and is not affected by individual differences. In the signal period without MA interference, DS is 0; when subjected to MA interference, DS is not 0, and its statistical parameters will deviate from 0.

[0126] Figure 6 The blood pressure calculation process of an example is given below, taking the standard deviation (SD) as an example:

[0127] 1. Based on heart rate (HR) true Calculate the periodic window length L = HR true *T / 60, where T is the sampling period.

[0128] 2. For PWS MA The PWS0 signal (including the MA segment) and DS are periodically divided, with each period having a length of L, resulting in n period segments {P1, P2, ... P}. i ,...P n} PWSMA and {D1,D2,...D} i,...D n} DS

[0129] 3. Calculate {D1, D2, ... D} i ,...D n} DS D i The standard deviation SD is used to obtain {SD1, SD2, ... SD}. i ,...SD n}

[0130] 4. For {SD1, SD2, ... SD... i ,...SD n} of SD i Perform threshold judgment, if SD i >SD th (SD th The threshold for distinguishing whether a periodic segment is affected by MA interference is a constant (determined experimentally). If the threshold is set, the i-th period is considered to be affected by MA interference; otherwise, it is not affected.

[0131] 5. Based on the result in 4, remove {P1, P2, ..., P} i ,...P n} PWSMA The corresponding periodic segment P i The remaining segments are reassembled to obtain a new PWS signal unaffected by MA interference. WMA .

[0132] From PWS WMA Feature values ​​are extracted, and blood pressure is calculated using a pre-established mathematical model.

[0133] Example 2

[0134] This embodiment provides a motion artifact removal system based on differential PPG signals, which includes:

[0135] The differential matrix construction module is used to dynamically adjust the light intensity of each photoelectric sensor in the wearable device according to different subjects, acquire PPG signals received by several groups of photoelectric sensors in real time, and then obtain the corresponding differential signal sequence by subtracting each group of PPG signals, and construct the differential matrix by assigning each column of signals to a group of photoelectric sensors.

[0136] The motion artifact interference level determination module is used to calculate the mean value of each column of the signal in the difference matrix, obtain the mean signal set, and determine the degree of motion artifact signal interference received by each group of photoelectric sensors based on the magnitude of each element in the mean signal set.

[0137] The preliminary motion artifact removal module is used to filter the PPG signal according to the degree of interference of the motion artifact signal to obtain the pulse wave signal with preliminary motion artifact removal.

[0138] The final motion artifact removal module is used to adaptively remove the pulse wave signal that is interfered with by the motion artifact signal from the pulse wave signal obtained by the initial motion artifact removal based on the set physiological index information and the characteristics of the motion artifact signal, so as to obtain the final motion artifact removed pulse wave signal.

[0139] It should be noted that each module in this embodiment corresponds one-to-one with each step in Embodiment 1, and their specific implementation processes are the same, so they will not be repeated here.

[0140] Example 3

[0141] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the motion artifact removal method based on differential PPG signals as described above.

[0142] Example 4

[0143] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the motion artifact removal method based on differential PPG signals as described above.

[0144] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0145] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for motion artifact removal based on differential PPG signals, characterized in that, include: The light intensity of each photoelectric sensor in the wearable device is dynamically adjusted according to different subjects. The PPG signals received by several groups of photoelectric sensors are acquired in real time. The corresponding differential signal sequence is obtained by subtracting the PPG signals of each group. The differential matrix is ​​constructed by assigning each signal to a group of photoelectric sensors. The mean value of each column of the signal in the difference matrix is ​​calculated to obtain the mean signal set. The degree of motion artifact interference received by each group of photoelectric sensors is determined based on the magnitude of each element in the mean signal set. The PPG signal is filtered based on the degree of motion artifact interference to obtain a preliminarily motion artifact-free pulse wave signal. The process of filtering the PPG signal based on the degree of motion artifact interference is as follows: Based on the degree of motion artifact signal interference received by each group of photoelectric sensors, the weighting coefficient of the average PPG signal of each group of photoelectric sensors is obtained. Then, the PPG signal mean of all groups of photoelectric sensors is weighted and summed to obtain the initial motion artifact-free pulse wave signal; Alternatively, the process of filtering the PPG signal based on the degree of motion artifact interference is as follows: Remove all signals detected by the group of photoelectric sensors with the highest level of interference; Then, the average value of the PPG signal of the remaining groups of photoelectric sensors is calculated to obtain the pulse wave signal with preliminary motion artifact removal. Based on the set physiological index information and motion artifact signal characteristics, the pulse wave signal that is interfered with by the motion artifact signal is adaptively removed again from the pulse wave signal that has been initially de-motion artifacted, so as to obtain the final pulse wave signal that has been de-motion artifacted. The process of dynamically adjusting the light source intensity of each photoelectric sensor in the wearable device according to different subjects is as follows: A pulse wave signal was acquired from the subject under resting conditions, and after preprocessing, a reference signal PWS free of motion artifacts was obtained. ref ; When the subject begins to move, the light intensity of the photoelectric sensor is set to the preset maximum light intensity I. MAX Record the current PPG signal and calculate its differential signal DS. The current light intensity is denoted as I. d I d =I MAX ; Calculate DS and PWS ref Correlation coefficient r1: If r1>R _th1 Then adjust the light intensity to I. d_next =1 / 2 I d_pre Calculate the corresponding differential signal DS, and then continue to calculate the current DS and PWS. ref The correlation coefficient is calculated until r1 is satisfied. <R _th1 At that time, the light intensity adjustment process stops; among which, R _th1 The correlation threshold; I d_next I is the light intensity for the next PPG signal acquisition. d_pre The light intensity of the PPG signal collected last time; The process of dynamically adjusting the light source intensity of each photoelectric sensor in the wearable device according to different subjects is as follows: A pulse wave signal was acquired from the subject under resting conditions, and after preprocessing, a reference signal PWS free of motion artifacts was obtained. ref ; As the subject begins to move, the light intensity of the photoelectric sensor is gradually adjusted from zero to a preset maximum light intensity I. MAX The adjustment step size is I, the adjustment time interval is T; within the time interval T, record the PPG signal under the current light intensity and calculate its differential signal DS; After the light intensity adjustment is completed, a set of PWS is obtained. _mat {PWS1, PWS2, … PWS n } and a set of DS _mat1 {DS1, DS2, … DS n The subscripts correspond one-to-one with the light intensity. Calculate DS _mat1 Each column of DS in i Signals and PWS ref The correlation coefficient 'a', and all those satisfying 'a' <R _th1 The differential signals DS form a set DS _mat2 Among them, R _th1 This is the correlation threshold; Calculate DS _mat2 Each column of DS in i Signals and PWS _mat The corresponding subscript PWS i The correlation coefficient b of the signal, when b is at its maximum, represents the corresponding DS. m With PWS m The correlation between motion artifacts and interference is the highest, so the light intensity corresponding to the current index is the optimal value.

2. The method for motion artifact removal based on differential PPG signals as described in claim 1, characterized in that, The mean PPG signal of each group of photoelectric sensors is negatively correlated with its weighting coefficient.

3. The motion artifact removal method based on differential PPG signals as described in claim 2, characterized in that, The sum of the weighting coefficients of the average PPG signal of all groups of photoelectric sensors equals 1.

4. A motion artifact removal system based on differential PPG signals, characterized in that, The motion artifact removal method based on differential PPG signals as described in any one of claims 1-3 includes: The differential matrix construction module is used to dynamically adjust the light intensity of each photoelectric sensor in the wearable device according to different subjects, acquire PPG signals received by several groups of photoelectric sensors in real time, and then obtain the corresponding differential signal sequence by subtracting each group of PPG signals, and construct the differential matrix by assigning each column of signals to a group of photoelectric sensors. The motion artifact interference level determination module is used to calculate the mean value of each column of the signal in the difference matrix, obtain the mean signal set, and determine the degree of motion artifact signal interference received by each group of photoelectric sensors based on the magnitude of each element in the mean signal set. The preliminary motion artifact removal module is used to filter the PPG signal according to the degree of interference of the motion artifact signal to obtain the pulse wave signal with preliminary motion artifact removal. The final motion artifact removal module is used to adaptively remove the pulse wave signal that is interfered with by the motion artifact signal from the pulse wave signal obtained by the initial motion artifact removal based on the set physiological index information and the characteristics of the motion artifact signal, so as to obtain the final motion artifact removed pulse wave signal.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the method for motion artifact removal based on differential PPG signals as described in any one of claims 1-3.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the method for removing motion artifacts based on differential PPG signals as described in any one of claims 1-3.

Images