Motion artifact removal method based on multiple photoelectric converters, medium and equipment

Abstract

The invention relates to a motion artifact removal method based on multiple photoelectric converters, a medium and equipment. The method comprises the following steps: symmetrically arranging a plurality of photoelectric converters, collecting photoelectric volume pulse wave signals and carrying out amplitude normalization; calculating a signal difference value of the photoelectric converter at the symmetrical position, and constructing a space displacement curve reflecting the relative displacement of the detection surface; judging whether motion disturbance exists in the signal based on amplitude change of the spatial displacement curve; if yes, correcting each photoelectric converter signal by using a space displacement curve, and removing motion artifacts; and finally, fusing the corrected signals to form artifact-removed output. The method can truly reflect the relative movement of the photoelectric converter and the skin, improves the artifact judgment and correction accuracy, reduces the calculation complexity, and is suitable for real-time processing of embedded equipment. The multi-PD symmetric layout forms a plurality of motion axes, so that the spatial displacement sensing dimension is enhanced, and the adaptability and robustness of a multi-direction motion scene are improved.

Description

Technical Field

[0001] The present application relates to, in particular to, a method, medium and device for removing motion artifacts based on multiple photoelectric converters. Background Art

[0002] Photoplethysmography (PPG) is a non-invasive physiological parameter monitoring technology that detects changes in blood flow based on optical principles. It is widely used in smart wearable devices for real-time monitoring of physiological indicators such as heart rate and blood oxygen. The acquisition of PPG signals relies on the photoelectric converter (PD) receiving the changes in the light signal of a specific wavelength emitted by the LED after it is reflected by the skin and blood vessels. However, in actual application scenarios, the relative displacement between the PD and the skin caused by the user's movement behavior can easily lead to the so-called "motion artifact" - that is, the low-frequency noise signal generated by inertia. This type of artifact has a great impact on the PPG signal, which will directly cause inaccurate measurement of physiological parameters and seriously reduce the reliability of wearable devices in dynamic scenarios.

[0003] Existing mainstream motion artifact suppression methods are mostly based on inertial sensors such as accelerometers and gyroscopes, combined with PPG signals for auxiliary analysis and filtering. However, accelerometers are fixedly mounted on the device and only detect the overall motion of the device, failing to accurately reflect the relative displacement between the PD and the skin. This significantly limits the accuracy of artifact modeling and removal. Summary of the Invention

[0004] The purpose of this application is to solve the above-mentioned problem of being unable to accurately identify and effectively eliminate motion artifacts caused by the relative displacement between the PD and the skin contact surface.

[0005] According to one aspect of the present application, a method for removing motion artifacts based on multiple photoelectric converters is provided, which is applied to a wearable device, comprising:

[0006] providing a plurality of photoelectric converters, wherein the plurality of photoelectric converters are symmetrically arranged around the center point of the surface to be measured;

[0007] Acquire photoplethysmography signals collected by multiple photoelectric converters, perform amplitude normalization processing on the multiple photoplethysmography signals, and generate a normalized photoplethysmography signal;

[0008] performing difference calculation on normalized photoplethysmography signals at symmetrical positions to obtain a plurality of difference signals;

[0009] Combining multiple difference signals to construct a spatial displacement curve, wherein the spatial displacement curve is used to reflect the change trend of the detection surface;

[0010] Based on the amplitude change of the spatial displacement curve, determine whether there is a disturbance in the photoplethysmography signal;

[0011] If yes, the photoplethysmography signal collected by each photoelectric converter is corrected according to the spatial displacement curve to remove motion artifacts;

[0012] The multiple corrected photoplethysmography signals are fused to form a fused photoplethysmography signal as a motion artifact removal result.

[0013] Preferably, the amplitude normalization process includes:

[0014] Select a photoelectric converter signal as a reference signal;

[0015] Calculating a scaling ratio of other photoelectric converter signals relative to the reference signal, the scaling ratio being a ratio of the reference signal amplitude range to the corresponding photoelectric converter signal amplitude range;

[0016] Each photoelectric converter signal is multiplied by its corresponding scaling ratio to make the amplitude range of all photoelectric converter signals uniform.

[0017] Preferably, constructing the spatial displacement curve includes:

[0018] Calculating the signal difference of at least two symmetrical photoelectric converter pairs in orthogonal directions as the axial displacement;

[0019] Perform square root operation on the sum of squares of each axial displacement to obtain the spatial displacement curve;

[0020] The ratio of the peak-to-peak amplitude of the spatial displacement curve to the peak-to-peak amplitude of the fused photoplethysmography signal was calculated as the displacement intensity value;

[0021] When the displacement intensity value exceeds a preset intensity threshold, the spatial displacement curve is subtracted from the fused photoplethysmography signal to filter out motion artifacts.

[0022] Preferably, the motion artifact removal method further includes a method for dynamically optimizing light source intensity, specifically comprising:

[0023] Determine whether the user is in motion based on accelerometer data;

[0024] If so, the light source intensity is gradually increased at fixed time intervals;

[0025] Calculating the correlation coefficient between the collected signal and the preset static reference signal at each light source intensity;

[0026] screening the collected signals whose correlation coefficient is lower than a first threshold;

[0027] The signal with the largest correlation coefficient with the current photoplethysmography signal is selected from the screening results, and the corresponding light source intensity is the optimal intensity.

[0028] Preferably, the motion artifact removal method further includes constructing reference heart rate information for correcting artifact signals in a manner based on a generative adversarial network model, specifically including:

[0029] Synthesize baseline heart rate information using a resting photoplethysmography signal and a moving heart rate belt signal;

[0030] Fusing accelerometer, gyroscope and magnetometer data to generate motion posture data in quaternion form;

[0031] The original photoplethysmography signal and motion posture data are converted into artifact-free signals through a generative network;

[0032] The artifact-free signal is compared with the baseline heart rate information through a discriminant network until they are indistinguishable.

[0033] Preferably, the correction process includes:

[0034] The spatial displacement curve is fused with the normalized photoplethysmography signal through weighted filtering to obtain a reference interference signal.

[0035] The original photoplethysmography signal is processed by minimum mean square error adaptive filtering using a reference interference signal to output a de-artifacted signal.

[0036] Preferably, it further comprises:

[0037] Perform time synchronization and phase alignment on multiple optoelectronic converter signals;

[0038] Use principal component analysis or independent component analysis to fuse multiple synchronized signals, extract the main pulse components, and enhance the signal-to-noise ratio;

[0039] The spatial displacement curve is combined to further optimize the fusion results and reduce the impact of motion interference.

[0040] Preferably, it further comprises:

[0041] Perform frequency domain feature analysis on the output artifact-free signal to extract the main frequency component and heart rate rhythm parameters;

[0042] Perform dynamic stability testing on the extracted results to determine whether the current signal meets the preset heart rate signal quality standards;

[0043] If the signal quality does not meet the standards, a prompt signal will be issued or a re-acquisition mechanism will be triggered.

[0044] The present invention also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to perform the steps of any one of the methods described above.

[0045] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of any one of the methods described above.

[0046] The present application has the following beneficial effects: It truly reflects the relative displacement changes between the PD and the skin: By symmetrically arranging multiple photoelectric converters, the motion axis is constructed using the signal differences at spatially symmetrical positions, and the spatial displacement curve is further calculated, effectively revealing the true source of motion interference in the PPG signal; it improves the accuracy of artifact determination and correction: normalization is used to unify the amplitude range of different PD signals, and the degree of artifact is determined by the displacement intensity (the ratio of the spatial displacement curve amplitude to the fused signal amplitude), avoiding the problem of artificial threshold mismatch in traditional methods; it achieves direct and effective artifact removal: the constructed spatial displacement curve is used to perform subtraction correction on the fused PPG signal, without the introduction of heterogeneous sensors or high-dimensional mapping, significantly reducing computational complexity and improving the real-time operation efficiency of embedded devices; it enhances the utilization and robustness of multiple PD signals: multiple symmetrical PDs are used to form multiple motion axes, increasing the spatial displacement perception dimension, making the artifact removal processing more adaptable and stable in multi-directional motion scenarios; it is applicable to different layouts and dynamic environments: the method of the present invention is adaptable to various PD layouts (such as 4, 6, or 8 PDs) and can flexibly adjust the artifact judgment threshold according to the application scenario, making the system have good versatility and dynamic adjustment capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the implementation methods of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the implementation methods or the description of the prior art. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 This is a logic block diagram of a motion artifact removal method based on multiple photoelectric converters according to one embodiment of the present application. DETAILED DESCRIPTION

[0049] To facilitate understanding of the present application, a more comprehensive description of the present application will be provided below with reference to the accompanying drawings. The accompanying drawings illustrate preferred embodiments of the present application. However, the present application may be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and comprehensive understanding of the disclosure of the present application.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are intended only to describe specific embodiments and are not intended to limit this application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0051] Please refer to Figure 1 An embodiment of the present application provides a method for removing motion artifacts based on multiple photoelectric converters, which is applied to wearable devices, including:

[0052] S10. Set up multiple photoelectric converters, and arrange the multiple photoelectric converters symmetrically around the center point of the surface to be measured. In this step, it should be noted that in order to improve the ability to identify and remove motion artifacts in optical signals, multiple photoelectric converters (PDs) need to be arranged in a centrally symmetrical manner in the optical detection structure. The center point is usually the position of the optical emission source (such as an LED) or the geometric center of the PPG signal detection area. Through the centrally symmetrical layout, any two PDs arranged symmetrically about the center point can form a set of "motion axes" for detecting relative displacement changes in the direction of the axis.

[0053] In a preferred embodiment, the number of PDs is an even number, preferably 4, 6, or 8, arranged symmetrically around a central point. For example, when using four PDs, PD1 and PD3 form a first motion axis symmetrically about the center, and PD2 and PD4 form a second motion axis, respectively used to detect relative motion interference in the vertical and horizontal directions. When using eight PDs, four more motion axes can be formed, including an oblique axis, thereby improving the spatial resolution and accuracy of motion artifacts.

[0054] This symmetrical arrangement not only ensures that each pair of PDs has similar optical path characteristics but also provides the data structure foundation for subsequently constructing motion artifact spatial displacement curves through signal differences. Furthermore, the PDs are preferably of the same model, sensitivity, and response bandwidth to ensure consistency and comparability of the detection data.

[0055] S20. Acquire the photoplethysmographic signals collected by multiple photoelectric converters, perform amplitude normalization on the multiple photoplethysmographic signals, and generate a normalized photoplethysmographic signal. It should be noted that in this step, multiple photoelectric converters (PDs) synchronously collect photoplethysmographic signals (PPG signals) at corresponding positions. The signals collected by each PD may have amplitude deviations due to factors such as differences in optical paths, light source illumination angles, and skin reflectivity. To ensure the accuracy of subsequent difference calculations, all PPG signals need to be uniformly amplitude normalized.

[0056] The normalization method includes but is not limited to the following steps: select the signal of one PD as the reference signal (for example, PD1), calculate the scaling ratio of the other PD signals relative to the amplitude range of the reference signal, and unify their signal amplitudes to the amplitude range of the reference signal. Specifically, it can be expressed as:

[0057] pd_y_i_norm=pd_y_i×(pd_y_ref_max-pd_y_ref_min)÷(pd_y_i_max-pd_y_i_min)

[0058] Among them, pd_y_i is the original signal of the i-th PD, pd_y_ref is the reference PD signal, and pd_y_i_norm is the normalized signal.

[0059] S30. Perform difference calculation on the normalized photoplethysmography signals at symmetrical positions to obtain multiple difference signals. In this step, it should be noted that a pair of PD normalized signals at central symmetrical positions are subjected to point-to-point difference calculations. Each pair of PDs corresponds to a "motion axis", and the difference signal represents the relative displacement change in the direction of the axis. The larger the absolute value of the difference, the more obvious the motion disturbance in this direction. The difference is calculated as follows:

[0060] Move_axis_i=pd_y_m_norm-pd_y_n_norm

[0061] Wherein, m and n are a pair of PDs symmetrical about the center.

[0062] S40. Combine multiple difference signals to construct a spatial displacement curve, which is used to reflect the change trend of the detection surface. In this step, it should be noted that multiple axial difference signals are synthesized and the spatial displacement curve is calculated using the square root of the sum of squares (Euclidean norm) to reflect the relative displacement amplitude of the entire detection surface in space. The expression is:

[0063] Move_res=sqrt(∑(Move_axis_i 2 ))

[0064] The spatial displacement curve can be regarded as an indicator reflecting the overall stability and relative motion state of the PD array, and can intuitively represent the fluctuation trend of the motion artifact on the time axis.

[0065] S50: Determine whether there is disturbance in the photoplethysmography signal based on the amplitude change of the spatial displacement curve. In this step, it should be noted that the peak-to-peak amplitude change of the spatial displacement curve is analyzed and proportional to the amplitude of the normalized fusion signal to define the disturbance intensity index:

[0066] Move_strength=(Move_res_max-Move_res_min) / (PPG_fusion_max-PPG_fusion_min)

[0067] If Move_strength is less than a preset threshold (such as 0.1), it is considered that the current signal is not significantly disturbed; if it exceeds the threshold, it is considered that the motion artifact interference is large and needs to be corrected.

[0068] S60: If yes, then each photoplethysmography signal is corrected based on the spatial displacement curve to remove motion artifacts. It should be noted that when disturbance (i.e., motion artifact) is detected, the core of this step is to use the spatial displacement curve to infer and deduct the motion interference component, and to correct each normalized photoplethysmography (PPG) signal collected by the photoelectric converter one by one.

[0069] The spatial displacement curve is constructed based on the differences between the symmetrical PD signals and accurately reflects the magnitude and direction of the relative displacement between the sensor detection surface and the skin contact surface. By using the spatial displacement curve as an estimate of motion disturbance and subtracting this interference component from each PD channel signal, an artifact-free PPG signal can be obtained.

[0070] The significance of this step is to restore each PD channel to a pure signal that is not affected by motion, laying the foundation for subsequent signal fusion.

[0071] S70. Fuse the multiple corrected photoplethysmography signals to form a fused photoplethysmography signal as the motion artifact removal result. It should be noted that after completing artifact removal for each PD channel, this step further fuses the PPG signals of multiple "clean channels." This fusion operation is used to improve the overall signal quality and suppress abnormal fluctuations in individual PDs caused by local factors (such as poor fit, differences in local tissue characteristics, temporary occlusion, etc.).

[0072] Fusion methods may include simple averaging, multi-channel weighted averaging, principal component analysis (PCA), etc. Its core goal is to enhance the common effective pulse components between channels and reduce the impact of noise and individual channel abnormalities.

[0073] The final output fused photoplethysmography signal, as the result of motion artifact removal, will be input into the subsequent health parameter analysis module as a reliable physiological data source, such as heart rate, blood oxygen saturation, etc.

[0074] In a specific embodiment, the amplitude normalization process in S20 includes:

[0075] S21. Select a photoelectric converter signal as a reference signal.

[0076] Preferably, a signal from a photoelectric converter arranged at any fixed position in the central symmetrical structure is selected as a reference, such as the PD1 channel, and is used to standardize the amplitude range of other channels.

[0077] S22 . Calculate the scaling ratio of other photoelectric converter signals relative to the reference signal, where the scaling ratio is the ratio of the reference signal amplitude range to the corresponding photoelectric converter signal amplitude range.

[0078] The scaling ratio is the ratio between the amplitude range of the reference signal and the amplitude range of the target photoelectric converter signal, and is specifically calculated as follows:

[0079]

[0080] Among them, PD ref Indicates the reference photoelectric signal, PD i represents the i-th photoelectric converter signal to be normalized.

[0081] S23. Multiply each photoelectric converter signal by its corresponding scaling factor to unify the amplitude range of all photoelectric converter signals. This scaling process achieves consistency in the amplitude dimension of the PPG signals across multiple channels, effectively eliminating light intensity differences caused by individual PD differences (e.g., position, contact state, and sensitivity), facilitating subsequent difference calculations and spatial displacement modeling.

[0082] The construction of spatial displacement curves in S40 includes:

[0083] S41. Calculate the signal difference between at least two orthogonal pairs of photoelectric converters as the axial displacement. Multiple photoelectric converters are symmetrically distributed around the center point of the surface to be measured. Two PDs at symmetrical positions are selected to form a "motion axis." For example, PD1 and PD3 form the X-axis, and PD2 and PD4 form the Y-axis. For each symmetrical PD channel, calculate the difference in its normalized signal as the corresponding axial displacement:

[0084] Movex=PD1 norm -PD3 norm

[0085] Movey=PD2 norm -PD4 norm

[0086] S42. Perform square root operation on the sum of the squares of the axial displacements to obtain a spatial displacement curve.

[0087] The displacements in each direction are synthesized to construct the total spatial displacement curve:

[0088]

[0089] This curve reflects the relative displacement change trend of the detection surface in the time dimension and can capture the temporal pattern generated by motion artifacts.

[0090] S43. Calculate the ratio of the peak-to-peak amplitude of the spatial displacement curve to the peak-to-peak amplitude of the fused photoplethysmography signal as the displacement intensity value.

[0091] The amplitude of the spatial displacement curve is calculated as the difference between the maximum and minimum values:

[0092] Move_res_amp=max(Move_res)-min(Move_res)

[0093] The fused photoplethysmography signal is the average of the normalized signals of multiple PD channels, and its peak-to-peak amplitude is:

[0094] PPG_amp=max(PPGfusion)-min(PPGfusion)

[0095] The calculation formula of displacement strength value is:

[0096]

[0097] S44. When the displacement intensity value exceeds a preset intensity threshold, subtract the spatial displacement curve from the fused photoplethysmography signal to filter out motion artifacts.

[0098] Set the intensity threshold T when:

[0099] Move_strength≥T

[0100] If significant motion artifacts exist, the fusion signal is corrected:

[0101] PPG filtered =PPG fusion -Move_resPPG

[0102] This operation essentially removes the common-mode interference component introduced by the displacement from the fused signal, thereby improving the PPG signal quality and functional measurement accuracy.

[0103] Furthermore, the motion artifact removal method also includes a method for dynamically optimizing the light source intensity, specifically including:

[0104] Determine whether the user is in motion based on accelerometer data: This function uses the device's built-in three-axis or six-axis accelerometer to obtain current acceleration information and analyze its fluctuations within a set time window. If the acceleration fluctuation exceeds a preset threshold, the user is considered in motion.

[0105] If so, gradually increase the light source intensity at fixed time intervals: set the light source intensity adjustment range and incremental step size, for example, starting from the initial intensity I0, gradually increase the light source intensity to the maximum allowable value, sample one cycle at each level, and obtain the corresponding photoplethysmography signal.

[0106] Calculate the correlation coefficient between the collected signal and the preset static reference signal at each light source intensity: Perform correlation analysis on the PPG signals collected under different light source intensities and a set of pre-stored static high-quality PPG signals (such as those collected in a resting state), and use indicators such as the Pearson correlation coefficient and cross-correlation function to evaluate the degree of matching:

[0107] Corr i =Correlation(PPG i ,PPG ref )

[0108] Among them PPG i is the signal under the intensity of the i-th light source.

[0109] Filter the collected signals with correlation coefficients lower than the first threshold: By setting the correlation threshold T1T_1T1, the light source positions with heavy artifact interference that are obviously mismatched with the reference signal are eliminated:

[0110] If Corr i <T1, then discard the signal

[0111] The signal with the largest correlation coefficient with the current photoplethysmography signal is selected from the screening results, and the corresponding light source intensity is the optimal intensity.

[0112] Assume that the fusion signal is PPG fusion PPG, select the one with the largest correlation as the optimal strength:

[0113] I optimal =argmaxCorrelation(PPG i ,PPG fusion )

[0114] The light source intensity can minimize the impact of motion artifacts in the current motion state, enhance the signal-to-noise ratio, and facilitate subsequent signal fusion and health indicator calculation.

[0115] Furthermore, the motion artifact removal method also includes constructing reference heart rate information based on a generative adversarial network model for correcting artifact signals, specifically including:

[0116] Use the resting photoplethysmography signal and the exercise heart rate belt signal to synthesize the baseline heart rate information:

[0117] The system collects high-quality photoplethysmography signals (PPG static signals) of users at rest and reference heart rate data obtained by a heart rate belt during exercise, and performs time alignment and fusion processing on them to construct baseline heart rate information as a training target. This information reflects the real and effective heart rate characteristics under different exercise states.

[0118] Fusing accelerometer, gyroscope and magnetometer data to generate motion attitude data in quaternion form:

[0119] The raw inertial data is obtained through the three-axis accelerometer, gyroscope and magnetometer carried by the wearable device, and further integrated into motion posture data in the form of quaternions to accurately describe the dynamic motion state of the detection object and provide spatial posture feature input for neural network modeling.

[0120] The original photoplethysmography signal and motion posture data are converted into artifact-free signals through the generative network:

[0121] The original disturbed photoplethysmography signal (i.e., the PPG signal containing motion artifacts) and the synchronously acquired motion posture data are input into the generator network (Generator). The generator network outputs the PPG signal after artifact removal based on a deep neural structure (such as LSTM+Transformer, etc.).

[0122] The discriminant network compares the artifact-free signal with the baseline heart rate information until they are indistinguishable. Through continuous training of the adversarial process between the generative network and the discriminant network, the generative network gradually learns the ability to recover the true components of the heart rate from the interference signal under different motion states, ultimately achieving the goal of removing motion artifacts and reconstructing accurate PPG signals.

[0123] Furthermore, the correction process in S60 includes:

[0124] S61. The spatial displacement curve is fused with the normalized photoplethysmography signal through weighted filtering to obtain a reference interference signal:

[0125] The spatial displacement curve is considered a characteristic quantity reflecting the intensity of motion artifacts. It is then fused with each normalized photoplethysmography signal through weighted filtering to construct a reference interference signal with time synchronization and trend consistency. The key to this step is to match the perturbation trends of the PPG signals in different channels to obtain an interference estimate that better reflects the characteristics of the actual artifact.

[0126] S62, using the reference interference signal to perform minimum mean square error adaptive filtering on the original photoplethysmography signal, and outputting a de-artifacted signal:

[0127] The least mean square error (LMS) adaptive filtering algorithm is used. The reference interference signal generated in S61 is used as the interference input, and the original photoplethysmography signal is used as the target signal. By iteratively adjusting the filter weights, the error between the output signal and the expected heart rate signal is dynamically minimized, thereby outputting a high-quality artifact-free signal.

[0128] In an optional embodiment, the motion artifact removal method further comprises:

[0129] Time synchronization and phase alignment of multiple optical / electrical converter signals:

[0130] The photoplethysmography signals collected by multiple photoelectric converters are processed synchronously on the time axis to ensure that the signals of each channel are compared and calculated under the same time reference. At the same time, phase alignment technology is used to eliminate phase offset problems caused by device response differences or sampling delays, thereby improving the consistency of data fusion.

[0131] Use principal component analysis or independent component analysis to fuse multiple synchronized signals, extract the main pulse components, and enhance the signal-to-noise ratio:

[0132] Multiple synchronized, normalized photoplethysmography signals are fed into a principal component analysis (PCA) or independent component analysis (ICA) model to extract the principal components or independent source signals that are highly correlated with pulse activity. This process effectively reduces redundant interference between channels and improves the signal-to-noise ratio of the target pulse signal.

[0133] Combined with the spatial displacement curve, the fusion results are further optimized to reduce the impact of motion interference:

[0134] The motion disturbance trend represented by the spatial displacement curve constructed above is used to further correct the PCA / ICA fused signal. By identifying and suppressing components with the same frequency or common mode as the motion artifact, more precise artifact removal is achieved, thereby improving the stability and accuracy of the fused signal.

[0135] In an optional embodiment, the motion artifact removal method further comprises:

[0136] Perform frequency domain feature analysis on the output artifact-free signal to extract the main frequency component and heart rate rhythm parameters:

[0137] The output de-artifacted signal is transformed in the frequency domain to extract the main frequency component of the signal and its corresponding parameters such as amplitude and frequency stability. At the same time, heart rate rhythm indicators (such as RR interval fluctuation, frequency domain heart rate variability, etc.) are calculated to evaluate the physiological relevance of the signal.

[0138] Perform dynamic stability testing on the extracted results to determine whether the current signal meets the preset heart rate signal quality standards:

[0139] Based on the frequency domain analysis results, the dynamic stability of the extracted heart rate frequency components is judged, and their frequency drift, amplitude fluctuation and noise ratio within a given time window are evaluated to determine whether the current signal meets the preset heart rate signal quality standards, such as the frequency stability threshold and the signal-to-noise ratio lower limit.

[0140] If the signal quality does not meet the standards, a prompt signal will be issued or a re-acquisition mechanism will be triggered:

[0141] If the test results show that the current signal quality does not meet the standards, the feedback mechanism will be triggered to issue a signal quality prompt (such as prompting unstable wearing, re-collection, etc.), or automatically start the re-collection process to ensure the reliability and accuracy of subsequent physiological parameter calculations.

[0142] This specific embodiment also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0143] This specific embodiment also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the above method.

[0144] The above-described embodiments merely represent several embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A motion artifact removal method based on multiple photoelectric converters, applied to wearable devices, characterized in that: include: providing a plurality of photoelectric converters, wherein the plurality of photoelectric converters are symmetrically arranged around the center point of the surface to be measured; Acquire photoplethysmography signals collected by multiple photoelectric converters, perform amplitude normalization processing on the multiple photoplethysmography signals, and generate a normalized photoplethysmography signal; performing difference calculation on normalized photoplethysmography signals at symmetrical positions to obtain a plurality of difference signals; Combining multiple difference signals to construct a spatial displacement curve, wherein the spatial displacement curve is used to reflect the change trend of the detection surface; Based on the amplitude change of the spatial displacement curve, determine whether there is a disturbance in the photoplethysmography signal; If yes, the photoplethysmography signal collected by each photoelectric converter is corrected according to the spatial displacement curve to remove motion artifacts; The multiple corrected photoplethysmography signals are fused to form a fused photoplethysmography signal as a motion artifact removal result.

2. The motion artifact removal method according to claim 1, wherein: The amplitude normalization process includes: Select a photoelectric converter signal as a reference signal; Calculating a scaling ratio of other photoelectric converter signals relative to the reference signal, the scaling ratio being a ratio of the reference signal amplitude range to the corresponding photoelectric converter signal amplitude range; Each photoelectric converter signal is multiplied by its corresponding scaling ratio to make the amplitude range of all photoelectric converter signals uniform.

3. The motion artifact removal method according to claim 1, wherein: The constructing of the spatial displacement curve comprises: Calculating the signal difference of at least two symmetrical photoelectric converter pairs in orthogonal directions as the axial displacement; Perform square root operation on the sum of squares of each axial displacement to obtain the spatial displacement curve; The ratio of the peak-to-peak amplitude of the spatial displacement curve to the peak-to-peak amplitude of the fused photoplethysmography signal was calculated as the displacement intensity value; When the displacement intensity value exceeds a preset intensity threshold, the spatial displacement curve is subtracted from the fused photoplethysmography signal to filter out motion artifacts.

4. The motion artifact removal method according to claim 1, wherein: The motion artifact removal method further includes a method for dynamically optimizing light source intensity, specifically comprising: Determine whether the user is in motion based on accelerometer data; If so, the light source intensity is gradually increased at fixed time intervals; Calculating the correlation coefficient between the collected signal and the preset static reference signal at each light source intensity; screening the collected signals whose correlation coefficient is lower than a first threshold; The signal with the largest correlation coefficient with the current photoplethysmography signal is selected from the screening results, and the corresponding light source intensity is the optimal intensity.

5. The motion artifact removal method according to claim 1, wherein: The motion artifact removal method further includes constructing reference heart rate information based on a generative adversarial network model for correcting artifact signals, specifically including: Synthesize baseline heart rate information using a resting photoplethysmography signal and a moving heart rate belt signal; Fusing accelerometer, gyroscope and magnetometer data to generate motion posture data in quaternion form; The original photoplethysmography signal and motion posture data are converted into artifact-free signals through a generative network; The artifact-free signal is compared with the baseline heart rate information through a discriminant network until they are indistinguishable.

6. The motion artifact removal method according to claim 1, characterized in that: The correction process includes: The spatial displacement curve is fused with the normalized photoplethysmography signal through weighted filtering to obtain a reference interference signal. The original photoplethysmography signal is processed by minimum mean square error adaptive filtering using a reference interference signal to output a de-artifacted signal.

7. The motion artifact removal method according to claim 1, characterized in that: Further including: Perform time synchronization and phase alignment on multiple optoelectronic converter signals; Use principal component analysis or independent component analysis to fuse multiple synchronized signals, extract the main pulse components, and enhance the signal-to-noise ratio; The spatial displacement curve is combined to further optimize the fusion results and reduce the impact of motion interference.

8. The motion artifact removal method according to claim 1, wherein: Further including: Perform frequency domain feature analysis on the output artifact-free signal to extract the main frequency component and heart rate rhythm parameters; Perform dynamic stability testing on the extracted results to determine whether the current signal meets the preset heart rate signal quality standards; If the signal quality does not meet the standards, a prompt signal will be issued or a re-acquisition mechanism will be triggered.

9. A computer-readable storage medium, characterized in that A computer program is stored, and when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 8.

10. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 8.

Images