An artifact-eliminated motion heart rate monitoring method and monitoring device

Abstract

The application discloses an artifact-eliminated motion heart rate monitoring method and a monitoring device, and relates to the technical field of health monitoring.The method comprises the following steps: acquiring motion data and a photoplethysmogram (PPG) signal in real time; generating an artifact feature vector according to current motion data, and matching the artifact feature vector with a preset reference artifact feature vector library to quickly determine a transfer function model that is optimally matched with a current motion mode; inputting real-time motion data into the optimal model to generate a simulated artifact signal; and finally, eliminating the simulated artifact signal from the real-time PPG signal, and performing filtering optimization on the processed signal to obtain a final heart rate. By independently predicting artifacts based on motion data and performing model matching, the application avoids complex real-time processing of contaminated signals, and realizes high real-time, high accuracy and high robustness of heart rate monitoring under various complex motion conditions.

Description

Technical Field

[0001] This application relates to the field of health monitoring technology, and in particular to a method and device for monitoring exercise heart rate with artifact elimination. Background Technology

[0002] With the rapid development of smart wearable devices, heart rate monitoring technology based on photoplethysmography (PPG) signals has been widely applied in the field of sports and health. PPG signals reflect human heart rate by detecting minute changes in blood flow, making it a non-invasive and portable method of heart rate measurement. However, during user exercise, PPG signals are often severely affected by motion artifacts due to body vibration, movement, and posture changes, leading to a significant decrease in the accuracy of heart rate monitoring results. Therefore, how to effectively eliminate artifact interference and accurately extract heart rate information in complex motion environments has become a research hotspot in the field of smart wearable devices.

[0003] In related technologies, commonly used motion artifact removal methods typically involve acquiring the PPG signal and then progressively analyzing and removing motion artifacts generated within it. While this method can ensure accuracy in processing motion artifacts in PPG signals, performing complex real-time analysis and processing on PPG signals already affected by artifacts can lead to signal processing delays, resulting in insufficient real-time performance. Summary of the Invention

[0004] This application provides a motion heart rate monitoring method and monitoring device with artifact elimination, which reduces motion artifacts generated by the monitoring device when monitoring the user's PPG signal, thereby improving the real-time performance of heart rate monitoring while ensuring the accuracy of heart rate monitoring.

[0005] In a first aspect, this application provides a method for motion heart rate monitoring with artifact elimination, applied to a monitoring device. The method includes: acquiring current motion data of a target user, the motion data including acceleration signals and angular velocity signals; determining the current motion pattern of the target user based on the motion data, and performing time-frequency domain analysis on the motion data to generate an artifact feature vector, the artifact feature vector being a data vector containing the vibration frequency and statistical characteristics of the current motion; determining a target feature transfer function mapping table from a preset set of feature transfer function mapping tables based on the motion pattern, the target feature transfer function mapping table containing a reference artifact feature vector and a transfer function model corresponding to the motion pattern; and matching the artifact feature vector with the target feature transfer function mapping table to obtain a matching artifact feature vector. An optimal transfer function model is used, which is a mathematical operator for converting motion data into a noise waveform. The motion data is input into this optimal transfer function model for processing to obtain a simulated artifact signal, which is used to simulate the real artifact signal that will be generated. The PPG signal of the target user is acquired in real time, and the simulated artifact signal is removed from the PPG signal by a waveform fitting method to obtain the initial heart rate signal, which is a photoplethysmography (PPG) signal. The successive heartbeat interval of the initial heart rate signal is extracted by a peak detection algorithm. The successive heartbeat interval is the time length between two adjacent valid heartbeat peaks of the target user. Based on a preset successive heartbeat interval threshold, the successive heartbeat interval is input into a Kalman filter to obtain the final heart rate signal.

[0006] By adopting the above technical solution, this method does not directly process the interfered PPG signal. Instead, it first identifies the current motion pattern through motion data and quickly matches the optimal transfer function model from a preset mapping table set based on the motion pattern. Therefore, it can generate a signal in advance and independently to simulate real artifacts. This simulated artifact signal is then used to remove interference from the real-time PPG signal, effectively solving the problems of signal processing delay and poor real-time performance caused by complex analysis of PPG signals already contaminated by artifact signals in related technologies. Thus, it achieves both accurate artifact elimination and effective improvement in the real-time response speed of heart rate monitoring in complex motion environments, providing users with immediate and reliable heart rate data feedback.

[0007] In conjunction with some embodiments of the first aspect, in some embodiments, the artifact feature vector is matched with the target feature transfer function mapping table to obtain the transfer function model that best matches the artifact feature vector. Specifically, this includes: calculating the Euclidean distance between the artifact feature vector and each of the reference artifact feature vectors, and determining the reference artifact feature vector corresponding to the smallest Euclidean distance as the target feature vector; matching the target feature vector with the transfer function model corresponding to the target feature transfer function mapping table to obtain the optimal transfer function model that matches the artifact feature vector.

[0008] By employing the above technical solution, the Euclidean distance between the artifact feature vector generated from the current motion data and the preset reference artifact feature vector can be calculated, enabling a quantitative matching of the current motion state with the motion states stored in the database. Selecting the reference vector with the smallest Euclidean distance means finding the "template" that best matches the current motion artifact features. This method can accurately and quickly identify the optimal transfer function model, avoiding the ambiguity and subjectivity of model selection. It effectively solves the problem of excessive deviation between simulated and real artifacts due to improper model selection, thereby significantly improving the targeting and accuracy of subsequent artifact removal.

[0009] In conjunction with some embodiments of the first aspect, in some embodiments, the real-time acquisition of the target user's PPG signal, and the removal of the simulated artifact signal from the PPG signal using a waveform fitting method to obtain the initial heart rate signal, specifically includes: real-time acquisition of the target user's PPG signal and preprocessing the PPG signal; segmenting and aligning the simulated artifact signal with the preprocessed PPG signal, extracting the corresponding first signal segment using a time window sliding method, and calculating the fitting parameters of the first signal segment using a waveform fitting method, which is a least squares algorithm or a signal correlation analysis method; adjusting the amplitude and phase of the simulated artifact signal according to the fitting parameters, the adjustment including fitting the simulated artifact signal with the artifact portion of the PPG signal corresponding to the first signal segment; subtracting the adjusted simulated artifact signal point by point from the preprocessed PPG signal, and smoothing the PPG signal after artifact removal to obtain the initial heart rate signal.

[0010] By employing the above technical solution, and aligning the simulated artifact signal with the PPG signal in segments, and calculating fitting parameters such as amplitude and phase using waveform fitting methods such as least squares, the initially generated simulated artifact signal can be dynamically adjusted with fine precision. This method does not perform simple rigid subtraction; instead, it ensures that the amplitude and temporal phase of the simulated artifact are highly fitted to the actual artifact portion in the PPG signal before removal. This "tailor-made" adjustment effectively solves the problem of incomplete matching between simulated and real artifacts caused by inherent model errors or sudden changes in motion state, thereby achieving higher precision artifact removal and minimizing artifact residue and misinterpretation of the real heart rate signal.

[0011] In conjunction with some embodiments of the first aspect, in some embodiments, the adjusted simulated artifact signal is subtracted point by point from the preprocessed PPG signal, and the artifact-free PPG signal is smoothed to obtain the initial heart rate signal. Specifically, this includes: mapping the adjusted simulated artifact signal to the preprocessed PPG signal point by point according to the time series; subtracting the corresponding adjusted simulated artifact signal value from the PPG signal value at each time point, and truncating the subtraction result based on a fixed threshold to obtain the artifact-free PPG signal, where the fixed threshold is used to limit the amplitude range of artifact removal to avoid signal distortion; and processing the artifact-free PPG signal using a smoothing algorithm to obtain the initial heart rate signal.

[0012] By employing the above technical solution, the adjusted simulated artifact signal is subtracted point-by-point from the PPG signal, thus ensuring the precise temporal correspondence of the artifact removal operation. Furthermore, a truncation process based on a fixed threshold is introduced, which limits the magnitude of artifact removal. While accurately eliminating artifacts, this approach also proactively avoids the risk of over-removal due to model errors or nonlinear interference, effectively solving the problem that traditional signal subtraction methods may introduce new signal distortions or warp the true heart rate waveform. This ensures the robustness of the artifact removal process and protects the physiological integrity of the original heart rate signal.

[0013] In conjunction with some embodiments of the first aspect, in some embodiments, after the steps of acquiring the target user's PPG signal in real time, removing the analog artifact signal from the PPG signal using a waveform fitting method to obtain the initial heart rate signal, the method further includes: performing autocorrelation analysis on the initial heart rate signal to obtain the autocorrelation function of the initial heart rate signal; extracting the peak value at the non-zero delay in the autocorrelation function, and determining whether the peak value corresponds to the periodic characteristics of the initial heart rate signal, the periodic characteristics including the dominant frequency, spectral energy ratio, and peak amplitude; if they do not correspond, it is determined that the initial heart rate signal does not conform to periodicity, and the parameters corresponding to the transfer function model are adjusted.

[0014] By employing the above technical solution, autocorrelation analysis of the initial heart rate signal after artifact removal allows for an objective assessment of the signal's periodicity, thus determining the effectiveness of artifact removal. If the periodicity is not met, the parameters of the transfer function model are deemed to have errors, and feedback is provided to adjust these parameters. This constitutes a closed-loop control system of "removal-verification-feedback-optimization." Therefore, this method transforms artifact removal from a unidirectional, fixed process into one with self-learning and adaptive adjustment capabilities. It effectively solves the problems of poor adaptability and unstable performance of traditional methods in complex and ever-changing motion scenarios, thereby achieving continuous optimization of the transfer function model and effectively improving the overall robustness and accuracy of the system.

[0015] In conjunction with some embodiments of the first aspect, in some embodiments, after the step of adjusting the parameters of the corresponding transfer function model if the initial heart rate signal does not correspond to the periodicity, the method further includes: acquiring a second signal segment in the initial heart rate signal that does not correspond to the periodicity, and extracting the time-domain features and frequency-domain features of the second signal segment, wherein the time-domain features include waveform amplitude and periodicity trend, and the frequency-domain features include the position of the dominant frequency and the spectral energy distribution; if the dominant frequency of the second signal segment exceeds a preset frequency band range, then bandpass filtering is performed on the second signal segment to retain only frequencies within the preset frequency band range; if the waveform amplitude of the second signal segment is abnormal, then the second signal segment is normalized based on the waveform amplitude of adjacent periodic signal segments; if there is a periodic interruption in the second signal segment, then the missing or abnormal waveform in the second signal segment is processed by an interpolation algorithm based on the periodicity trend of adjacent signal segments.

[0016] By employing the above technical solution, when non-periodic segments are detected in the initial heart rate signal, their time and frequency domain features are further extracted, and targeted correction measures, such as bandpass filtering, normalization, or interpolation, are taken based on these features. Therefore, this method does not simply discard these poor-quality data segments, but attempts to repair and compensate for them, effectively solving the problem of heart rate data loss and signal discontinuity caused by brief strong interference. This maximizes the preservation of signal integrity, providing more continuous and reliable input data for subsequent Kalman filtering, and improving the stability of the final heart rate output.

[0017] In conjunction with some embodiments of the first aspect, in some embodiments, the step-by-step heartbeat interval, based on a preset threshold, is input into a Kalman filter to obtain a final heart rate signal. Specifically, this includes: obtaining a preset threshold range for the step-by-step heartbeat interval, which is dynamically adjusted based on the target user's age, exercise mode, and historical heart rate data; inputting the step-by-step heartbeat interval into a Kalman filter to obtain a measured value and a predicted value for the step-by-step heartbeat interval; when the measured value exceeds the preset threshold range, replacing the measured value with the predicted value; and recursively updating the Kalman filter to smooth the step-by-step heartbeat interval signal to obtain the final heart rate signal.

[0018] By employing the aforementioned technical solution, the successive heart rate intervals are processed using a Kalman filter, combined with a preset threshold that is dynamically adjusted based on the user's age, exercise pattern, and other factors. When the measured value exceeds this dynamic threshold, a more reliable predicted value is used instead. Therefore, this method not only utilizes the smoothing and noise reduction capabilities of the Kalman filter itself but also improves its accuracy in identifying and processing outliers through intelligent dynamic thresholds. This effectively solves the problem of interference from isolated, physiologically inaccurate jump points caused by residual artifacts or noise on the final result, thereby achieving deep optimization of the heart rate signal and outputting extremely smooth, stable, and highly reliable final heart rate data.

[0019] In a second aspect, embodiments of this application provide a monitoring device, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the monitoring device to perform the method described in the first aspect and any possible implementation thereof.

[0020] Thirdly, embodiments of this application provide a computer program product containing instructions that, when the computer program product is run on a monitoring device, cause the monitoring device to perform the method described in the first aspect and any possible implementation thereof.

[0021] Fourthly, embodiments of this application provide a computer-readable storage medium including instructions that, when executed on a monitoring device, cause the monitoring device to perform the method described in the first aspect and any possible implementation thereof.

[0022] Understandably, the monitoring equipment provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the methods provided in the embodiments of this application. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] 1. By employing a method that predicts artifacts based on motion data and generates simulated artifact signals in advance, and matching the artifact feature vectors generated from real-time motion data with a preset feature transfer function mapping table corresponding to the motion mode, the optimal transfer function model can be quickly determined. Therefore, when the monitoring device acquires the interfered PPG signal, it can almost simultaneously obtain a simulated artifact signal that is highly similar to the real artifact. This effectively solves the problem in related technologies that require complex analysis of the interfered PPG signal, resulting in processing delays and insufficient real-time performance. Thus, without sacrificing the accuracy of artifact elimination, the real-time performance of exercise heart rate monitoring is significantly improved, providing users with more immediate and accurate feedback on their exercise physiological status.

[0025] 2. By employing the method of calculating the Euclidean distance between the artifact feature vector generated from the current motion data and the preset reference artifact feature vector, and selecting the transfer function model corresponding to the minimum distance, it is possible to accurately and quantitatively evaluate which typical state in the database is most similar to the current motion state, and precisely lock in the optimal model. This effectively solves the problem of excessive deviation between simulated artifacts and real artifacts and poor subsequent removal effect caused by fuzzy or mismatched model selection. In this way, personalized and high-precision modeling of motion artifacts is achieved, laying a solid foundation for the final generation of high-fidelity heart rate signals.

[0026] 3. By employing a recursive update method combining Kalman filters and dynamic threshold processing, this threshold can be adjusted in real time based on personalized information such as the user's age and exercise mode. When the measured value exceeds the dynamic threshold, it is replaced with a more reliable predicted value. Therefore, this method can intelligently and accurately identify and process abnormal heart rate interval data that do not conform to physiological laws due to residual artifacts. It effectively solves the problems of poor adaptability of fixed thresholds and the pollution of the overall heart rate calculation results by a single abnormal data point. This achieves deep optimization and smoothing of the heart rate signal and outputs extremely stable and highly reliable final heart rate data. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of a scenario for an exercise heart rate monitoring method with artifact elimination in an embodiment of this application;

[0028] Figure 2 A flowchart illustrating the motion heart rate monitoring method for artifact elimination in this application embodiment;

[0029] Figure 3 This is another flowchart illustrating the motion heart rate monitoring method for artifact elimination in the embodiments of this application;

[0030] Figure 4 This is a schematic diagram of the physical device structure of a monitoring device in the embodiments of this application. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to any or all possible combinations including one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] To facilitate understanding, the application scenarios of the embodiments of this application are described below. For example... Figure 1 The image shown is a schematic diagram of a scenario for an exercise heart rate monitoring method for artifact elimination in an embodiment of this application.

[0034] In related technologies, motion artifacts can be separated from heart rate signals by performing direct post-processing (such as adaptive filtering and independent component analysis) on the PPG signal acquired by the monitoring device 101, which is already mixed with motion artifacts. The following describes a scenario using the motion heart rate monitoring method with artifact elimination from related technologies. When a user wears the monitoring device 101 for exercise such as running, the monitoring device 101 acquires a severely interfered PPG signal. At this time, the processor inside the monitoring device 101 begins to process this complex mixed signal, attempting to separate the artifacts. This process is time-consuming; therefore, the heart rate value seen by the user on the screen may reflect their physiological state several seconds prior, exhibiting a significant delay.

[0035] The artifact prediction and real-time removal method in this application involves parallel acquisition of motion data and PPG signals. A simulated artifact signal highly similar to the real artifact is predicted and generated using the motion data, and then synchronously removed from the PPG signal. This achieves real-time and accurate monitoring of exercise heart rate, ensuring not only the accuracy of the monitoring results but, more importantly, significantly reducing data processing latency. The following describes a scenario using the artifact removal exercise heart rate monitoring method described in this application. While the user is running, the monitoring device 101 acquires PPG signals while simultaneously acquiring motion data using sensors such as an accelerometer. Based on the motion data, the processing unit in the monitoring device 101 quickly generates the corresponding simulated artifact signal. When the PPG signal enters the processor, the simulated artifact is also ready; the two are immediately subtracted to obtain a pure heart rate signal. Therefore, the heart rate value seen by the user on the screen is a true reflection of their current physiological state.

[0036] As can be seen, by adopting the artifact prediction and real-time elimination technology in the embodiments of this application, high-precision exercise heart rate monitoring can be achieved, while effectively solving the problems of monitoring result delay and poor real-time performance caused by complex signal processing in related technologies. This enables a high degree of synchronization between the monitoring device and the user's physiological state, providing the user with truly immediate and reliable exercise status feedback.

[0037] To facilitate understanding, the method provided in this implementation will be described in detail below, using the above scenario as an example. Please refer to [link / reference]. Figure 2 This is a flowchart illustrating a motion heart rate monitoring method for artifact elimination in an embodiment of this application.

[0038] S201. Obtain the current motion data of the target user, which includes acceleration signals and angular velocity signals;

[0039] Among them, motion data refers to a set of quantitative information that can reflect the motion state of the target user, including at least acceleration signals and angular velocity signals; acceleration signals refer to the rate of change of the target user's velocity during motion measured by sensors, such as the acceleration change signal generated by the up-and-down movement of the legs when running; angular velocity signals refer to the rate of change of the angle of rotation of the target user around a certain axis during motion, such as the angular velocity signal of the rotation around the shoulder joint when the arm swings.

[0040] Specifically, when monitoring exercise heart rate, the monitoring device first needs to acquire the target user's current exercise data. The monitoring device can collect data through built-in accelerometers (such as triaxial accelerometers) and angular velocity sensors (such as gyroscopes): accelerometers can sense the target user's acceleration changes in the x, y, and z directions; for example, when a user walks, the foot landing generates a large vertical acceleration. Angular velocity sensors can record the rotational angular velocity of the user's limbs around the three axes; for example, when a user rides a bicycle, the leg pedaling motion generates periodic angular velocity changes.

[0041] S202. Based on the motion data, determine the current motion pattern of the target user, and perform time-frequency domain analysis on the motion data to generate an artifact feature vector. The artifact feature vector is a data vector containing the vibration frequency and statistical characteristics of the current motion.

[0042] Among them, the motion mode refers to the type of motion currently being performed by the target user, used to distinguish different motion states, such as running, walking, cycling, sitting, etc.; the artifact feature vector is a data combination containing specific motion-related features, used to characterize the characteristics of interference signals generated by motion; statistical features refer to the feature quantities of motion data extracted through statistical methods, used to describe the overall distribution or trend of the data, such as mean, variance, peak value, kurtosis, etc.

[0043] Specifically, after acquiring motion data, the monitoring device performs two processes: first, determining the motion pattern, and second, generating an artifact feature vector. These two steps can be performed in parallel or sequentially. To determine the motion pattern, the monitoring device extracts features from the motion data and then matches them with a pre-defined motion pattern feature library. For example, running involves a larger peak acceleration and a higher frequency (typically 1-3Hz), while walking involves a smaller peak acceleration and a lower frequency (typically 0.5-1.5Hz). By comparing these, it can be determined whether the user is currently running or walking. To generate the artifact feature vector, the monitoring device first performs time-frequency domain analysis on the motion data: time-domain analysis yields the time-domain features of the data, and frequency-domain analysis yields the vibration frequency. Then, statistical features are calculated, such as the mean, variance, and root mean square of the motion data. Finally, these vibration frequencies and statistical features are integrated into a single vector, i.e., the artifact feature vector.

[0044] S203. Based on the motion mode, determine the target feature transfer function mapping table from the preset feature transfer function mapping table set. The target feature transfer function mapping table contains the reference artifact feature vector and transfer function model corresponding to the motion mode.

[0045] The preset feature transfer function mapping table set is a collection of feature transfer function mapping tables corresponding to various motion modes, used to store reference data and models under different motion modes; the target feature transfer function mapping table refers to the feature transfer function mapping table that matches the current motion mode, used to provide reference information under that motion mode; the reference artifact feature vector is a preset artifact feature vector corresponding to a specific motion mode, used as a comparison standard; the transfer function model is a mathematical model used to describe the conversion relationship between motion data and artifact signals, through which the corresponding artifact signal can be calculated from the motion data.

[0046] The transfer function model is a mathematical model that includes multiple preset motion patterns. The construction process of the transfer function model includes:

[0047] First, a synchronous dataset for model training is acquired. Specifically, test subjects are organized to perform a specific exercise pattern in a controlled environment (e.g., using a treadmill). During this process, at least two sets of data are acquired synchronously using acquisition devices: the first set is the first motion data, i.e., kinematic data characterizing the exercise pattern acquired by accelerometers and / or angular velocity sensors; the second set is the first PPG signal, i.e., a mixed signal containing the true heart rate component and motion artifacts caused by the exercise acquired by optical sensors. Optionally, to improve the accuracy of subsequent artifact extraction, a third set of data can be acquired simultaneously, i.e., a reference physiological signal, such as a true heart rate signal unaffected by exercise acquired by a chest patch electrocardiogram (ECG) device. The acquired multi-source data is then time-stamped, filtered, and segmented to obtain high-quality data segments that strictly correspond to the specific exercise pattern.

[0048] Next, the target artifact signal is extracted and a reference artifact feature vector is generated. This step aims to separate the pure artifact component from the first PPG signal. Specifically, using the aforementioned obtained reference physiological signal (ECG signal), an algorithm is used to generate a theoretically pure PPG waveform free of artifacts. Subsequently, signal processing techniques such as adaptive filtering are employed to subtract this theoretical PPG waveform from the first PPG signal, thereby separating the pure target artifact signal. Then, time-domain and frequency-domain analysis is performed on the target artifact signal to extract its key features (e.g., dominant frequency, spectral energy distribution), and these features are combined into a multi-dimensional vector, which is defined as the reference artifact feature vector corresponding to this specific motion pattern.

[0049] Next, the specific parameters of the transfer function model are identified and determined. This step aims to establish a mathematical mapping relationship from the first motion data (input) to the target artifact signal (output). Optionally, the transfer function model can adopt a multiple-input single-output (MISO) system architecture, such as a finite impulse response (FIR) filter model. Using the first motion data as the model input and the target artifact signal as the desired output, a system identification algorithm (e.g., recursive least squares) is used to iteratively solve for the model parameters (i.e., the coefficients of the FIR filter). This minimizes the error between the simulated artifact signal generated by the model based on the input first motion data and the target artifact signal. Through this process, a set of optimal model parameters that accurately describes the "motion-artifact" transformation relationship under this motion mode can be obtained.

[0050] Finally, the model is validated and a feature transfer function mapping table is constructed. Using a reserved, untrained test dataset, the performance of the trained transfer function model is validated, evaluating its accuracy and generalization ability on unknown data. For each validated motion mode, the reference artifact feature vector is associated with the transfer function model (i.e., a set of optimal model parameters) to jointly construct a feature transfer function mapping table. The multiple feature transfer function mapping tables generated for all preset motion modes (such as walking, running, cycling, etc.) are summarized to form the "preset feature transfer function mapping table set," which is then stored in the monitoring device's storage medium for use during online monitoring.

[0051] Specifically, after determining the target user's movement pattern, the monitoring device needs to access the reference data and model corresponding to that movement pattern. The monitoring device pre-stores a set of feature transfer function mapping tables, which are established for different movement patterns (such as running, walking, and cycling). Each mapping table contains multiple reference artifact feature vectors for that movement pattern (covering different intensities or postures of that movement pattern, such as reference vectors for slow walking and fast walking) and corresponding transfer function models (different reference vectors correspond to different model parameters). Based on the determined movement pattern, the monitoring device selects the corresponding target feature transfer function mapping table from the set; for example, it finds the "running mode feature transfer function mapping table" from the set.

[0052] S204. Match the artifact feature vector with the target feature transfer function mapping table to obtain the transfer function model that is optimal with the artifact feature vector. The transfer function model is a mathematical operator used to convert the motion data into a noise waveform.

[0053] Among them, the optimal transfer function model refers to the transfer function model that has the highest matching degree with the current artifact feature vector, and is used to most accurately convert motion data into simulated artifact signals.

[0054] Specifically, after determining the target feature transfer function mapping table, the monitoring device needs to match the currently generated artifact feature vector with the reference artifact feature vectors in the mapping table to find the optimal transfer function model. The monitoring device calculates the similarity (such as Euclidean distance, cosine similarity, etc.) between the current artifact feature vector and each reference artifact feature vector in the mapping table. The smaller the distance or the higher the similarity, the higher the matching degree. Through this matching method, the monitoring device can select the transfer function model most suitable for the current motion pattern.

[0055] S205. Input the motion data into the optimal transfer function model for processing to obtain the simulated artifact signal, which is used to simulate the real artifact signal that will be generated.

[0056] Among them, simulated artifact signal refers to waveform data used to simulate interference signals generated by motion, which is calculated by the transfer function model; real artifact signal refers to interference signals generated on PPG signal due to factors such as limb swaying and muscle contraction during motion, such as high-frequency noise mixed into the signal collected by PPG sensor caused by wrist vibration during running.

[0057] Specifically, after determining the optimal transfer function model, the monitoring device inputs the real-time collected motion data into the model for calculation to generate simulated artifact signals. The transfer function model calculates based on the characteristics of the motion data: for example, when the model receives an acceleration signal from a rapid arm swing, it outputs a corresponding high-frequency, high-amplitude simulated artifact signal; when the motion data shows relatively smooth limb movement, the model generates a low-frequency, low-amplitude simulated signal. These simulated artifact signals are highly similar to real artifact signals in terms of frequency and amplitude variation trends, providing a "template" for subsequent artifact removal from the PPG signal, thereby reducing the impact of real artifacts on heart rate calculation.

[0058] S206. Acquire the PPG signal of the target user in real time, and remove the analog artifact signal from the PPG signal by waveform fitting method to obtain the initial heart rate signal. The PPG signal is a photoplethysmography (PPG) signal.

[0059] Among them, PPG signal refers to photoplethysmography (PPG) signal, which is a signal generated by detecting changes in blood volume in blood vessels as the heart beats through photoelectric sensors, and is used to indirectly reflect heart rate information.

[0060] Specifically, while generating the simulated artifact signal, the monitoring equipment acquires the target user's PPG signal in real time and preprocesses it. The monitoring equipment acquires the signal through its built-in PPG sensor to obtain the raw PPG signal. Since the raw signal may contain motion artifacts, high-frequency noise, baseline drift, etc., the monitoring equipment performs preprocessing, including removing high-frequency noise through low-pass filtering, correcting baseline drift through high-pass filtering, and possibly performing signal amplification and DC component removal to make the waveform characteristics (such as peaks and troughs) of the PPG signal clearer.

[0061] S207. Extract the successive heartbeat interval of the initial heart rate signal using a peak detection algorithm. The successive heartbeat interval is the time length between two adjacent valid heartbeat peaks of the target user.

[0062] The initial heart rate signal refers to the signal that initially reflects the heart rate change after artifact removal and smoothing. Its waveform contains periodic peaks corresponding to the heartbeat. The successive heartbeat interval refers to the time interval between two adjacent effective heartbeat peaks, which is used to quantify the period of the heartbeat. The effective heartbeat peak refers to the peak in the initial heart rate signal that truly corresponds to the heartbeat, excluding false peaks caused by noise or interference.

[0063] Specifically, after obtaining the initial heart rate signal, the monitoring device needs to extract the successive heartbeat intervals. The monitoring device uses a peak detection algorithm to process the initial heart rate signal: first, it performs local extremum analysis on the signal to identify all possible peaks (such as positions where the signal value is greater than the adjacent points); then, it filters valid heartbeat peaks according to preset rules, such as excluding peaks with too small amplitude (which may be noise) or intervals that are too short / too long (which do not conform to physiological laws); finally, it calculates the time difference between two adjacent valid peaks to obtain the successive heartbeat intervals.

[0064] S208. Based on the preset successive heartbeat interval threshold, the successive heartbeat interval is input into the Kalman filter to obtain the final heart rate signal.

[0065] Among them, the preset successive heartbeat interval threshold refers to the numerical range set according to physiological laws and user characteristics to judge whether the successive heartbeat interval is reasonable, and is used to filter abnormal data; the Kalman filter is a recursive estimation algorithm used to optimize the signal by fusing measured values ​​and predicted values, and reduce noise and errors; the final heart rate signal refers to the accurate and stable heart rate data obtained after Kalman filtering.

[0066] Specifically, after extracting the successive heartbeat intervals, the monitoring device needs to optimize them using a Kalman filter to obtain the final heart rate signal. The monitoring device inputs the successive heartbeat intervals into the Kalman filter: the filter first predicts the current heartbeat interval (predicted value) based on historical interval data, then merges the actual extracted interval (measured value) with the predicted value. If the measured value is within a threshold range, it is retained and used to update the filter; if the measured value exceeds the threshold (possibly due to incomplete artifact removal), the predicted value is used instead. Through this recursive update method, the filter can smooth out abnormal fluctuations, resulting in a more stable successive heartbeat interval. Finally, the monitoring device converts the optimized successive heartbeat intervals into heart rate values, forming the final heart rate signal. This data can be displayed on the device in real time, allowing users to understand their heart rate status.

[0067] In this embodiment, by predicting artifacts based on motion data and generating simulated artifact signals in advance, and by matching the artifact feature vectors generated from real-time motion data with a preset feature transfer function mapping table corresponding to the motion mode, the optimal transfer function model can be quickly determined. Therefore, when the monitoring device acquires the interfered PPG signal, it can almost simultaneously obtain a simulated artifact signal that is highly similar to the real artifact. This effectively solves the problem in related technologies that require complex analysis of the interfered PPG signal, resulting in processing delays and insufficient real-time performance. Thus, without sacrificing the accuracy of artifact elimination, the real-time performance of exercise heart rate monitoring is significantly improved, providing users with more immediate and accurate feedback on their exercise physiological state.

[0068] In light of the above scenarios, the method provided in this implementation will now be described in more detail. Please refer to [link / reference]. Figure 3 This is another flowchart illustrating the motion heart rate monitoring method for artifact elimination in this application embodiment.

[0069] S301. Obtain the current motion data of the target user, which includes acceleration signals and angular velocity signals;

[0070] S302. Based on the motion data, determine the current motion pattern of the target user, and perform time-frequency domain analysis on the motion data to generate an artifact feature vector. The artifact feature vector is a data vector containing the vibration frequency and statistical characteristics of the current motion.

[0071] S303. Based on the motion mode, determine the target feature transfer function mapping table from the preset feature transfer function mapping table set. The target feature transfer function mapping table contains the reference artifact feature vector and transfer function model corresponding to the motion mode.

[0072] Steps S301~S303 and Figure 2The descriptions of steps S201 to S203 in the embodiments are similar, and can be referred to the descriptions of the corresponding steps, which will not be repeated here.

[0073] S304. Calculate the Euclidean distance between the artifact feature vector and each of the reference artifact feature vectors, and determine the reference artifact feature vector corresponding to the smallest Euclidean distance as the target feature vector;

[0074] Euclidean distance refers to the straight-line distance between two points in multidimensional space, which is used to measure the similarity between two feature vectors. The smaller the distance, the closer the features are. The target feature vector is the reference artifact feature vector with the smallest Euclidean distance to the current artifact feature vector, which is used to match the corresponding transfer function model.

[0075] Specifically, after determining the target feature transfer function mapping table, the monitoring device needs to find the reference artifact feature vector that is most similar to the current artifact feature vector by calculating the Euclidean distance. The monitoring device will calculate the Euclidean distance between the current artifact feature vector (such as an n-dimensional vector containing n features such as step frequency, mean acceleration, peak angular velocity, etc.) and all reference artifact feature vectors (of the same dimension) in the mapping table one by one. For example, the Euclidean distance between vectors (x1, y1) and (x2, y2) in two-dimensional space is √[(x1-x2)²+(y1-y2)²]. The monitoring device selects the reference artifact feature vector with the smallest distance and determines it as the target feature vector. The transfer function model corresponding to this vector is more likely to be applicable to the current motion state.

[0076] S305. Match the target feature vector with the corresponding transfer function model in the target feature transfer function mapping table to obtain the transfer function model that is optimal with the artifact feature vector.

[0077] The target feature transfer function mapping table refers to a preset mapping table corresponding to the current motion pattern, in which each reference artifact feature vector is associated with a transfer function model; the optimal transfer function model is the model that matches the target feature vector and can most accurately simulate the current motion artifact.

[0078] Specifically, after determining the target feature vector, the monitoring device needs to find the corresponding transfer function model from the target feature transfer function mapping table. The target feature transfer function mapping table stores a one-to-one correspondence between "reference artifact feature vectors and transfer function models." For example, in the mapping table for a running mode, the reference vector with a stride frequency of 1.8Hz corresponds to model A, and the reference vector with a stride frequency of 2.2Hz corresponds to model B. By looking up the mapping table, the monitoring device finds the transfer function model associated with the target feature vector. The parameters of this model are preset based on a motion scenario similar to the current feature, thus enabling more accurate conversion of motion data into simulated artifact signals and ensuring the effectiveness of subsequent artifact removal.

[0079] S306. Input the motion data into the optimal transfer function model for processing to obtain the simulated artifact signal, which is used to simulate the real artifact signal that will be generated.

[0080] Step S306 and Figure 2 The description of step S205 in the above embodiments is similar, and the description of the corresponding step can be referred to, and will not be repeated here.

[0081] S307. Acquire the PPG signal of the target user in real time and preprocess the PPG signal;

[0082] S308. Align the simulated artifact signal with the preprocessed PPG signal by segmentation, extract the corresponding first signal segment by time window sliding method, and calculate the fitting parameters of the first signal segment by waveform fitting method, which is the least squares algorithm or signal correlation analysis method.

[0083] Among them, the simulated artifact signal refers to the signal generated by the optimal transfer function model to simulate motion artifacts; the preprocessed PPG signal refers to the photoplethysmography pulse wave signal that has been processed by filtering, denoising and other methods, retaining the pulse characteristics and unremoved artifacts; the time window sliding method refers to sliding a fixed-length time window across the signal segment by segment to extract the signal segment within the window; the fitting parameters refer to the parameters that describe the waveform characteristics, including amplitude, phase, frequency and so on.

[0084] Specifically, after obtaining the simulated artifact signal and the preprocessed PPG signal, the monitoring equipment needs to segment and align them, extract segments, and calculate fitting parameters. First, the monitoring equipment aligns the simulated artifact signal and the PPG signal along the time axis to ensure they are synchronized in time. Then, a time window (e.g., 500 milliseconds) is set, and the equipment slides across the aligned signal in steps (e.g., 100 milliseconds). Each slide extracts a segment of the simulated artifact signal and the corresponding segment of the PPG signal within the window, recording this as the first signal segment. For each first signal segment, the monitoring equipment uses the least squares method to calculate the fitting parameters: if the least squares method is used, it assumes the artifact waveform is a function (e.g., a sine function), and by minimizing the sum of squared errors between the simulated artifact segment and the artifact portion of the PPG signal, the optimal amplitude, phase, and other fitting parameters are obtained.

[0085] S309. Based on the fitting parameters, adjust the amplitude and phase of the simulated artifact signal, the adjustment including fitting the simulated artifact signal to the artifact portion of the PPG signal corresponding to the first signal segment.

[0086] Among them, amplitude refers to the magnitude of the signal waveform, reflecting the signal strength; phase refers to the time offset of the signal waveform, reflecting the starting position of the signal; fitting refers to adjusting parameters to make the simulated artifact signal as consistent as possible with the waveform of the real artifact part.

[0087] Specifically, after obtaining the fitting parameters, the monitoring equipment needs to adjust the amplitude and phase of the simulated artifact signal based on these parameters to make the simulated signal closer to the real artifact portion of the PPG signal. Specifically, the monitoring equipment adjusts the overall amplitude of the simulated artifact signal according to the amplitude ratio in the fitting parameters. For example, if the fitting parameters show that the simulated signal amplitude is 0.8 times that of the real artifact, the simulated signal amplitude is multiplied by 1.25 to match the real artifact amplitude. The equipment also adjusts the phase of the simulated signal based on the phase difference. For example, if the fitting parameters show that the simulated signal lags behind the real artifact by 100 milliseconds, the simulated signal is shifted forward by 100 milliseconds to eliminate the phase difference. Through these adjustments, the simulated artifact signal can be highly fitted to the artifact portion of the PPG signal in each first signal segment, ensuring more accurate removal of interference during subsequent artifact elimination.

[0088] S310. Match the adjusted analog artifact signal with the preprocessed PPG signal point by point according to the time sequence.

[0089] Here, time series refers to the order in which signals are arranged chronologically; point-to-point correspondence means that the adjusted analog artifact signal and PPG signal are matched one-to-one at each time point.

[0090] After adjusting the analog artifact signal, the monitoring equipment needs to correlate it with the preprocessed PPG signal point by point in time series. Specifically, since the analog artifact signal and the PPG signal have already been time-aligned, the monitoring equipment will correlate them according to more precise time points (such as one sampling point every millisecond) and record the signal value corresponding to each time point.

[0091] S311. Subtract the adjusted simulated artifact signal value at the corresponding time point from the PPG signal value at each time point, and truncate the subtraction result based on a fixed threshold to obtain the PPG signal after artifact removal. The fixed threshold is used to limit the amplitude range of artifact removal to avoid signal distortion.

[0092] Among them, the PPG signal value refers to the value of the preprocessed photoplethysmography (PPG) signal at each time point, for example, the signal value at t=0.1 seconds is 0.8mV; the adjusted analog artifact signal value refers to the signal value of the analog artifact signal at the corresponding time point after amplitude and phase adjustment, for example, the signal value at t=0.1 seconds is 0.3mV; the fixed threshold refers to the preset numerical range used to limit the signal subtraction amplitude, for example, [-0.5mV, 1.5mV], to avoid signal distortion caused by excessive subtraction; the truncation process refers to the operation of limiting the subtraction result to the threshold range when it exceeds the fixed threshold.

[0093] After completing the point-by-point correspondence between the adjusted simulated artifact signal and the PPG signal, the monitoring equipment removes the artifact signal. Specifically, the monitoring equipment first subtracts the adjusted simulated artifact signal value at the corresponding time point from the preprocessed PPG signal value at each time point. Because motion artifacts may have nonlinearity or model errors, the subtraction results at some time points may exceed the physiologically reasonable range. Therefore, the monitoring equipment uses a fixed threshold for truncation: if the preliminary result is higher than the upper limit of the threshold (e.g., 1.5mV), it is set as the upper limit value; if it is lower than the lower limit of the threshold (e.g., -0.5mV), it is set as the lower limit value.

[0094] S312. The PPG signal after artifact removal is processed using a smoothing algorithm to obtain the initial heart rate signal;

[0095] The initial heart rate signal refers to the signal obtained after smoothing, which initially reflects changes in heart rate and forms the basis for subsequent optimization.

[0096] Specifically, after obtaining the PPG signal with artifacts removed, the monitoring equipment immediately smooths it to eliminate residual noise and make the signal more stable. The monitoring equipment uses smoothing algorithms to process the signal; for example, using Gaussian filtering, different weights are assigned to signal values ​​at different locations within the window (higher weight for the center, lower weight for the edges), and then a weighted average is calculated to obtain the smoothed signal, which is recorded as the initial heart rate signal. Through smoothing, high-frequency noise in the signal is suppressed, the pulse waveform outline is clearer, and the final initial heart rate signal more accurately reflects the periodicity of the heartbeat.

[0097] S313. Perform autocorrelation analysis on the initial heart rate signal to obtain the autocorrelation function of the initial heart rate signal;

[0098] Among them, autocorrelation analysis refers to the analysis method that calculates the correlation between a signal and its delayed version, which is used to reveal the periodic characteristics of the signal; the autocorrelation function is a function that describes the degree of similarity between a signal and itself at different delay times. For example, a signal with period t will show an autocorrelation peak at a delay of its period or a multiple of its period.

[0099] Specifically, after obtaining the initial heart rate signal, the monitoring device immediately performs autocorrelation analysis to extract the periodic characteristics of the signal. Specifically, the monitoring device calculates the autocorrelation function of the initial heart rate signal: for the initial heart rate signal x(t), the autocorrelation function R(τ) is defined as R(τ) = E[x(t)∙x(t+τ)], where τ is the delay time and E is the expectation operator. In actual calculations, the monitoring device processes the discrete initial heart rate signal, calculating the average of the product of the signal at time t and time t+τ for each possible delay τ (e.g., 0.1 seconds, 0.2 seconds…2 seconds), thus obtaining the value of R(τ). For example, if the period of the initial heart rate signal is 1 second, then at positions such as τ = 1 second, 2 seconds, etc., the autocorrelation function will show obvious peaks (because the signal is highly similar to itself at these delays); while for signals with strong noise, the peaks of the autocorrelation function will be more blurred. Through the autocorrelation function, the monitoring device can more clearly identify the periodic characteristics of the initial heart rate signal.

[0100] S314. Extract the peak value at the non-zero delay point in the correlation function and determine whether the peak value corresponds to the periodic characteristics of the initial heart rate signal. The periodic characteristics include the main frequency, the proportion of spectral energy, and the peak amplitude.

[0101] The peak value at a non-zero delay refers to the maximum value of the autocorrelation function when the delay time τ≠0, such as the peak value of the initial heart rate signal when the delay is equal to the heart rate cycle.

[0102] Specifically, after obtaining the autocorrelation function of the initial heart rate signal, the monitoring device determines the periodicity of the initial heart rate signal by analyzing the correspondence between the autocorrelation peaks and periodic characteristics. Specifically, the monitoring device first extracts the peak values ​​at all non-zero delays in the autocorrelation function (excluding peak values ​​at τ=0, as these only reflect the signal's own energy), and records the corresponding delay time and amplitude. Then, it analyzes the periodic characteristics of the initial heart rate signal, including: calculating the signal spectrum through Fourier transform to determine the dominant frequency and the spectral energy proportion of the dominant frequency; next, it determines whether the autocorrelation peaks correspond to these periodic characteristics. If the peaks correspond to the periodic characteristics, the signal periodicity is good; if the peak position deviates significantly from the dominant frequency period, or the peak amplitude is extremely low, it indicates a lack of correspondence, and the signal periodicity is poor.

[0103] S315. If they do not correspond, the initial heart rate signal is determined to be non-periodic, and the parameters of the corresponding transfer function model are adjusted.

[0104] Specifically, when the monitoring equipment determines that the autocorrelation peak value of the initial heart rate signal does not correspond to the periodic characteristics, it performs feedback optimization of the transfer function model to improve the artifact removal effect. The monitoring equipment first marks that the initial heart rate signal does not conform to periodicity, indicating that the simulated artifact signal generated by the current transfer function model deviates from the real artifact (such as insufficient amplitude or phase misalignment); then it analyzes the reasons for the deviation: if the autocorrelation peak value is blurred, it may be that the artifact removal is incomplete, and the amplitude of the simulated artifact needs to be enhanced, so the amplitude scaling factor of the transfer function is increased; if the peak value position is offset, it may be that the phase mismatch needs to be adjusted, and the phase compensation parameter of the transfer function needs to be adjusted.

[0105] S316. Obtain the second signal segment that does not conform to the periodicity in the initial heart rate signal, and extract the time domain features and frequency domain features of the second signal segment. The time domain features include waveform amplitude and periodic trend, and the frequency domain features include main frequency and spectral energy distribution.

[0106] The second signal segment refers to a local signal segment in the initial heart rate signal that lacks stable periodic characteristics.

[0107] Specifically, after determining that the initial heart rate signal does not conform to periodicity, the monitoring device extracts non-periodic signal segments from the initial heart rate signal. First, the monitoring device scans the initial heart rate signal through a sliding time window (e.g., 2 seconds), calculates the periodicity index (e.g., autocorrelation peak amplitude) of the signal within each window, and marks windows with indices below a preset threshold (e.g., 0.4) as second signal segments; then, it extracts the time-domain features of these segments, including the mean and variance of the waveform amplitude, and determines the periodicity trend through trend analysis; simultaneously, it extracts the frequency-domain features, including the dominant frequency and the distribution of spectral energy in different frequency bands.

[0108] S317. Based on the time-domain and frequency-domain characteristics, the second signal segment is corrected or compensated to obtain the periodically adjusted initial heart rate signal, specifically including:

[0109] If the main frequency of the second signal segment exceeds the preset frequency band range, then the second signal segment is bandpass filtered to retain only the frequencies within the preset frequency band range;

[0110] If the waveform amplitude of the second signal segment is abnormal, the second signal segment is normalized based on the waveform amplitude of the adjacent periodic signal segments.

[0111] If there is a periodic interruption in the second signal segment, then based on the periodic trend of the adjacent signal segments, the missing or abnormal waveforms in the second signal segment are processed by an interpolation algorithm.

[0112] Specifically, after extracting the time-domain and frequency-domain features of the second signal segment, the monitoring equipment immediately performs targeted corrections.

[0113] If the main frequency of the second signal segment exceeds the preset frequency range (e.g., 3.5Hz, which is higher than the upper limit of 3Hz), the monitoring device uses a bandpass filter (passband 0.5-3Hz) to filter the segment, filter out high-frequency noise (e.g., 5Hz interference caused by exercise), and retain possible heart rate signal components. For example, the main frequency of the filtered segment is adjusted to 1.5Hz (within the range).

[0114] If the waveform amplitude is abnormal (e.g., the average amplitude is 0.2mV, which is much lower than the 0.8mV of the adjacent normal segment), the monitoring device calculates the average amplitude of the adjacent normal segment (e.g., 0.8mV), multiplies the amplitude of the second signal segment by a scaling factor (0.8 / 0.2=4) to normalize its amplitude to the normal range, while preserving the original waveform trend.

[0115] If there is a periodic interruption (such as the absence of two heartbeat waveforms), the monitoring equipment analyzes the periodic trend of adjacent normal segments (such as the period of the previous segment gradually shortening from 1.0 seconds to 0.9 seconds, and the period of the next segment gradually shortening from 0.8 seconds to 0.7 seconds), and uses a cubic spline interpolation algorithm to generate a transition waveform at the interruption point, so that the period smoothly transitions from 0.9 seconds to 0.8 seconds, filling in the missing waveform.

[0116] Through the above corrections, the periodicity of the second signal segment is improved, and the final periodically adjusted initial heart rate signal is closer to the actual heart rate changes.

[0117] S318. Extract the successive heartbeat interval of the initial heart rate signal using a peak detection algorithm. The successive heartbeat interval is the time length between two adjacent valid heartbeat peaks of the target user.

[0118] Step S318 and Figure 2 The description of step S207 in the above embodiments is similar, and the description of the corresponding step can be referred to, and will not be repeated here.

[0119] S319. Obtain the preset threshold range of the successive heartbeat interval, which is dynamically adjusted according to the target user's age, exercise mode and historical heart rate data;

[0120] Among them, successive heartbeat interval refers to the time length between two adjacent valid heartbeat peaks of the target user; preset threshold range refers to a pre-set numerical range used to judge whether the successive heartbeat interval is reasonable.

[0121] Specifically, after obtaining a series of successive heart rate intervals (SBEs) through the peak detection algorithm, the monitoring device does not use them immediately. Instead, it first sets a validity criterion for these data, namely a preset threshold range. This preset threshold range is not a fixed threshold but is dynamically adjusted based on the target user's age and current exercise mode. For example, for a 25-year-old user, their resting heart rate might correspond to an SBE of 800-1200 milliseconds (50-75 beats / minute). However, during high-intensity exercise like running, the monitoring device will automatically adjust the threshold range to 300-500 milliseconds (120-200 beats / minute). Simultaneously, the monitoring device also refers to the user's historical heart rate data. If the user's heart rate is consistently high, the baseline threshold range will be adjusted accordingly.

[0122] S320. Input the successive heartbeat interval into the Kalman filter to obtain the measured and predicted values ​​of the successive heartbeat interval;

[0123] Among them, the Kalman filter refers to an efficient recursive estimation algorithm that can extract an estimate closer to the true value from a noisy signal by fusing measurement data and system state prediction; the measured value refers to the unfiltered successive heartbeat interval data actually extracted from the initial heart rate signal; the predicted value refers to the possible value of the heartbeat interval at the current moment predicted by the Kalman filter based on the system state at the previous moment (i.e., the trend of historical heartbeat interval changes).

[0124] Specifically, after determining the personalized threshold range, the monitoring device begins filtering the successive heartbeat interval sequence. Specifically, the monitoring device inputs the successive heartbeat interval data one by one into a Kalman filter. For each input data point, the filter performs two core actions: First, based on the previously processed heartbeat interval data and the internally established motion model, the filter generates a "predicted value." Then, it compares this predicted value with the newly input "measured value" (e.g., the actual measured 490ms).

[0125] S321. When the measured value exceeds the preset threshold range, the measured value is replaced by the predicted value.

[0126] Specifically, after acquiring the measured and predicted values, the monitoring device filters out obviously abnormal heartbeat intervals. The monitoring device first checks whether the currently input measured value falls within the dynamic threshold range determined in step S319. For example, if the current exercise mode is brisk walking, the threshold range is set to [500ms, 750ms]. If the measured value is 300ms (possibly due to artifact interference caused by sudden, violent arm swings), this value clearly exceeds the lower limit of the threshold range. The system then determines that the measured value is unreliable and discards it. To avoid interrupting the data stream, the system replaces this abnormal measured value with the predicted value from the current Kalman filter (e.g., a value predicted based on previous data of 580ms). Through this mechanism, sudden, extreme data points that do not conform to physiological patterns caused by incompletely eliminated motion artifacts can be effectively filtered out.

[0127] S322. The Kalman filter is used to recursively update the successive heartbeat interval signal to obtain the final heart rate signal.

[0128] Recursive update refers to the process by which the Kalman filter repeatedly performs a "prediction-update" cycle, where the output of each round serves as the basis for the next round of prediction, thereby continuously optimizing the estimated value.

[0129] Specifically, this is the final step in the heart rate monitoring method, designed to output a stable and reliable result. The Kalman filter is continuously updated recursively: at each time point, it weights and fuses the measured value (after S321 filtering) with the predicted value to generate a currently optimal estimate of the successive heart rate interval. This newly generated optimal estimate is then used to update the internal state of the filter and serves as the basis for the next prediction. Through multiple rounds of such recursive processing, spikes and jitters caused by noise or minor artifacts in the original successive heart rate interval signal are effectively smoothed. Finally, the monitoring device converts the smoothed successive heart rate interval sequence into heart rate values, forming the final heart rate signal. This data can be displayed in real time, providing users with an accurate reference for their exercise heart rate.

[0130] In this embodiment, the method employs dynamic adjustment of the amplitude and phase of the simulated artifact signal, fine fitting with the PPG signal, and point-by-point removal. Furthermore, it introduces a closed-loop feedback and correction mechanism based on autocorrelation analysis. This mechanism verifies the signal quality after artifact removal and adjusts the transfer function model in reverse. Finally, it combines a Kalman filter with a dynamic threshold for optimization. Therefore, this method not only focuses on unidirectional artifact prediction and removal but also constructs an adaptive processing closed loop of "prediction-fine removal-verification-feedback correction." This effectively solves the problems in related technologies, such as the inability to change the transfer function model once selected, poor adaptability to complex or abrupt motion scenarios, and signal distortion caused by incomplete or excessive artifact removal. Consequently, it achieves highly robust and high-fidelity artifact removal under various complex motion conditions, greatly improving the stability and reliability of heart rate monitoring results.

[0131] The monitoring device in the embodiments of this invention is described below from the perspective of hardware processing. Please refer to [link / reference needed]. Figure 4 This is a schematic diagram of the physical device structure of the monitoring equipment in the embodiments of this application.

[0132] It should be noted that, Figure 4 The structure of the monitoring device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0133] like Figure 4 As shown, the monitoring device includes a CPU 401, which can perform various appropriate actions and processes according to a program stored in the read-only memory ROM 402 or a program loaded from the storage section 408 into the random access memory RAM 403, such as performing the methods described in the above embodiments. The RAM 403 also stores various programs and data required for system operation. The CPU 401, ROM 402, and RAM 403 are interconnected via a bus 404. An I / O interface 405 is also connected to the bus 404.

[0134] The following components are connected to I / O interface 405: input section 406 including audio input devices, push-button switches, etc.; output section 407 including a liquid crystal display (LCD) and audio output devices, indicator lights, etc.; storage section 408 including a hard disk, etc.; and communication section 409 including a network interface card such as a LAN (Local Area Network) card, modem, etc. Communication section 409 performs communication processing via a network such as the Internet. Drive 410 is also connected to I / O interface 405 as needed. Removable media 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 410 as needed so that computer programs read from them can be installed into storage section 408 as needed.

[0135] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing computer programs for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by CPU 401, it performs the various functions defined in the present invention.

[0136] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0137] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. Each block in a flowchart or block diagram may represent a module, program segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those shown in the drawings.

[0138] Specifically, the monitoring device in this embodiment includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, it implements the motion heart rate monitoring method with artifact elimination provided in the above embodiment.

[0139] In another aspect, the present invention also provides a computer-readable storage medium, which may be included in the monitoring device described in the above embodiments; or it may exist independently and not assembled into the monitoring device. The storage medium carries one or more computer programs that, when executed by a processor of the monitoring device, cause the monitoring device to implement the motion heart rate monitoring method with artifact elimination provided in the above embodiments.

[0140] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0141] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0142] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for monitoring exercise heart rate with artifact elimination, applied to monitoring equipment, characterized in that, include: Acquire the target user's current motion data, which includes acceleration signals and angular velocity signals; Based on the motion data, the current motion pattern of the target user is determined, and time-frequency domain analysis is performed on the motion data to generate an artifact feature vector. The artifact feature vector is a data vector containing the vibration frequency and statistical characteristics of the current motion. Based on the motion pattern, a target feature transfer function mapping table is determined from a preset set of feature transfer function mapping tables. The target feature transfer function mapping table includes a reference artifact feature vector and a transfer function model corresponding to the motion pattern. The artifact feature vector is matched with the target feature transfer function mapping table to obtain the transfer function model that is optimal with the artifact feature vector. The transfer function model is a mathematical operator used to convert the motion data into a noise waveform. The motion data is input into the optimal transfer function model for processing to obtain a simulated artifact signal, which is used to simulate the real artifact signal that will be generated. The PPG signal of the target user is acquired in real time. The simulated artifact signal is removed from the PPG signal by a waveform fitting method to obtain the initial heart rate signal. The PPG signal is a photoplethysmography (PPG) signal. The successive heartbeat interval of the initial heart rate signal is extracted by a peak detection algorithm. The successive heartbeat interval is the time length between two adjacent valid heartbeat peaks of the target user. Based on a preset successive heartbeat interval threshold, the successive heartbeat interval is input into a Kalman filter to obtain the final heart rate signal.

2. The method according to claim 1, characterized in that, The step of matching the artifact feature vector with the target feature transfer function mapping table to obtain the transfer function model that best matches the artifact feature vector specifically includes: Calculate the Euclidean distance between the artifact feature vector and each of the reference artifact feature vectors, and determine the reference artifact feature vector corresponding to the smallest Euclidean distance as the target feature vector; The target feature vector is matched with the corresponding transfer function model in the target feature transfer function mapping table to obtain the optimal transfer function model for the artifact feature vector.

3. The method according to claim 1, characterized in that, The process of acquiring the target user's PPG signal in real time and removing the simulated artifact signal from the PPG signal using a waveform fitting method to obtain the initial heart rate signal specifically includes: The PPG signal of the target user is acquired in real time, and the PPG signal is preprocessed. The simulated artifact signal is segmented and aligned with the preprocessed PPG signal. The corresponding first signal segment is extracted by time window sliding method, and the fitting parameters of the first signal segment are calculated by waveform fitting method, which is the least squares algorithm or signal correlation analysis method. According to the fitting parameters, the amplitude and phase of the simulated artifact signal are adjusted, and the adjustment includes fitting the simulated artifact signal with the artifact portion of the PPG signal corresponding to the first signal segment. The adjusted simulated artifact signal is subtracted point by point from the preprocessed PPG signal, and the PPG signal after artifact removal is smoothed to obtain the initial heart rate signal.

4. The method according to claim 3, characterized in that, The step of subtracting the adjusted simulated artifact signal point by point from the preprocessed PPG signal and smoothing the artifact-free PPG signal to obtain the initial heart rate signal specifically includes: The adjusted analog artifact signals are matched point by point with the preprocessed PPG signals according to the time series. The PPG signal value at each time point is subtracted from the corresponding time point's adjusted simulated artifact signal value, and the subtraction result is truncated based on a fixed threshold to obtain the PPG signal after artifact removal. The fixed threshold is used to limit the amplitude range of artifact removal to avoid signal distortion. The PPG signal after artifact removal is processed using a smoothing algorithm to obtain the initial heart rate signal.

5. The method according to claim 1, characterized in that, After the steps of acquiring the target user's PPG signal in real time and removing the analog artifact signal from the PPG signal using a waveform fitting method to obtain the initial heart rate signal, the method further includes: Autocorrelation analysis was performed on the initial heart rate signal to obtain the autocorrelation function of the initial heart rate signal; The peak value at the non-zero delay point in the correlation function is extracted, and it is determined whether the peak value corresponds to the periodic characteristics of the initial heart rate signal. The periodic characteristics include the main frequency, the proportion of spectral energy, and the peak amplitude. If they do not correspond, the initial heart rate signal is determined to be non-periodic, and the parameters of the corresponding transfer function model are adjusted.

6. The method according to claim 5, characterized in that, After the steps of determining that the initial heart rate signal does not conform to periodicity if it does not correspond, and adjusting the parameters of the corresponding transfer function model, the method further includes: Obtain a second signal segment that does not conform to periodicity from the initial heart rate signal, and extract the time-domain features and frequency-domain features of the second signal segment. The time-domain features include waveform amplitude and periodic trend, and the frequency-domain features include the main frequency position and spectral energy distribution. If the main frequency of the second signal segment exceeds the preset frequency band range, then the second signal segment is bandpass filtered to retain only the frequencies within the preset frequency band range; If the waveform amplitude of the second signal segment is abnormal, the second signal segment is normalized based on the waveform amplitude of the adjacent periodic signal segments. If there is a periodic interruption in the second signal segment, then based on the periodic trend of the adjacent signal segments, the missing or abnormal waveforms in the second signal segment are processed by an interpolation algorithm.

7. The method according to claim 1, characterized in that, The step of inputting the successive heartbeat interval into a Kalman filter based on a preset successive heartbeat interval threshold to obtain the final heart rate signal specifically includes: A preset threshold range for the successive heartbeat intervals is obtained, and the preset threshold range is dynamically adjusted based on the target user's age, exercise mode, and historical heart rate data; The successive heartbeat interval is input into a Kalman filter to obtain the measured and predicted values ​​of the successive heartbeat interval; When the measured value exceeds the preset threshold range, the measured value is replaced by the predicted value; The Kalman filter is used to recursively update the successive heartbeat interval signal to obtain the final heart rate signal.

8. A monitoring device, characterized in that, The monitoring device includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code including computer instructions, and the one or more processors call the computer instructions to cause the monitoring device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the monitoring device, the monitoring device performs the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on the monitoring device, the monitoring device performs the method as described in any one of claims 1-7.

Images