Exercise assessment method, device and equipment based on multi-modal physiological data and medium

Abstract

The invention relates to a motion evaluation method, device and equipment based on multi-modal physiological data and a medium, and the method comprises the steps: solving the problem of space-time mismatch of the multi-modal data through sampling timestamps of a hardware clock protocol for multi-source physiological signals such as a makeup rate, myoelectricity, blood lactic acid and the like; equipment interference and motion artifacts are eliminated, and the signal quality is improved; dynamic characteristics such as heart rate variability, myoelectricity root mean square and blood lactic acid gradient in the sliding window are calculated; dividing exercise intensity intervals based on the individually calibrated heart rate percentage and the myoelectricity activation degree threshold, and detecting conversion candidate points; and recognizing a motion intensity critical state in real time through a self-adaptive threshold model driven by historical data, and generating a comprehensive evaluation result containing a thermodynamic diagram and an early warning report. According to the method, the limitation of a traditional fixed threshold model is broken through, multi-modal data deep fusion and individual dynamic adaptation are achieved, the exercise intensity critical point detection precision is improved, and real-time decision support is provided for training load optimization and rehabilitation progress evaluation.

Description

Technical Field

[0001] The present invention relates to the technical field of multimodal data processing, and in particular to a motion assessment method, apparatus, device and medium based on multimodal physiological data. Background Art

[0002] Exercise assessment methods have important application value in sports training and rehabilitation medicine. By analyzing the body's physiological responses during exercise, they provide a scientific basis for training plan development and rehabilitation progress monitoring. These methods are typically based on multi-dimensional physiological data such as heart rate, electromyographic signals, and metabolic indicators to comprehensively evaluate the body's exercise load tolerance, fatigue status, and energy metabolism efficiency. In competitive sports training, accurate exercise assessment helps optimize athletes' intensity distribution; in rehabilitation therapy, it can provide patients with personalized functional recovery guidance. With the development of wearable sensor technology, real-time acquisition of multimodal physiological data has become a key way to improve assessment accuracy.

[0003] However, existing exercise assessment technologies still face significant bottlenecks. First, the acquisition of multi-source physiological data suffers from temporal and spatial mismatches, preventing effective collaborative analysis of key parameters such as heart rate, electromyography, and metabolic indicators. Second, traditional assessment models rely on fixed threshold rules, making them difficult to adapt to individual physiological differences and dynamic changes. This static assessment mechanism results in delayed judgment of critical exercise intensity levels and an inability to capture anaerobic metabolic transition points in real time, ultimately leading to inaccurate training load recommendations and delayed adjustments to rehabilitation plans. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide a motion assessment method, device, equipment and medium based on multimodal physiological data that can achieve deep fusion of multimodal data, dynamically adapt to individual characteristics and provide real-time feedback on the critical state of motion intensity.

[0005] The purpose of the present invention is achieved by the following scheme:

[0006] In a first aspect, the present invention provides a motion assessment method based on multimodal physiological data, comprising the following steps:

[0007] S1: Perform spatiotemporal synchronization on the multimodal physiological raw data collected by sensors. The sampling timestamps of heart rate, electromyography, and blood lactate signals are aligned using a hardware clock protocol to generate a sports physiological dataset. The sports physiological dataset includes time-aligned heart rate waveforms, electromyography time-domain signals, and continuous blood lactate concentration curves.

[0008] S2: Perform noise filtering on the sports physiological dataset to remove the power frequency interference signal of the acquisition equipment and the electromyographic artifact noise in the sports environment, and generate a denoised physiological dataset with enhanced signal quality;

[0009] S3: Extract motion features from the denoised physiological data set, calculate heart rate variability, myoelectric root mean square value, and blood lactate concentration gradient change rate within the sliding window, and generate a multidimensional feature matrix containing time-varying physiological indicators;

[0010] S4: Perform exercise intensity classification on the multidimensional feature matrix, divide it into low, medium and high intensity intervals according to the preset heart rate percentage threshold and myoelectric activation threshold rules, and generate an intensity classification result including intensity level labels and conversion candidate points;

[0011] S5: Perform dynamic threshold detection on the intensity grading results, perform multimodal consistency verification and adaptive threshold decision on candidate transition points based on historical data, and generate an individualized threshold model. The individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point;

[0012] S6: Process the exercise physiological data set based on the individualized threshold model, output the exercise intensity heat map and threshold conversion warning report in real time, and generate dynamic exercise assessment results. The dynamic exercise assessment results are used to indicate the training load optimization plan and rehabilitation progress assessment report.

[0013] In a second aspect, the present invention provides a motion assessment device based on multimodal physiological data, the device being configured with the following modules:

[0014] The physiological data spatiotemporal synchronization module is used to synchronize the multimodal physiological raw data collected by sensors. It aligns the sampling timestamps of heart rate, electromyography, and blood lactate signals through the hardware clock protocol to generate a sports physiological data set. The sports physiological data set includes time-aligned heart rate waveforms, electromyography time domain signals, and continuous blood lactate concentration curves.

[0015] The physiological signal denoising processing module is used to perform noise filtering on the sports physiological data set, remove the power frequency interference signal of the acquisition equipment and the electromyographic artifact noise in the sports environment, and generate a denoised physiological data set with enhanced signal quality;

[0016] The motion feature extraction module is used to extract motion features from the denoised physiological data set, calculate the heart rate variability, myoelectric root mean square value and blood lactate concentration gradient change rate within the sliding window, and generate a multidimensional feature matrix containing time-varying physiological indicators;

[0017] The exercise intensity grading module is used to grade the exercise intensity of the multidimensional feature matrix, divide the exercise intensity into low, medium and high intensity intervals according to the preset heart rate percentage threshold and electromyographic activation threshold rules, and generate an intensity grading result including intensity level labels and conversion candidate points;

[0018] The personalized threshold modeling module is used to perform dynamic threshold detection on intensity grading results, perform multimodal consistency verification and adaptive threshold decision-making on candidate transition points based on historical data, and generate a personalized threshold model. The personalized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point;

[0019] The dynamic motion assessment module is used to process the motion physiological data set based on the individualized threshold model, output the motion intensity heat map and threshold conversion warning report in real time, and generate dynamic motion assessment results. The dynamic motion assessment results are used to indicate the training load optimization plan and rehabilitation progress assessment report.

[0020] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements any of the above-mentioned motion assessment methods based on multimodal physiological data.

[0021] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned motion assessment methods based on multimodal physiological data.

[0022] In summary, the motion assessment method based on multimodal physiological data provided by the present application can achieve a systematic breakthrough in motion assessment technology through the deep fusion and dynamic adaptation mechanism of multimodal physiological data. At the data processing level, the spatiotemporal synchronization mechanism based on the hardware clock protocol can eliminate the time misalignment problem of multi-source signal acquisition and achieve high-precision alignment of the original data; the adaptive filtering technology can achieve a substantial improvement in the signal-to-noise ratio by synergistically suppressing power frequency interference and motion artifacts. At the feature analysis level, the joint calculation of heart rate variability, myoelectric root mean square and blood lactate gradient within the sliding window can achieve a comprehensive quantification of exercise load characteristics; individualized calibration of heart rate percentage and myoelectric activation threshold rules are used to achieve physiological adaptability optimization of intensity grading. At the core decision-making level, the multimodal consistency verification mechanism based on historical data can reduce the misjudgment rate of transition point identification; the adaptive threshold decision model dynamically tracks individual physiological baseline drift to achieve real-time capture of the critical state of exercise intensity. The resulting integrated system of heatmaps and early warning reports enables visual monitoring of anaerobic metabolic transition points, providing a basis for intensity allocation decisions for training load optimization and establishing a quantitative feedback loop for rehabilitation progress assessment. This approach overcomes three major limitations of traditional static movement assessment: inadequate collaborative analysis due to insufficient multimodal data fusion, lack of individual adaptation caused by fixed threshold rules, and delayed intervention due to delayed assessment results. This approach fundamentally improves the accuracy and timeliness of sports science decision-making.

[0023] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A flowchart of a motion assessment method based on multimodal physiological data provided in an embodiment of the present application;

[0025] Figure 2 A schematic diagram of a process for generating a multidimensional feature matrix containing time-varying physiological indicators provided in an embodiment of the present application;

[0026] Figure 3 A schematic diagram of a process for generating an intensity grading result including intensity grade labels and conversion candidate points provided in an embodiment of the present application;

[0027] Figure 4 A schematic structural diagram of a motion assessment device based on multimodal physiological data provided in another embodiment of the present application. DETAILED DESCRIPTION

[0028] To facilitate understanding of the present invention, the present invention will be described more fully below with reference to the accompanying drawings. The drawings illustrate preferred embodiments of the present invention. However, the present invention 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.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which this invention pertains. The terms used in this specification are for the purpose of describing specific embodiments only and are not intended to limit the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0030] In one embodiment, Figure 1 As shown, a motion assessment method based on multimodal physiological data is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0031] S1: Perform spatiotemporal synchronization on the multimodal physiological raw data collected by sensors. Align the sampling timestamps of heart rate, electromyography, and blood lactate signals through the hardware clock protocol to generate a sports physiological dataset. The sports physiological dataset includes time-aligned heart rate waveforms, electromyography time-domain signals, and continuous blood lactate concentration curves.

[0032] Specifically, the system collects multimodal physiological raw data through various high-precision sensors. These sensors include but are not limited to heart rate sensors, electromyography sensors, and blood lactate detection devices. The high-precision sensors capture the physiological responses of the human body during exercise at high frequency and high resolution. Because different sensors may have different sampling frequencies and data output formats, this data needs to be synchronized in time and space. Preferably, the system uses a hardware clock protocol to provide a unified time base for different sensors, allowing the sampling timestamps of heart rate, electromyography, and blood lactate signals to be accurately aligned. Based on a timestamp calibration algorithm, the system identifies and corrects the time deviations between the sensors to generate an exercise physiological dataset. The resulting exercise physiological dataset integrates the time-aligned heart rate waveform, electromyography time-domain signal, and a continuous blood lactate concentration curve. The heart rate waveform records the changes in the heart beat interval with millisecond accuracy, the electromyography time-domain signal details the waveform characteristics of muscle electrical activity, and the blood lactate concentration curve plots the continuous change trajectory of blood lactate concentration over exercise time, providing a comprehensive and accurate data foundation for subsequent analysis and processing.

[0033] S2: Perform noise filtering on the sports physiological dataset to remove the power frequency interference signal of the acquisition equipment and the electromyographic artifact noise in the sports environment, and generate a denoised physiological dataset with enhanced signal quality.

[0034] Specifically, due to the complexity of the acquisition environment and the characteristics of the equipment itself, the raw data is inevitably contaminated with various noise components. The system uses digital filtering technology to remove power frequency interference signals generated by the acquisition equipment. Taking ECG signal acquisition as an example, common 50Hz or 60Hz power frequency interference often superimposes on physiological signals at a fixed frequency. The system uses a band-stop filter to precisely locate and filter out signal components within this frequency range, while minimizing the impact on the heart rate waveform itself and preserving its complete physiological characteristics. Furthermore, the system uses the Independent Component Analysis (ICA) algorithm to address EMG artifact noise generated during exercise. When the human body is exercising, the electrical activity of surrounding muscles can interfere with the signal acquisition of the target EMG sensor, generating artifact noise. The system uses the ICA algorithm to decompose the multi-channel EMG signals, identifying independent signal source components. By analyzing the statistical characteristics of each component, it distinguishes the independent components corresponding to the true EMG signal from the artifact noise, and then removes the artifact noise components, effectively purifying the EMG time domain signal. After this series of filtering operations, the signal quality of the generated denoised physiological dataset is significantly improved. The contours of the heart rate waveform are clearer, the true muscle electrical activity characteristics in the electromyographic time domain signal are more prominent, and the blood lactate concentration curve is smoother and more accurately reflects metabolic dynamics. This lays a solid data foundation for subsequent in-depth feature extraction work and greatly improves the accuracy and reliability of subsequent analysis.

[0035] S3: Extract motion features from the denoised physiological dataset, calculate the heart rate variability, myoelectric root mean square value, and blood lactate concentration gradient change rate within the sliding window, and generate a multidimensional feature matrix containing time-varying physiological indicators.

[0036] Specifically, the setting of the sliding window comprehensively considers the dynamic characteristics of the physiological process of exercise and the computational stability of the characteristic indicators. Its length has been verified by a large number of experiments and determined to be an appropriate time interval to ensure that a sufficient number of data points can be captured within the window to accurately reflect the physiological state, while ensuring that the window sliding step can reflect the dynamic change trend of the signal.

[0037] To calculate heart rate variability (HRV), the system measures the duration of continuous cardiac cycles within each sliding window and uses time domain analysis methods to calculate multiple indicators such as the standard deviation of the difference between adjacent RR intervals (SDNN) and the mean square error (RMSSD). These indicators can reflect the intensity and balance of the autonomic nervous system's regulatory effect on cardiac activity from different perspectives, providing a key basis for evaluating the impact of exercise load on the cardiovascular system.

[0038] In terms of calculating the root mean square (RMS) value of myoelectricity, the system squares the EMG signal within the window to eliminate the influence of positive and negative polarity, then calculates the mean and takes the square root. The resulting RMS value quantitatively characterizes the contraction strength of the muscle within the time window. By continuously calculating the RMS values ​​of multiple windows, a curve showing the change in muscle activation over time can be drawn, clearly showing the muscle's force at different stages of movement.

[0039] For the rate of change of blood lactate concentration gradient, the system uses a numerical differentiation algorithm to calculate its rate of change based on continuous blood lactate concentration data points. This rate of change intuitively reflects the dynamic process of the body's metabolism switching from aerobic metabolism to anaerobic metabolism. Areas with high rates of change often indicate that the anaerobic metabolism level is rising rapidly, and it is an important reference indicator for judging the exercise intensity threshold.

[0040] Through the above-mentioned feature extraction operation, the system constructs a multidimensional feature matrix containing time-varying physiological indicators. This matrix organizes data in a matrix form, where rows represent time series and columns correspond to the calculation results of different types of feature indicators. It comprehensively and systematically reflects the dynamic change characteristics of the body in multiple physiological dimensions such as cardiovascular regulation, muscle contraction and metabolic conversion during exercise, providing detailed data support for the subsequent accurate assessment of exercise intensity.

[0041] S4: Perform exercise intensity classification on the multidimensional feature matrix, divide it into low, medium and high intensity intervals according to the preset heart rate percentage threshold and electromyographic activation threshold rules, and generate an intensity classification result including intensity level labels and conversion candidate points.

[0042] Specifically, the system makes judgments based on pre-set heart rate percentage thresholds and electromyographic activation threshold rules. For the setting of the heart rate percentage threshold, this application refers to a large number of sports physiology research results, combined with the maximum heart rate estimation formula for people of different age groups and different health conditions, to determine the specific percentage values ​​for dividing the heart rate range into low, medium and high intensity intervals. For example, the low-intensity interval is usually set to a certain percentage range below the maximum heart rate, while the medium and high intensity intervals correspond to higher percentage ranges, respectively. The system quickly determines the intensity interval in which the heart rate is located by calculating in real time the proportion of the current heart rate to the estimated maximum heart rate.

[0043] Furthermore, for EMG activation, the system sets thresholds based on the ratio of the root mean square (RMS) value to the RMS value during maximum voluntary contraction. Through statistical analysis of a large amount of exercise data, the system identifies typical ranges of EMG activation at different exercise intensities and, in turn, defines corresponding threshold limits. When processing a multidimensional feature matrix, the system integrates heart rate and EMG data and applies a logical judgment algorithm to assign intensity levels to the exercise states corresponding to each time window. When both heart rate and EMG indicators meet the conditions for a specific intensity range, the system assigns the corresponding intensity level label to that time window. Furthermore, the system specifically identifies candidate points for intensity transitions. In a continuous sequence of time windows, the system monitors the changing trends of heart rate and EMG indicators. Once a transition from one intensity range to another is detected, the specific location and associated feature information of the candidate transition point are recorded. The resulting intensity grading results not only include the intensity level label for each moment but also detail potential points where intensity transitions occur. This provides accurate identification of key moments of intensity changes, facilitates in-depth analysis of physiological load fluctuations during exercise, and provides a key basis for developing personalized exercise plans.

[0044] S5: Perform dynamic threshold detection on the intensity grading results, perform multimodal consistency verification and adaptive threshold decision on candidate transition points based on historical data, and generate an individualized threshold model. The individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point.

[0045] Specifically, to improve the accuracy and personalization of exercise assessment, the system conducts in-depth multimodal consistency verification of candidate transition points based on historical data. This historical data encompasses a large number of multimodal physiological data records from different individuals in various exercise scenarios. Using data mining algorithms, the system extracts typical change patterns and associated features of multimodal data such as heart rate, electromyography, and blood lactate at the moment of exercise intensity transition from this historical data. For each candidate transition point detected, the system compares and analyzes the multimodal data within a nearby time window with the typical patterns in the historical data. By calculating similarity metrics between the data, such as Euclidean distance and correlation coefficient, it determines whether the candidate transition point conforms to the true physiological characteristics of exercise intensity transition. If verification is successful, the system further utilizes the transition point data, combined with the individual's historical exercise data, and applies adaptive threshold decision algorithms in machine learning, such as regression models based on support vector machines (SVMs) or neural networks, to dynamically adjust the initially set fixed threshold.

[0046] The system continuously collects individual physiological data from multiple exercise sessions, including heart rate trajectories, myoelectric activation patterns, and blood lactate metabolism curves across different exercise types and intensities. This data is used to construct a personalized data model. By analyzing the differences between individual and group data, the system identifies unique characteristics of individual physiological responses to exercise. For example, some individuals may experience significant blood lactate accumulation before their heart rate reaches a certain intensity threshold, or their myoelectric activation patterns may exhibit a distinctive rhythmic change during intensity transitions. The system then adjusts threshold parameters accordingly, adapting them to the individual's physiological characteristics and dynamic changes in exercise status. The resulting personalized threshold model accurately indicates the critical state of exercise intensity and the specific location of the anaerobic metabolic transition point. This model is stored in the system database as a mathematical model, and its parameters are continuously updated and optimized based on the individual's ongoing exercise data. This allows for a personalized and precise assessment of each user's exercise status, ensuring that the assessment results are closely tailored to their individual needs and providing a highly targeted scientific basis for subsequent training guidance and rehabilitation programs.

[0047] S6: Process the exercise physiological data set based on the individualized threshold model, output the exercise intensity heat map and threshold conversion warning report in real time, and generate dynamic exercise assessment results. The dynamic exercise assessment results are used to indicate the training load optimization plan and rehabilitation progress assessment report.

[0048] Specifically, when monitoring the exercise process in real time, the system performs instant analysis of the collected physiological data based on the parameter settings in the individualized threshold model. By comparing the real-time heart rate data with the individualized low, medium, and high intensity heart rate threshold intervals, combined with the real-time calculation results of the electromyography activation, the system can accurately determine the current level of exercise intensity and draw an exercise intensity heat map accordingly. The heat map uses an intuitive visualization method, with different colors or grayscale levels representing different intensity levels, marking the intensity distribution at each moment on the time axis. At the same time, in the spatial dimension, based on the electromyography activity data of different muscle groups, it can present the force intensity distribution of the muscles throughout the body during exercise, so that users can understand the dynamic changes of their own exercise intensity at a glance, and promptly identify the stages of excessive or insufficient intensity and the corresponding body parts.

[0049] In addition, the system closely monitors the proximity of various physiological indicators during exercise to the critical thresholds in the individualized threshold model. When it detects that indicators such as heart rate, electromyography, or blood lactate concentration are about to reach or have exceeded the anaerobic metabolism transition point threshold, the threshold conversion warning mechanism is quickly triggered. The warning report will be issued in a variety of pre-set ways, such as displaying a striking warning message on the monitoring terminal, issuing a sound prompt, or pushing a notification to an associated mobile device. It will list in detail the type of threshold that is about to be broken, the expected breakthrough time, and the corresponding physiological indicator change trend, providing users with timely warning information to help them adjust their exercise rhythm in a timely manner and avoid the adverse consequences of excessive exercise. The final dynamic exercise assessment results not only include the above-mentioned exercise intensity heat map and threshold conversion warning report, but also integrate optimization suggestions for training load and assessment conclusions on rehabilitation progress.

[0050] In terms of training load optimization, the system applies exercise physiology principles and training plan design algorithms based on data such as the duration of exercise intensity, conversion frequency, and energy consumption estimates for each intensity stage. It then proposes specific suggestions such as adjusting exercise duration, changing the distribution of exercise intensity intervals, and optimizing exercise intervals to maximize training effectiveness and improve training safety. For rehabilitation progress assessment, the system combines multiple exercise assessment data from the rehabilitation process to compare and analyze changes in physiological indicators before and after injury and at different stages of rehabilitation. It assesses multiple factors such as muscle strength recovery, cardiopulmonary function improvement, and metabolic capacity enhancement, and generates a detailed rehabilitation progress assessment report. This provides a quantitative basis for adjustments to rehabilitation treatment plans and helps promote the scientific and personalized advancement of rehabilitation training, thereby fully realizing the precise exercise assessment function based on multimodal physiological data to meet the professional needs of sports training and rehabilitation medicine.

[0051] In summary, the motion assessment method based on multimodal physiological data provided by the present application can achieve a systematic breakthrough in motion assessment technology through the deep fusion and dynamic adaptation mechanism of multimodal physiological data. At the data processing level, the spatiotemporal synchronization mechanism based on the hardware clock protocol can eliminate the time misalignment problem of multi-source signal acquisition and achieve high-precision alignment of the original data; the adaptive filtering technology can achieve a substantial improvement in the signal-to-noise ratio by synergistically suppressing power frequency interference and motion artifacts. At the feature analysis level, the joint calculation of heart rate variability, myoelectric root mean square and blood lactate gradient within the sliding window can achieve a comprehensive quantification of exercise load characteristics; individualized calibration of heart rate percentage and myoelectric activation threshold rules are used to achieve physiological adaptability optimization of intensity grading. At the core decision-making level, the multimodal consistency verification mechanism based on historical data can reduce the misjudgment rate of transition point identification; the adaptive threshold decision model dynamically tracks individual physiological baseline drift to achieve real-time capture of the critical state of exercise intensity. The resulting integrated system of heatmaps and early warning reports enables visual monitoring of anaerobic metabolic transition points, providing a basis for intensity allocation decisions for training load optimization and establishing a quantitative feedback loop for rehabilitation progress assessment. This approach overcomes three major limitations of traditional static movement assessment: inadequate collaborative analysis due to insufficient multimodal data fusion, lack of individual adaptation caused by fixed threshold rules, and delayed intervention due to delayed assessment results. This approach fundamentally improves the accuracy and timeliness of sports science decision-making.

[0052] In one embodiment, S1 of a motion assessment method based on multimodal physiological data provided by the present invention specifically includes the following steps:

[0053] S11: Perform heterogeneous data reception and processing on the original physiological signals collected by multiple sensors, synchronously obtain the original signals output by the heart rate monitoring device, electromyography sensor, blood lactate analyzer and respiratory monitoring device, and generate an original heterogeneous physiological data set.

[0054] Specifically, in an exercise assessment scenario, the system simultaneously receives raw signals from various monitoring devices, including heart rate monitors, electromyography sensors, blood lactate analyzers, and respiratory monitors. Based on their respective sensing principles and technical standards, these devices generate raw data with varying formats, sampling frequencies, and data volumes. Heart rate monitors continuously output electrical signals reflecting heartbeats at a high sampling rate, generating a heart rate waveform sequence. Myoelectric sensors capture the weak bioelectrical signals generated by muscle contraction, generating a myoelectric waveform sequence. Blood lactate analyzers collect blood samples at specific time points during exercise to analyze blood lactate concentration, generating discrete blood lactate concentration data points. Respiratory monitors record parameters such as respiratory rate and depth, generating a respiratory parameter sequence. To address this heterogeneous data, the system utilizes a specially designed data interface module to ensure stable and efficient reception of these diverse data streams and integrate them into a raw, heterogeneous physiological dataset. This dataset fully preserves the characteristics, timestamps, and original data formats of each raw signal, providing a comprehensive, unaltered raw data foundation for subsequent processing steps, ensuring data integrity and authenticity.

[0055] S12: Perform time reference alignment processing on the original heterogeneous physiological data set, use a hardware clock synchronization protocol to assign a unified time reference tag to all data streams of the original heterogeneous physiological data set, and generate a time synchronized data set.

[0056] Specifically, because different monitoring devices may have independent clock systems, the timestamps of the collected physiological signals may differ. To address this issue, in this embodiment, the system can use a hardware clock synchronization protocol to assign a unified time reference to all data streams as data is received. The synchronization module within the system time-calibrates the data output by each device based on the precise time of the hardware clock. Specifically, the system analyzes the deviation between each device clock and the hardware clock and uses algorithms such as linear interpolation and time offset correction to accurately adjust the timestamp of each data point, ensuring that heart rate, electromyography, blood lactate, and respiratory monitoring data are aligned in the same time coordinate system. The resulting time-synchronized dataset allows the integration of various physiological signals in a unified time sequence, achieving precise matching of different physiological indicators in the time dimension. This processing step lays a solid foundation for subsequent data fusion and analysis, ensuring that the relationships between different physiological indicators and their dynamic changes over time can be accurately assessed during the subsequent analysis process, thereby improving the accuracy and reliability of motion assessment.

[0057] S13: Perform blood lactate signal reconstruction processing on the time-synchronized dataset, convert the intermittent sampling data into a continuous time series, and generate a sports physiological dataset. The sports physiological dataset includes a time-aligned heart rate waveform sequence, an electromyographic signal waveform sequence, a respiratory parameter sequence, and a blood lactate concentration sequence in the continuous time domain.

[0058] Specifically, because blood lactate analyzers typically acquire data using intermittent sampling, the blood lactate concentration data exhibits discontinuities in the time series. To address this issue, the system converts the intermittently sampled data into a continuous time series. The system first preprocesses the discrete blood lactate concentration data points, including removing outliers and smoothing to improve data quality. The system can then apply interpolation algorithms, such as spline interpolation and polynomial fitting, to generate reasonable intermediate values ​​between adjacent data points based on the concentration values ​​and time positions of known data points, thereby constructing a continuous blood lactate concentration curve.

[0059] During this process, the system fully considers the physiological variations in blood lactate concentration to ensure that the reconstructed signal conforms to the physiological metabolic characteristics of the human body. The resulting exercise physiology dataset integrates time-aligned heart rate waveform sequences, electromyographic signal waveform sequences, respiratory parameter sequences, and blood lactate concentration sequences in the continuous time domain. This dataset not only achieves temporal synchronization but also improves the integrity of physiological indicators, enabling various physiological data to be presented in a continuous and coordinated manner. This provides high-quality data support for subsequent exercise feature extraction and analysis, and facilitates more accurate assessment of key exercise physiology indicators such as exercise intensity, fatigue status, and energy metabolism efficiency.

[0060] In one embodiment, S2 of a motion assessment method based on multimodal physiological data provided by the present invention specifically includes the following steps:

[0061] S21: Based on the adaptive notch filtering technology, the power frequency interference elimination processing is performed on the exercise physiological data set to filter out the 50Hz power line interference component in the ECG signal and generate a preliminary filtered data set.

[0062] Specifically, power frequency interference usually refers to 50Hz or 60Hz power line interference, which will be superimposed on the ECG signal and affect the accurate measurement of heart rate. The system uses an adaptive notch filter that can dynamically adjust its frequency response to accurately suppress interference signals of specific frequencies. In this embodiment, the system continuously monitors the input signal, estimates and tracks the amplitude and phase of the 50Hz interference in real time, and generates a signal with the same amplitude but opposite phase as the interference signal to offset it. Based on the minimum mean square error (LMS) algorithm, the system iteratively adjusts the filter weights to minimize the power frequency interference component in the output signal. The system processes each ECG signal sample in real time to ensure that the integrity and authenticity of the ECG signal are retained while removing interference.

[0063] After this processing step, the system generates a preliminary filtered data set, in which the 50Hz power frequency interference component in the ECG signal is effectively filtered out, providing a purer data basis for subsequent heart rate analysis and heart rate variability calculation. The system further performs spectral analysis on the data after preliminary filtering to verify the filtering effect and ensure that there is no obvious 50Hz peak in the filtered signal spectrum. At the same time, the system also compares the time domain and frequency domain features of the filtered signal to ensure that the key features of the ECG signal, such as the morphology and amplitude of the P wave, QRS complex and T wave, are not adversely affected. In addition, the system automatically adjusts and optimizes the parameters in the filtering process to adapt to the ECG signal characteristics of different individuals and different power frequency interference intensities, thereby ensuring the consistency and reliability of the filtering effect and providing high-quality data support for subsequent physiological data analysis.

[0064] S22: Perform motion artifact suppression processing on the preliminary filtered dataset, construct a limb motion-noise correlation model and inversely compensate for the distortion of the electromyographic signal waveform to generate a motion artifact suppression dataset. The motion artifact suppression dataset is used to indicate the muscle activation state after removing limb motion interference.

[0065] Specifically, motion artifacts are often caused by limb movement, which can distort the EMG waveform and affect the accurate assessment of true muscle activation. To this end, the system constructs a limb motion-noise correlation model based on feature analysis and statistical modeling of motion artifacts. By extracting the time and frequency domain features of motion artifacts, the system can identify the artifact's manifestation in the EMG signal and establish a corresponding mathematical model to describe its relationship with the true EMG signal. Based on this model, the system can use an inverse compensation algorithm to correct the distorted EMG waveform. This inverse compensation estimates the amplitude and phase of the artifact and generates a corresponding compensation signal, which is subtracted from the original EMG signal to restore an EMG waveform that is closer to the true state. The resulting motion artifact suppression dataset not only effectively removes limb motion interference but also preserves the true activation pattern and intensity information of muscles during movement, providing reliable data support for subsequent muscle function analysis and exercise intensity assessment.

[0066] When building the model, the system analyzes a large amount of EMG signal data under different motion modes to extract typical features of motion artifacts, such as their specific frequency components in the frequency domain and their change patterns in the time domain. At the same time, the system also takes individual differences into account and adaptively models the muscle characteristics of different individuals to improve the universality and accuracy of the model. During the reverse compensation process, the system monitors the compensation effect in real time, calculates the signal difference before and after compensation, and automatically adjusts the compensation parameters to achieve the best signal recovery effect. In addition, the system also performs a quality assessment on the compensated EMG signal to ensure that it meets the preset signal-to-noise ratio and waveform fidelity standards, thereby ensuring the accuracy of subsequent analysis.

[0067] S23: Performing environmental noise filtering on the motion artifact suppression dataset to filter out the environmental noise component in the respiratory signal and generate a denoised physiological dataset.

[0068] Specifically, respiratory signals are susceptible to interference from ambient noise during acquisition, such as background noise and equipment noise. This noise can mask subtle changes in the respiratory signal and hinder accurate analysis of respiratory patterns and parameters. The system uses advanced environmental noise filtering algorithms, such as adaptive filtering or wavelet transform, to process the respiratory signal. Adaptive filters adjust their coefficients in real time to match the characteristics of the ambient noise, thereby separating it from the respiratory signal. Wavelet transforms utilize their localized properties in the time and frequency domains to decompose and reconstruct the signal, effectively removing noise components. The filtered respiratory signal is smoother and clearer, more accurately reflecting key parameters such as respiratory flow, respiratory rate, and respiratory depth. The resulting denoised physiological dataset integrates the cleaned respiratory signal with other physiological signals, providing a high-quality data foundation for comprehensive assessment of respiratory function and overall physiological status during exercise. When processing the respiratory signal, the system first analyzes the characteristics of the ambient noise to determine its primary frequency range and power spectral density. Based on these characteristics, it then selects appropriate filtering algorithms and parameters to effectively suppress the ambient noise.

[0069] During the adaptive filtering process, the system uses a reference noise signal to assist in adjusting the filter coefficients, thereby improving filtering effectiveness. For wavelet transform processing, the system selects an appropriate mother wavelet function and decomposition scale to ensure that key features of the respiratory signal, such as the peak and valley positions and amplitude variations of the respiratory wave, are preserved while removing noise. Furthermore, the system smoothes the filtered respiratory signal to remove any remaining high-frequency glitches while maintaining its natural form and physiological significance.

[0070] In one embodiment, Figure 2 As shown, S3 of a motion assessment method based on multimodal physiological data provided by the present invention specifically includes the following steps:

[0071] S31: Perform motion cycle segmentation processing on the denoised physiological data set, automatically divide the action cycle boundaries based on the joint kinematic characteristics, and generate segmented motion physiological data.

[0072] Specifically, the system performs motion cycle segmentation on the denoised physiological dataset, automatically demarcating motion cycle boundaries based on joint kinematic features to generate segmented motion physiological data. Based on real-time monitoring and analysis of joint kinematic features during movement, the system precisely identifies the start and end points of movements, thereby accurately segmenting motion cycles. By analyzing the changing patterns of kinematic parameters such as joint angles, angular velocities, and angular accelerations, the system identifies key characteristic points within each motion cycle. For example, in gait analysis, the system monitors changes in knee and ankle angles to identify key phases such as heel strike, stance phase, and swing phase, thereby segmenting the continuous motion process into multiple complete gait cycles. In more complex movement patterns, such as resistance training, the system analyzes the coordinated motion characteristics of multiple joints to identify the cyclical patterns of muscle contraction and relaxation. Through this automated cycle segmentation, the segmented motion physiological data generated by the system not only preserves detailed physiological information within each motion cycle but also provides a clear time frame for subsequent feature extraction and analysis. When demarcating motion cycle boundaries, the system comprehensively considers the kinematic characteristics of multiple joints and employs machine learning algorithms trained on large amounts of motion data to improve recognition accuracy and robustness. During processing, the system also verifies the boundary demarcation results to ensure the integrity and consistency of each segmented data, providing a reliable data foundation for subsequent physiological feature calculations.

[0073] S32: performing heart rate variability calculation processing on the segmented exercise physiological data, quantifying the cardiac autonomic nervous system regulation state by analyzing the standard deviation of the continuous RR interval sequence, and generating a heart rate variability feature sequence.

[0074] Specifically, the system calculates heart rate variability (HRV) on segmented exercise physiological data, quantifying the cardiac autonomic nervous system's regulatory state by analyzing the standard deviation of continuous RR interval sequences, and generating a characteristic sequence of heart rate variability. Heart rate variability (HRV) is an important indicator of the autonomic nervous system's regulatory function on the heart. Its calculation involves precise measurement and statistical analysis of RR intervals in continuous ECG signals. Within each segmented motion cycle, the system extracts the RR interval sequence and calculates statistical indicators such as its standard deviation. The size of the standard deviation reflects the degree of fluctuation in the RR interval, and thus reflects the balance between the sympathetic and parasympathetic nervous systems. A higher standard deviation generally indicates stronger parasympathetic nervous system activity and better cardiac adaptability; a lower standard deviation may indicate a predominance of the sympathetic nervous system, placing greater stress on the heart.

[0075] Through this detailed HRV analysis, the system generates a characteristic sequence of heart rate variability, providing critical data for assessing the state of the heart's autonomic nervous system during exercise. When calculating HRV, the system uses multiple algorithms to pre-process the RR intervals, such as removing outliers and correcting artifacts, to ensure the accuracy of the results. Furthermore, the system analyzes HRV trends across different exercise intensities and types, taking into account the different phases of the exercise cycle. This provides important information for subsequent exercise intensity assessment and training optimization.

[0076] S33: Calculate and process the myoelectric root mean square value of the segmented motion physiological data, quantify the amplitude intensity of the myoelectric signal within the sliding window through integration operation, and generate an electromyographic activation intensity sequence.

[0077] Specifically, the root mean square value (RMS) of myoelectricity is an important indicator to measure the degree of muscle activation. In each segmented movement cycle, the system sets a sliding window. The selection of the window length T comprehensively considers the physiological characteristics of muscle contraction and the duration of the movement cycle to ensure that the key features of muscle activation can be captured. The system squares the electromyographic signal in each window to eliminate the influence of the positive and negative polarity of the signal; then integrates the squared signal to calculate its area in the time window; finally, the square root is taken to obtain the RMS value of the window. Through the gradual movement of the sliding window, the system generates a continuous sequence of myoelectric activation intensity, which intuitively reflects the changes in the activation intensity of the muscle during exercise. The calculation formula of the myoelectric activation intensity sequence is:

[0078]

[0079] Among them, RMS is the myoelectric activation intensity sequence, that is, the root mean square value of myoelectricity, T is the sliding window length, x t is the electromyographic signal value at time t. During the calculation process, the system automatically adjusts the sliding window length and step size based on muscle type and movement pattern to optimize the resolution and representativeness of the calculation results. Furthermore, the system normalizes the calculated RMS sequence to eliminate differences in signal amplitude between different muscles, thereby achieving a unified assessment of the activation status of multiple muscles.

[0080] S34: Calculate and process the blood lactate concentration gradient of the segmented exercise physiological data, reflect the anaerobic metabolic state through the concentration difference change rate of adjacent sampling points, and generate a blood lactate dynamic gradient sequence.

[0081] Specifically, the blood lactate concentration gradient is a key indicator for assessing anaerobic metabolic activity during exercise. Its calculation is based on continuous blood lactate concentration monitoring data. Within each segmented exercise cycle, the system extracts continuous blood lactate concentration sampling points, calculates the concentration difference between adjacent sampling points, and further determines the rate of change. This rate of change reflects the rate of increase or decrease in blood lactate concentration over time. A positive rate of change indicates increased anaerobic metabolic activity, with the lactate production rate exceeding the clearance rate; a negative rate of change may indicate decreased exercise intensity and effective lactate clearance. Through this gradient calculation, the system generates a dynamic blood lactate gradient sequence, providing key data for analyzing metabolic state transitions during exercise. When calculating the blood lactate concentration gradient, the system uses numerical differentiation and optimizes parameters based on the duration of the exercise cycle and sampling frequency to improve the accuracy and stability of the results. The system also smoothes the gradient sequence to remove any potential high-frequency noise, ensuring that the data accurately reflects the dynamic changes in anaerobic metabolism.

[0082] S35: Perform matrix integration processing on the heart rate variability feature sequence, the electromyographic activation intensity sequence, and the blood lactate dynamic gradient sequence to generate a multidimensional feature matrix.

[0083] Specifically, the construction of a multidimensional feature matrix aims to comprehensively reflect the multidimensional physiological changes during exercise, providing comprehensive data support for subsequent exercise intensity assessment and personalized analysis. The system arranges the data for each sequence in chronological order and integrates it into a matrix format, with each row representing a time point or an exercise cycle, and each column corresponding to a characteristic indicator. Through this integration, the system can simultaneously display multiple physiological information, including the state of cardiac autonomic nervous system regulation, muscle activation intensity, and dynamic changes in anaerobic metabolism, within the same time frame. The multidimensional feature matrix not only preserves the original data of each feature sequence but also facilitates subsequent data analysis and pattern recognition through its mathematical structure. During the integration process, the system normalizes the data to eliminate dimensional differences between different characteristic indicators and ensure balanced weighting of each feature in the matrix. Furthermore, the system performs dimensionality reduction on the matrix to extract key characteristic components, improving the efficiency and accuracy of subsequent analysis. The resulting multidimensional feature matrix provides structured input for the training and prediction of exercise assessment models, ensuring that the assessment results fully reflect the comprehensive physiological responses during exercise.

[0084] The above-mentioned motion assessment method based on multimodal physiological data can achieve accurate and systematic improvement in the quantitative assessment of motion state through the collaborative extraction and fusion mechanism of multimodal motion physiological features. The motion cycle segmentation process automatically divides the motion boundaries based on the kinematic characteristics of the joints, which can eliminate the subjective errors of manual segmentation and achieve accurate alignment of motion physiological data with the motion cycle. The heart rate variability calculation can achieve dynamic quantification of the cardiac autonomic nervous system regulation state through the standard deviation analysis of continuous RR intervals to reflect the cardiovascular stress level under exercise load. The integral operation of the root mean square value of the electromyography can achieve the time domain energy representation of the muscle activation intensity within the sliding window, providing an objective basis for fatigue state assessment. The blood lactate concentration gradient calculation can achieve real-time monitoring of the anaerobic metabolic state through the rate of change analysis to capture the key nodes of energy metabolism conversion. Finally, through the matrix integration of multiple feature sequences, the unified spatiotemporal expression of time-varying physiological indicators can be achieved, establishing a structured foundation for the collaborative analysis of multimodal data. This process breaks through the limitations of traditional single-indicator evaluation and achieves the joint analysis of the triple physiological systems of heart rate, electromyography, and metabolism. It fundamentally solves the problem of insufficient evaluation accuracy caused by the fragmentation of motion characteristics and provides high-confidence feature input for intensity grading and threshold decisions.

[0085] In one embodiment, Figure 3 As shown, S4 of the motion assessment method based on multimodal physiological data provided by the present invention specifically includes the following steps:

[0086] S41: Perform individualized physiological benchmark calibration on the multidimensional feature matrix, dynamically adjust the maximum heart rate and maximum electromyographic activation reference values ​​based on resting physiological data, and generate personalized intensity threshold parameters.

[0087] Specifically, the system dynamically adjusts reference values ​​for maximum heart rate and maximum EMG activation based on resting physiological data to generate personalized intensity threshold parameters. This process involves analyzing an individual's resting heart rate and EMG data. The resting state is typically defined as a stable physiological state before exercise, when the body is relaxed and not under significant exertion. The system statistically analyzes resting heart rate data to calculate an individual's predicted maximum heart rate, correcting it based on personal information such as age, gender, and physical condition. Similarly, for EMG data, the system analyzes the resting EMG signal baseline to determine a reference value for maximum EMG activation. This typically involves a comprehensive assessment of the noise floor level in the non-contracted state and the signal characteristics during mild voluntary contraction. Through this personalized calibration, the system eliminates the influence of individual physiological differences, making subsequent intensity assessments more targeted and accurate. The generation of personalized intensity threshold parameters provides a critical reference for subsequent conversion of heart rate and EMG data, ensuring that the assessment results truly reflect the individual's physiological load during exercise.

[0088] S42: Performing heart rate percentage conversion processing on the multidimensional feature matrix based on the personalized intensity threshold parameter, calculating the percentage of the current heart rate to the personalized maximum heart rate, and generating a heart rate percentage feature vector.

[0089] Specifically, the system calculates the percentage of the current heart rate to the personalized maximum heart rate and generates a heart rate percentage feature vector. This conversion process requires the system to possess high-precision numerical processing capabilities to ensure that each heart rate data point is accurately converted to its corresponding percentage value. Specifically, the system compares the real-time monitored heart rate data with the personalized maximum heart rate reference value and obtains the percentage value through a simple division operation. For example, if the individual's maximum heart rate reference value is 180 beats / minute and the currently monitored heart rate is 120 beats / minute, the heart rate percentage is 66.7%. This heart rate percentage feature vector can intuitively reflect the relative position of the current heart rate within the individual's maximum heart rate range, providing a standardized metric for subsequent exercise intensity assessment. This metric helps eliminate the impact of individual maximum heart rate differences, allowing heart rate data to be compared and analyzed on a unified scale, thereby providing a more reliable basis for the classification of exercise intensity.

[0090] S43: Performing myoelectric activation conversion processing on the multidimensional feature matrix based on the personalized intensity threshold parameter, calculating the percentage of the current myoelectric root mean square value to the personalized maximum myoelectric activation, and generating a myoelectric activation feature vector.

[0091] Specifically, the system calculates the percentage of the current EMG RMS value to the individual's maximum EMG activation, generating an EMG activation feature vector. During this process, the system first determines the individual's maximum EMG activation reference value, typically derived from analyzing specific muscle activation patterns at rest. The system then calculates the RMS value of each EMG signal segment and compares it to this reference value to obtain a corresponding percentage. For example, if the individual's maximum EMG activation reference value is 5.0 mV and the currently calculated EMG RMS value is 2.5 mV, the EMG activation percentage is 50.0%. This EMG activation feature vector reflects the ratio of muscle activation during exercise to the individual's maximum activation capacity, providing an important quantitative indicator for assessing muscle work intensity and fatigue. Through this conversion, the system can analyze EMG data from different individuals on a unified scale, further improving the accuracy and comparability of exercise intensity assessment.

[0092] S44: Based on the preset heart rate percentage threshold and electromyographic activation threshold rules, the heart rate percentage feature vector and the electromyographic activation feature vector are subjected to intensity interval decision processing, and are divided into low intensity, medium intensity and high intensity level intervals to generate a preliminary intensity label set.

[0093] Specifically, the system processes these two feature vectors based on preset heart rate percentage thresholds and EMG activation thresholds. The system then categorizes these two feature vectors into low, medium, and high intensity levels, generating a preliminary set of intensity labels. These thresholds are typically based on extensive experimental data and physiological research. For example, a heart rate percentage below 60% might be classified as low intensity, 60% to 80% as medium intensity, and above 80% as high intensity. The EMG activation percentage intervals can also be adjusted based on different muscle groups and exercise types. Using a logical judgment algorithm, the system evaluates the two feature vector values ​​at each time point or movement cycle, determines the intensity range they belong to, and assigns the corresponding intensity label. This process requires the system to possess efficient data processing and logical judgment capabilities to ensure rapid and accurate classification of large amounts of feature data. The generation of this preliminary set of intensity labels provides the foundation for subsequent intensity grading, enabling the system to initially identify and record intensity changes during exercise.

[0094] S45: Perform state transition point detection processing on the preliminary intensity label set, identify the window position where the heart rate gradient and the myoelectric activation gradient mutate simultaneously, and generate an intensity grading result including intensity grade labels and transition candidate points.

[0095] Specifically, the system identifies window locations where both the heart rate gradient and the EMG activation gradient abruptly change, generating an intensity classification result that includes intensity level labels and candidate transition points. This process involves gradient calculation and mutation detection for the time series of heart rate percentage feature vectors and EMG activation feature vectors. The system calculates the gradient value at each time point using numerical differentiation and sets a threshold for gradient change to identify abrupt change points. When both the heart rate gradient and the EMG activation gradient exceed a preset threshold within a window, the system marks that window location as a candidate intensity transition point. These candidate points may represent a significant change in exercise intensity, such as from low to moderate intensity or from moderate to high intensity. The system further verifies and analyzes these candidate points, combining them with other physiological feature data in the multidimensional feature matrix to ultimately determine the location of the intensity transition point and generate a complete intensity classification result. This result not only includes an intensity level label for each time point or movement cycle, but also clearly identifies the moment when the intensity transition occurred, providing detailed and accurate physiological assessment information for subsequent exercise training optimization and rehabilitation program adjustments. Through this series of processing steps, the computer system can comprehensively and accurately reflect the changes in an individual's physiological load during exercise, providing strong data support for personalized exercise guidance and scientific rehabilitation treatment.

[0096] The above-mentioned method of exercise assessment based on multimodal physiological data can achieve a substantial improvement in the accuracy and reliability of exercise intensity grading through an individualized adaptive mechanism and multi-indicator collaborative decision-making. Individualized physiological baseline calibration dynamically adjusts key threshold parameters based on resting state data, which can eliminate the applicability deviation of the general formula for the group and achieve personalized adaptation of maximum heart rate and electromyographic activation. The dual-channel conversion processing of heart rate percentage and electromyographic activation can achieve standardized expression of different physiological indicators under a unified intensity dimension, solving the comparability problem of multimodal data. The dual-eigenvector intensity interval decision based on preset rules can achieve objective division of low, medium and high intensity intervals, avoiding the risk of misjudgment caused by subjective experience. The state transition point detection can accurately capture the critical state of exercise intensity through synchronous mutation analysis of heart rate gradient and electromyographic activation gradient, so as to identify the golden window period for anaerobic metabolic conversion. This technology system overcomes the three limitations of traditional intensity assessment: threshold failure caused by physiological baseline differences, the risk of misjudgment caused by single-indicator decision-making, and delayed intervention due to lags in transition point detection. It fundamentally establishes a hierarchical decision-making mechanism of "individual adaptation-multi-indicator collaboration-real-time capture", providing a high-confidence input basis for the dynamic threshold model, significantly improving the scientific nature of training load optimization and the timeliness of rehabilitation progress assessment.

[0097] In one embodiment, S5 of a motion assessment method based on multimodal physiological data provided by the present invention specifically includes the following steps:

[0098] S51: extracting and processing historical physiological data of the candidate conversion points in the intensity grading results, obtaining heart rate variability, electromyographic activation, and blood lactate gradient change data in the previous exercise cycle, and generating a historical feature set of the candidate points.

[0099] Specifically, the system retrieves historical physiological data stored in a database to extract physiological feature data from previous exercise cycles associated with the current transition candidate point. Specifically, based on the timestamp of the transition candidate point, the system traces back multiple exercise cycles and extracts heart rate variability, electromyographic activation, and blood lactate gradient change data within each cycle. This process requires the system to possess efficient data retrieval and management capabilities to ensure that the required historical data can be quickly and accurately acquired. The system can use time series matching and data alignment algorithms to synchronize the extracted historical data with the current physiological features, ensuring consistency and coherence across the temporal dimension. The resulting historical feature set for the candidate point not only contains rich historical physiological information but also provides a detailed data foundation for subsequent multimodal consistency verification. The construction of this feature set fully considers the integrity and accuracy of the data, ensuring that the historical data truly reflects the physiological changes occurring during exercise.

[0100] S52: Perform multimodal consistency verification on the candidate point historical feature set, analyze the synchronous change trend of different physiological signals in the conversion window, and generate a high-confidence candidate conversion point set.

[0101] Specifically, the system uses a multimodal data analysis algorithm to comprehensively analyze extracted historical physiological data and verify the synchronized trends of different physiological signals within the transition window. The system first extracts features from heart rate variability, electromyographic activation, and blood lactate gradient data, identifying key change points and trend characteristics for each signal. The system then evaluates whether the changes of different physiological signals within the transition window are synchronized and coordinated by calculating correlation coefficients, synchronization indices, and consistency scores between these features. For example, the system analyzes whether an increase in heart rate variability coincides with an increase in electromyographic activation and a change in the blood lactate gradient to determine whether a true transition in exercise intensity has occurred. This rigorous verification process identifies candidate transition points with high confidence and generates a set of high-confidence candidate transition points. Each transition point in this set has been verified for consistency with multimodal data, ensuring its reliability and authenticity in terms of physiological change trends, providing solid data support for subsequent individualized threshold decisions. During the verification process, the system dynamically adjusts the parameters of the verification algorithm to adapt to the physiological characteristics of different individuals and different exercise patterns, thereby improving the accuracy and adaptability of the verification.

[0102] S53: Perform adaptive threshold decision processing on the high-confidence candidate transition point set, dynamically establish individualized judgment threshold rules based on the distribution characteristics of historical data, and generate an individualized threshold model. The individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point.

[0103] Specifically, the system dynamically establishes personalized threshold rules based on extracted high-confidence historical physiological data using machine learning algorithms and statistical analysis methods. Specifically, the system performs cluster analysis and pattern recognition on the data from a set of high-confidence candidate transition points, identifying characteristic patterns that represent transitions between different exercise intensities and anaerobic metabolism. The system then dynamically adjusts the thresholds based on these characteristic patterns, combined with statistical indicators such as the mean, standard deviation, and distribution density of the individual's historical data. The system continuously iterates and optimizes threshold parameters to ensure that the threshold rules accurately reflect individual physiological characteristics and exercise responses. The resulting personalized threshold model not only accurately indicates critical states of exercise intensity but also effectively identifies transition points to anaerobic metabolism, providing a key basis for personalized exercise assessment and training guidance. During the generation of the personalized threshold model, the system continuously monitors its performance and accuracy. Through real-time feedback and adjustment mechanisms, the system ensures that the model adapts to individual physiological changes and diverse exercise scenarios, thereby enabling accurate assessment and prediction of exercise status.

[0104] In one embodiment, S6 of a motion assessment method based on multimodal physiological data provided by the present invention specifically includes the following steps:

[0105] S61: Perform real-time feature mapping processing on the motion physiological data set based on the individualized threshold model, convert the physiological signal stream into a feature space vector, and generate a real-time physiological state vector.

[0106] Specifically, the system performs real-time feature mapping on exercise physiological data sets based on a personalized threshold model. The system converts continuously acquired physiological signal streams into feature space vectors, generating real-time physiological state vectors. This process requires efficient data processing and feature extraction capabilities to ensure the conversion of signals into feature vectors is completed within milliseconds. Using a pre-set feature extraction algorithm, the system analyzes physiological signals such as heart rate, electromyography, and blood lactate in real time, extracting key features reflecting the current physiological state, such as heart rate variability, electromyography activation intensity, and blood lactate concentration gradient. These features are then mapped into a multidimensional feature space to form feature space vectors. Each vector not only contains the physiological feature value at the current moment but may also incorporate statistical features from short-term historical data, such as the mean and standard deviation over the past few seconds, to provide a more comprehensive description of the physiological state. The generation of real-time physiological state vectors provides immediate data support for subsequent threshold transition state identification, ensuring that the system can dynamically track physiological changes during exercise and respond promptly.

[0107] S62: Performing threshold conversion state recognition processing on the real-time physiological state vector, detecting the similarity relationship between the current physiological state and the critical state in the threshold model, and generating a threshold conversion warning signal.

[0108] Specifically, the system generates a threshold transition warning signal by detecting the similarity between the current physiological state and the critical state in the threshold model. This identification process is based on the preset physiological feature ranges and change patterns in the personalized threshold model. The system calculates the similarity between the real-time physiological state vector and the critical state feature vector in the model in real time. Similarity can be calculated using various methods, such as Euclidean distance, cosine similarity, or Mahalanobis distance, depending on the distribution characteristics of the physiological features and the model training results. The system compares the similarity score with a preset threshold to determine whether the current physiological state is approaching or has reached the critical state of exercise intensity. For example, if the real-time heart rate percentage and electromyographic activation percentage simultaneously approach the high-intensity threshold in the personalized threshold model, and the dynamic blood lactate gradient exceeds the predicted range of the anaerobic metabolic transition point, the system will trigger a threshold transition warning signal. This signal not only indicates a possible change in current exercise intensity but also predicts an impending intensity transition, such as a shift from aerobic to anaerobic exercise. The generation of the threshold transition warning signal provides key information for subsequent visual reports, helping users to timely adjust exercise intensity or rehabilitation training plans to optimize exercise results and prevent overtraining or sports injuries.

[0109] S63: Generate a visual report for the threshold conversion warning signal, integrate historical motion data and real-time warning status to generate a graphic report containing a heat map and optimization suggestions, and generate dynamic motion evaluation results. The dynamic motion evaluation results are used to indicate personalized training load optimization plans and rehabilitation progress evaluation reports.

[0110] Specifically, the system integrates historical exercise data and real-time warning status to generate a graphical report containing heat maps and optimization suggestions, forming a dynamic exercise assessment result. Generating this report requires the system to possess powerful data visualization and intelligent analysis capabilities. The heat map uses color coding to intuitively display the distribution of physiological load during exercise. For example, different colors are used to represent the duration and frequency of different intensity zones, helping users quickly identify areas of high load and potential fatigue accumulation during exercise. Simultaneously, the system combines real-time warning signals with historical data analysis results to provide users with personalized exercise optimization suggestions, such as adjusting exercise intensity, increasing rest time, or changing exercise patterns. These suggestions are based on in-depth analysis of the user's physiological responses and the principles of exercise physiology, and are designed to help users achieve their training goals and promote their recovery. The dynamic exercise assessment results are not only displayed to the user through a visual interface but may also be stored or transmitted as data files to professionals such as coaches and rehabilitation therapists for further evaluation and adjustment of the user's exercise and rehabilitation plans. Through this series of processing steps, the system can provide users with a more intuitive, personalized, and scientific exercise assessment service, significantly improving the effectiveness and efficiency of exercise training and rehabilitation treatment.

[0111] For example, during marathon training, runners wear physiological monitoring devices, and a computer system collects real-time physiological data such as heart rate, EMG, and blood lactate. The system generates a real-time physiological state vector based on a personalized threshold model, accurately depicting the runner's physiological load under varying paces and terrain conditions. When a runner's heart rate approaches 85% of its maximum, EMG activation exceeds 70%, and blood lactate concentration rises sharply, the system immediately issues a threshold transition warning signal. This signal not only alerts the runner that their current exercise intensity is in the high-intensity range, but also indicates that continued exercise at this level could trigger an anaerobic metabolic transition point, accelerating fatigue accumulation and impacting subsequent training or competition performance. The system's generated graphical report includes a heat map visually depicting the intensity distribution throughout the training session: red areas indicate high-intensity phases, blue areas indicate low-intensity warm-up and recovery phases, and green areas indicate moderate-intensity aerobic exercise phases. By comparing historical data, the system discovered significant fluctuations in heart rate and EMG activation in the middle and late stages of long-distance runs, suggesting that exercise intensity may need to be adjusted during this period. Recommendations include reducing the pace or increasing the frequency of fueling to maintain a stable aerobic state. At the same time, the report provides personalized training optimization suggestions based on the runners' physiological reactions and training goals, such as reasonably inserting interval running into training to improve cardiopulmonary function, while avoiding sports injuries caused by excessive training.

[0112] During swimming training, physiological signal patterns differ from those in land-based sports due to the body's horizontal position and the influence of water buoyancy. The system accurately captures changes in a swimmer's physiological characteristics across different strokes and speeds. When a swimmer's heart rate percentage reaches 75%, their electromyographic activation percentage reaches 65%, and their blood lactate concentration shows an upward trend during butterfly stroke training, the system issues a warning signal. This indicates that the swimmer may be approaching the upper limit of their aerobic metabolism and is about to enter a high-intensity anaerobic metabolism phase. The system-generated graphical report uses heat maps to visually display the physiological load of the swimmer in different strokes. The report found that the physiological load during the butterfly stroke is significantly higher than that during the freestyle and backstroke phases, suggesting that swimmers should prioritize their stroke sequence and rest intervals during training to avoid excessive local muscle fatigue. Furthermore, analysis of historical data revealed that the dynamic blood lactate gradient recovers slowly after high-intensity swimming. The system recommends increasing stretching and cool-down time after swimming to promote lactate metabolism and accelerate recovery.

[0113] In a rehabilitation training scenario, taking knee surgery rehabilitation as an example, patients undergo rehabilitation training while wearing physiological monitoring devices. The system monitors the patient's heart rate, EMG, and blood lactate levels in real time while performing knee flexion and extension exercises. When the patient's heart rate percentage reaches 60%, quadriceps EMG activation reaches 40%, and blood lactate concentration shows a slight increase, the system issues a warning signal, indicating that the patient's current exercise intensity may be approaching the upper limit of their physiological load during rehabilitation. The system-generated graphical report includes a heat map detailing the distribution of physiological load under different rehabilitation exercises. It reveals that squats and stands have a greater impact on heart rate and EMG, while isometric exercises are relatively mild. Analyzing the patient's historical rehabilitation data, the system found a gradual increase in EMG activation in the later stages of training, indicating that muscle strength is gradually recovering, but overactivation, which can cause excessive joint stress, is important. The system therefore recommends increasing the proportion of isometric exercises during rehabilitation training and reducing the number of squat and stand repetitions. This promotes muscle recovery while protecting the knee joint and preventing secondary injury caused by overtraining.

[0114] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0115] Based on the same inventive concept, embodiments of the present application further provide a multimodal physiological data-based motion assessment device 700 for implementing the aforementioned multimodal physiological data-based motion assessment method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the multimodal physiological data-based motion assessment device provided below can be found in the above-mentioned limitations of the multimodal physiological data-based motion assessment method, and will not be further elaborated here.

[0116] In an exemplary embodiment, Figure 4 As shown, a motion assessment device 700 based on multimodal physiological data is provided, comprising:

[0117] Physiological data spatiotemporal synchronization module 710, for performing spatiotemporal synchronization on multimodal physiological raw data collected by sensors, aligning the sampling timestamps of heart rate, electromyography, and blood lactate signals through a hardware clock protocol, and generating a motion physiological dataset. The motion physiological dataset includes time-aligned heart rate waveforms, electromyography time-domain signals, and continuous blood lactate concentration curves.

[0118] The physiological signal denoising processing module 720 is used to perform noise filtering on the sports physiological data set to remove the power frequency interference signal of the acquisition equipment and the electromyographic artifact noise in the sports environment, thereby generating a denoised physiological data set with enhanced signal quality;

[0119] A motion feature extraction module 730 is used to extract motion features from the denoised physiological data set, calculate heart rate variability, myoelectric root mean square value, and blood lactate concentration gradient change rate within a sliding window, and generate a multidimensional feature matrix containing time-varying physiological indicators;

[0120] An exercise intensity grading module 740 is used to grade exercise intensity based on the multidimensional feature matrix, dividing the intensity into low, medium, and high intensity intervals according to preset heart rate percentage thresholds and myoelectric activation thresholds, and generating an intensity grading result including intensity level labels and candidate conversion points;

[0121] Individualized threshold modeling module 750 is used to perform dynamic threshold detection on intensity grading results, perform multimodal consistency verification and adaptive threshold decision-making on candidate transition points based on historical data, and generate an individualized threshold model. The individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point;

[0122] The dynamic motion evaluation module 760 is used to process the motion physiological data set based on the individualized threshold model, output the motion intensity heat map and threshold conversion warning report in real time, and generate dynamic motion evaluation results. The dynamic motion evaluation results are used to indicate the training load optimization plan and rehabilitation progress evaluation report.

[0123] In summary, the motion assessment device based on multimodal physiological data provided by the present application can achieve a systematic breakthrough in motion assessment technology through the deep fusion and dynamic adaptation mechanism of multimodal physiological data. At the data processing level, the spatiotemporal synchronization mechanism based on the hardware clock protocol can eliminate the time misalignment problem of multi-source signal acquisition and achieve high-precision alignment of the original data; the adaptive filtering technology can achieve a substantial improvement in the signal-to-noise ratio by synergistically suppressing power frequency interference and motion artifacts. At the feature analysis level, the joint calculation of heart rate variability, myoelectric root mean square and blood lactate gradient within the sliding window can achieve a comprehensive quantification of exercise load characteristics; individualized calibration of heart rate percentage and myoelectric activation threshold rules are used to achieve physiological adaptability optimization of intensity grading. At the core decision-making level, the multimodal consistency verification mechanism based on historical data can reduce the misjudgment rate of transition point identification; the adaptive threshold decision model dynamically tracks individual physiological baseline drift to achieve real-time capture of the critical state of exercise intensity. The resulting integrated system of heatmaps and early warning reports enables visual monitoring of anaerobic metabolic transition points, providing a basis for intensity allocation decisions for training load optimization and establishing a quantitative feedback loop for rehabilitation progress assessment. This device overcomes three major limitations of traditional static movement assessment: inadequate collaborative analysis due to insufficient multimodal data fusion, lack of individual adaptation caused by fixed threshold rules, and delayed intervention due to delayed assessment results. This fundamentally improves the accuracy and timeliness of sports science decision-making.

[0124] Preferably, the physiological data spatiotemporal synchronization module 710 provided in this application is configured with the following units:

[0125] The heterogeneous data receiving unit is used to perform heterogeneous data reception and processing on the original physiological signals collected by multiple sensors, synchronously obtain the original signals output by the heart rate monitoring device, electromyography sensor, blood lactate analyzer and respiratory monitoring device, and generate the original heterogeneous physiological data set;

[0126] The time base alignment unit is used to perform time base alignment processing on the original heterogeneous physiological data set, and uses the hardware clock synchronization protocol to assign a unified time base tag to all data streams of the original heterogeneous physiological data set to generate a time synchronized data set;

[0127] The blood lactate signal reconstruction unit is used to reconstruct the blood lactate signal of the time-synchronized data set, convert the intermittent sampling data into a continuous time series, and generate a sports physiological data set. The sports physiological data set includes a time-aligned heart rate waveform sequence, an electromyographic signal waveform sequence, a respiratory parameter sequence, and a blood lactate concentration sequence in the continuous time domain.

[0128] Preferably, the physiological signal denoising processing module 720 provided in this application is configured with the following units:

[0129] The power frequency interference elimination unit is used to perform power frequency interference elimination processing on the exercise physiological data set based on the adaptive notch filtering technology, filter out the 50Hz power line interference component in the ECG signal, and generate a preliminary filtered data set;

[0130] a motion artifact suppression unit, configured to perform motion artifact suppression processing on the preliminary filtered data set, construct a limb motion-noise correlation model, and inversely compensate for the distortion of the electromyographic signal waveform to generate a motion artifact suppression data set. The motion artifact suppression data set is used to indicate the muscle activation state after removing limb motion interference;

[0131] The environmental noise filtering unit is used to perform environmental noise filtering on the motion artifact suppression data set, filter out the environmental noise component in the respiratory signal, and generate a denoised physiological data set.

[0132] Preferably, the motion feature extraction module 730 provided in this application is configured with the following units:

[0133] The motion cycle segmentation unit is used to perform motion cycle segmentation processing on the denoised physiological data set, automatically divide the action cycle boundaries based on the joint kinematic characteristics, and generate segmented motion physiological data;

[0134] A heart rate variability calculation unit is used to calculate and process the segmented exercise physiological data, quantify the cardiac autonomic nervous system regulation state by analyzing the standard deviation of the continuous RR interval sequence, and generate a heart rate variability feature sequence;

[0135] The myoelectric activation intensity calculation unit is used to calculate the myoelectric root mean square value of the segmented motion physiological data, quantify the amplitude intensity of the myoelectric signal in the sliding window through integration operation, and generate a myoelectric activation intensity sequence;

[0136] The blood lactate dynamic gradient calculation unit is used to calculate and process the blood lactate concentration gradient of segmented exercise physiological data, reflect the anaerobic metabolic state through the concentration difference change rate of adjacent sampling points, and generate a blood lactate dynamic gradient sequence;

[0137] The multidimensional feature integration unit is used to perform matrix integration processing on the heart rate variability feature sequence, the myoelectric activation intensity sequence and the blood lactate dynamic gradient sequence to generate a multidimensional feature matrix.

[0138] Preferably, the exercise intensity grading module 740 provided in this application is configured with the following units:

[0139] Individualized physiological baseline calibration unit, used to perform individualized physiological baseline calibration processing on the multi-dimensional feature matrix, dynamically adjust the maximum heart rate and maximum electromyographic activation reference values ​​based on resting physiological data, and generate personalized intensity threshold parameters;

[0140] A heart rate percentage conversion unit is used to perform heart rate percentage conversion processing on the multidimensional feature matrix based on the personalized intensity threshold parameter, calculate the percentage of the current heart rate to the personalized maximum heart rate, and generate a heart rate percentage feature vector;

[0141] The myoelectric activation conversion unit is used to perform myoelectric activation conversion processing on the multidimensional feature matrix based on the personalized intensity threshold parameter, calculate the percentage of the current myoelectric root mean square value to the personalized maximum myoelectric activation, and generate a myoelectric activation feature vector;

[0142] An intensity interval decision unit is used to perform intensity interval decision processing on the heart rate percentage feature vector and the myoelectric activation feature vector based on preset heart rate percentage threshold and myoelectric activation threshold rules, divide the heart rate percentage feature vector and the myoelectric activation feature vector into low intensity, medium intensity and high intensity level intervals, and generate a preliminary intensity label set;

[0143] The state transition point detection unit is used to perform state transition point detection processing on the preliminary intensity label set, identify the window position where the heart rate gradient and the myoelectric activation gradient change simultaneously, and generate an intensity grading result including intensity level labels and transition candidate points.

[0144] Preferably, the individualized threshold modeling module 750 provided in this application is configured with the following units:

[0145] A historical feature extraction unit is used to extract and process historical physiological data of the candidate conversion points in the intensity grading results, obtain heart rate variability, electromyographic activation, and blood lactate gradient change data in the previous exercise cycle, and generate a historical feature set of the candidate points;

[0146] The multimodal consistency verification unit is used to perform multimodal consistency verification on the historical feature set of the candidate points, analyze the synchronous change trend of different physiological signals in the conversion window, and generate a high-confidence candidate conversion point set;

[0147] The adaptive threshold decision unit is used to perform adaptive threshold decision processing on a set of high-confidence candidate transition points, dynamically establish individualized judgment threshold rules based on the distribution characteristics of historical data, and generate an individualized threshold model. The individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point.

[0148] Preferably, the dynamic motion assessment module 760 provided in this application is configured with the following units:

[0149] A real-time physiological state mapping unit is used to perform real-time feature mapping processing on the motion physiological data set based on an individualized threshold model, convert the physiological signal stream into a feature space vector, and generate a real-time physiological state vector;

[0150] A threshold transition recognition unit is used to perform threshold transition state recognition processing on the real-time physiological state vector, detect the similarity relationship between the current physiological state and the critical state in the threshold model, and generate a threshold transition warning signal;

[0151] The dynamic assessment report generation unit is used to generate a visual report for the threshold conversion warning signal, integrate historical motion data and real-time warning status to generate a graphic report containing a heat map and optimization suggestions, and generate dynamic motion assessment results. The dynamic motion assessment results are used to indicate personalized training load optimization plans and rehabilitation progress assessment reports.

[0152] In one embodiment, the present application further provides a computer device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned motion assessment method based on multimodal physiological data when executing the computer program.

[0153] In one embodiment, the present application further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the computer program implements the above-mentioned motion assessment method based on multimodal physiological data.

[0154] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and integrate different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0155] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0156] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any person skilled in the art can easily conceive of various modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should be included within the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A motion assessment method based on multimodal physiological data, characterized in that: The following steps are involved: S1: Performing spatiotemporal synchronization on multimodal physiological raw data collected by sensors, aligning the sampling timestamps of heart rate, electromyography, and blood lactate signals through a hardware clock protocol, and generating a sports physiological dataset. The sports physiological dataset includes time-aligned heart rate waveforms, electromyography time-domain signals, and continuous blood lactate concentration curves. S2: performing noise filtering on the sports physiological dataset to remove power frequency interference signals of the acquisition equipment and myoelectric artifact noise in the sports environment, thereby generating a denoised physiological dataset with enhanced signal quality; S3: extracting motion features from the denoised physiological data set, calculating heart rate variability, myoelectric root mean square value, and blood lactate concentration gradient change rate within a sliding window, and generating a multidimensional feature matrix containing time-varying physiological indicators; S4: performing exercise intensity classification on the multidimensional feature matrix, dividing the exercise intensity into low, medium, and high intensity intervals according to preset heart rate percentage thresholds and myoelectric activation thresholds, and generating an intensity classification result including intensity level labels and conversion candidate points; S5: Performing dynamic threshold detection on the intensity grading result, performing multimodal consistency verification and adaptive threshold decision on the candidate transition points based on historical data, and generating an individualized threshold model, wherein the individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point; S6: Processing the exercise physiological data set based on the individualized threshold model, outputting an exercise intensity heat map and a threshold conversion warning report in real time, and generating a dynamic exercise evaluation result, which is used to indicate a training load optimization plan and a rehabilitation progress evaluation report.

2. The method according to claim 1, characterized in that Said S1 comprises: S11: Perform heterogeneous data reception and processing on the original physiological signals collected by multiple sensors, synchronously obtain the original signals output by the heart rate monitoring device, electromyography sensor, blood lactate analyzer and respiratory monitoring device, and generate an original heterogeneous physiological data set; S12: performing time reference alignment processing on the original heterogeneous physiological data set, assigning a unified time reference tag to all data streams of the original heterogeneous physiological data set using a hardware clock synchronization protocol, and generating a time synchronized data set; S13: Perform blood lactate signal reconstruction processing on the time-synchronized dataset, convert the intermittent sampling data into a continuous time series, and generate a sports physiological dataset, wherein the sports physiological dataset includes a time-aligned heart rate waveform sequence, an electromyographic signal waveform sequence, a respiratory parameter sequence, and a blood lactate concentration sequence in the continuous time domain.

3. The method according to claim 1, characterized in that The S2 includes: S21: performing power frequency interference elimination processing on the exercise physiological data set based on the adaptive notch filtering technology, filtering out the 50 Hz power line interference component in the electrocardiogram signal, and generating a preliminary filtered data set; S22: performing motion artifact suppression processing on the preliminary filtered data set, constructing a limb motion-noise correlation model and inversely compensating for electromyographic signal waveform distortion to generate a motion artifact suppressed data set, wherein the motion artifact suppressed data set is used to indicate the muscle activation state after removing limb motion interference; S23: Performing environmental noise filtering processing on the motion artifact suppression dataset to filter out environmental noise components in the respiratory signal to generate a denoised physiological dataset.

4. The method according to claim 1, wherein The S3 includes: S31: performing motion cycle segmentation processing on the denoised physiological data set, automatically dividing the action cycle boundaries based on joint kinematic characteristics, and generating segmented motion physiological data; S32: performing heart rate variability calculation processing on the segmented exercise physiological data, quantifying the cardiac autonomic nervous system regulation state by analyzing the standard deviation of a continuous RR interval sequence, and generating a heart rate variability characteristic sequence; S33: Calculate the myoelectric root mean square value of the segmented motion physiological data, quantify the amplitude strength of the myoelectric signal in the sliding window through integration operation, and generate an electromyographic activation intensity sequence. The calculation formula of the electromyographic activation intensity sequence is: Among them, RMS is the myoelectric activation intensity sequence, that is, the root mean square value of myoelectricity, T is the sliding window length, x t is the electromyographic signal value at time t; S34: performing blood lactate concentration gradient calculation processing on the segmented exercise physiological data, reflecting the anaerobic metabolic state through the concentration difference change rate of adjacent sampling points, and generating a blood lactate dynamic gradient sequence; S35: Performing matrix integration processing on the heart rate variability feature sequence, the myoelectric activation intensity sequence, and the blood lactate dynamic gradient sequence to generate a multidimensional feature matrix.

5. The method according to claim 1, wherein The S4 includes: S41: performing individualized physiological benchmark calibration processing on the multidimensional feature matrix, dynamically adjusting the maximum heart rate and maximum electromyographic activation reference values ​​based on resting physiological data, and generating personalized intensity threshold parameters; S42: performing heart rate percentage conversion processing on the multidimensional feature matrix based on the personalized intensity threshold parameter, calculating the percentage of the current heart rate to the personalized maximum heart rate, and generating a heart rate percentage feature vector; S43: performing myoelectric activation conversion processing on the multidimensional feature matrix based on the personalized intensity threshold parameter, calculating the percentage of the current myoelectric root mean square value to the personalized maximum myoelectric activation, and generating a myoelectric activation feature vector; S44: performing intensity interval decision processing on the heart rate percentage feature vector and the myoelectric activation degree feature vector based on preset heart rate percentage threshold and myoelectric activation degree threshold rules, dividing the heart rate percentage feature vector and the myoelectric activation degree feature vector into low intensity, medium intensity and high intensity level intervals, and generating a preliminary intensity label set; S45: Perform state transition point detection processing on the preliminary intensity label set, identify the window position where the heart rate gradient and the myoelectric activation gradient suddenly change at the same time, and generate an intensity grading result including intensity grade labels and transition candidate points.

6. The method according to claim 1, characterized in that The S5 includes: S51: extracting historical physiological data from candidate conversion points in the intensity grading result to obtain heart rate variability, electromyographic activation, and blood lactate gradient change data during the preceding exercise cycle, and generating a historical feature set of candidate points; S52: performing multimodal consistency verification processing on the candidate point historical feature set, analyzing the synchronous change trend of different physiological signals in the conversion window, and generating a high-confidence candidate conversion point set; S53: Performing adaptive threshold decision processing on the high-confidence candidate transition point set, dynamically establishing individualized determination threshold rules based on historical data distribution characteristics, and generating an individualized threshold model, wherein the individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point.

7. The method according to any one of claims 1 to 6, characterized in that The S6 includes: S61: performing real-time feature mapping processing on the motion physiological data set based on the individualized threshold model, converting the physiological signal stream into a feature space vector, and generating a real-time physiological state vector; S62: performing threshold conversion state recognition processing on the real-time physiological state vector, detecting the similarity relationship between the current physiological state and the critical state in the threshold model, and generating a threshold conversion warning signal; S63: Perform visual report generation processing on the threshold conversion warning signal, integrate historical motion data and real-time warning status to generate a graphic report containing a heat map and optimization suggestions, and generate dynamic motion evaluation results. The dynamic motion evaluation results are used to indicate a personalized training load optimization plan and a rehabilitation progress evaluation report.

8. A motion assessment device based on multimodal physiological data, characterized in that: The device comprises: A physiological data spatiotemporal synchronization module is used to perform spatiotemporal synchronization on the multimodal physiological raw data collected by sensors, align the sampling timestamps of heart rate, electromyography, and blood lactate signals through a hardware clock protocol, and generate a sports physiological data set. The sports physiological data set includes time-aligned heart rate waveforms, electromyography time-domain signals, and continuous blood lactate concentration curves; A physiological signal denoising processing module is used to perform noise filtering on the sports physiological data set to remove the power frequency interference signal of the acquisition equipment and the myoelectric artifact noise in the sports environment, thereby generating a denoised physiological data set with enhanced signal quality; a motion feature extraction module, configured to extract motion features from the denoised physiological data set, calculate heart rate variability, myoelectric root mean square value, and blood lactate concentration gradient change rate within a sliding window, and generate a multidimensional feature matrix containing time-varying physiological indicators; An exercise intensity grading module is used to grade the exercise intensity of the multidimensional feature matrix, divide the exercise intensity into low, medium and high intensity intervals according to preset heart rate percentage thresholds and myoelectric activation thresholds, and generate an intensity grading result including intensity level labels and conversion candidate points; An individualized threshold modeling module is used to perform dynamic threshold detection on the intensity grading results, perform multimodal consistency verification and adaptive threshold decision-making on candidate transition points based on historical data, and generate an individualized threshold model. The individualized threshold model is used to indicate the critical state of exercise intensity and the anaerobic metabolism transition point; The dynamic motion assessment module is used to process the motion physiological data set based on the individualized threshold model, output the motion intensity heat map and threshold conversion warning report in real time, and generate dynamic motion assessment results. The dynamic motion assessment results are used to indicate the training load optimization plan and rehabilitation progress assessment report.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images