Qinggong lijin manipulation evaluation method and system based on multi-modal dynamics fusion

Abstract

The application discloses a palace tendon method evaluation method and system based on multi-modal dynamics fusion, and relates to the technical field of massage method evaluation. Including the following steps: S1, collecting multi-modal hand dynamics data of operators and professional tendon technicians, constructing a reference sample library, and carrying out data preprocessing; S2, reconstructing the force transfer process, extracting the main direction representation target force direction, judging the stability of force flow, and generating force flow characteristic record; S3, identifying the force exertion sequence of wrist, palm and finger, combining the reference sample library to form a typical transmission rhythm, judging the time sequence integrity of the current force chain, and identifying the abnormal chain segment; S4, comprehensively evaluating the strength performance, grading the strength, and generating targeted training prompts; at the same time, updating the threshold parameters based on the historical excellent sample distribution. The problems that the traditional analysis technology can only record the changes of pressure and displacement, and cannot depict the time sequence and strength conduction state of multi-segment force chain are solved.

Description

Technical Field

[0001] This invention relates to the field of massage technique evaluation technology, specifically to a method and system for evaluating the Qinggong Lijin technique based on multimodal dynamic fusion. Background Technology

[0002] Traditional Chinese massage plays a vital role in traditional conditioning, muscle and bone relaxation, and meridian unblocking. Its core emphasis is on the flow of force, rhythmic connection, and the pervasive sensation of pressure during manipulation, which are considered important indicators of skill level. With the increasing demands for modern TCM inheritance, teaching demonstrations, and skills assessments, the industry is gradually focusing on the objective presentation and systematic expression of the manipulation process, hoping to more systematically organize and standardize the characteristics of force changes, transmission rhythm, and muscle coordination during massage. Driven by the trends of digitalization and intelligentization, constructing an evaluation system that can assist in judging the expression of force, support teaching and training, and promote the inheritance of experience has become an important direction for the development of the TCM massage field.

[0003] For example, the invention patent with publication number CN113935623A discloses a performance management system for traditional Chinese medicine acupuncture and massage education, belonging to the field of traditional Chinese medicine acupuncture. It addresses the problems of manual operation and publicly available scores in performance management for traditional Chinese medicine acupuncture and massage education, and the lack of adaptive adjustments to performance evaluation standards based on actual student performance. The system includes a verification and adjustment module and a performance calculation module. The performance calculation module feeds back the performance level of student terminals to the server and teacher terminals. The teacher terminal verifies the performance level of the student terminal. After verification, the server sends the performance level to the corresponding student terminal. The verification and adjustment module is used to verify and adjust the performance evaluation of student terminals, generating a performance evaluation adjustment signal or a normal performance evaluation signal. This invention effectively avoids human manipulation of performance scores in traditional Chinese medicine acupuncture and massage education, adopts a performance level system, and allows performance evaluation standards to be adaptively adjusted based on actual student performance.

[0004] However, the Qing Dynasty's massage techniques for regulating tendons and muscles primarily emphasize the force, not simply the amount of pressure, but the rhythmic pattern of force accumulation and release, the path of the pushing force transmitted segment by segment through the palm, fingertips, and joint chains, and the dynamic force vector migration of the point of application position with changes in micro-movements. However, existing systems mainly collect force curves and trajectory curves, which are difficult to express whether the force is correctly transmitted from the wrist to the palm and then to the fingers, whether the force release is excessively premature or delayed, and which joint chain has a break in force or blockage of force. These are the key capabilities of the traditional Chinese massage system, which are difficult to describe using traditional techniques.

[0005] Therefore, in order to address the above problems, there is an urgent need for an evaluation method and system for the Qinggong Lijin manipulation technique based on multimodal dynamics fusion. Summary of the Invention

[0006] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an evaluation method and system for Qinggong Lijin manipulation techniques based on multimodal dynamics fusion, which solves the problem that traditional analysis techniques can only record pressure and displacement changes and cannot characterize the timing and force transmission state of multi-segment force chains.

[0007] Technical solution To achieve the above objectives, this invention provides the following technical solution: a method for evaluating Qing Dynasty-style hand manipulation techniques based on multimodal dynamics fusion, comprising the following steps: S1, collecting multimodal hand dynamics data from operators and professional hand manipulation technicians, constructing a reference sample library using the multimodal hand dynamics data from professional hand manipulation technicians, and preprocessing the multimodal hand dynamics data; S2, reconstructing the thrust migration process based on the preprocessed multimodal hand dynamics data, characterizing the target force direction using the principal direction extraction method, and analyzing the consistency of velocity direction and the change in jitter amplitude. S3. Utilize preprocessed multimodal hand dynamics data to identify the sequential relationship of force application in the wrist, palm, and fingers, and combine this with a reference sample library to form a typical transmission rhythm, determine the temporal integrity of the current force chain, and identify the chain segment to which anomalies belong; S4. Comprehensively analyze the preprocessed multimodal hand dynamics data, thrust flow stability, and temporal integrity of the force chain to classify the operator's strength performance and generate targeted training prompts; simultaneously, update threshold parameters based on the distribution of historical excellent samples to form a closed-loop optimization feedback.

[0008] Furthermore, the process of collecting multimodal hand dynamics data from operators and professional muscle-relaxing technicians, and constructing a reference sample library using the multimodal hand dynamics data from professional muscle-relaxing technicians, is as follows: The process from the start to the end of a movement is defined as a manipulation cycle. Within each manipulation cycle, multimodal hand dynamics data of the operator is collected in real time using a smart sensing glove. The multimodal hand dynamics data includes: triaxial acceleration of the hand, triaxial angular velocity of the hand, pressure values ​​of each sampling unit of the tactile array, electromyographic envelope signals of superficial muscle groups, and electromyographic envelope signals of deep muscle groups. At the same time, multimodal hand dynamics data of professional muscle-relaxing technicians within the manipulation cycle are collected to construct a reference sample library. Specifically, when the triaxial acceleration of the hand exceeds the motion threshold and is accompanied by activation of the electromyographic envelope signal, the movement is determined to have started. When the triaxial acceleration, angular velocity, pressure value, and electromyographic envelope signal of the hand all fall back to the resting threshold and continue to exceed the resting time window, the movement is determined to have ended.

[0009] Furthermore, the specific process of data preprocessing for multimodal hand dynamics data is as follows: multimodal time synchronization is performed based on the sampling timestamps of each channel, and the multimodal hand dynamics data are uniformly mapped to the same time axis using hardware clock alignment; low-pass filtering and drift compensation are applied to the three-axis acceleration and three-axis angular velocity of the hand; spatial convolution smoothing is used to suppress isolated noise for the pressure values ​​of each sampling unit of the tactile array, and the contact center coordinates and contact area are calculated using a two-dimensional centroid algorithm based on the pressure values, and the contact center coordinates are linearly interpolated to obtain the continuous contact center trajectory; effective electromyographic envelope signals of superficial and deep muscle groups are obtained through bandpass filtering, rectification, and smoothing; Z-score normalization is performed on the hand dynamics data; a manipulation assessment database is established, in which the original and preprocessed multimodal hand dynamics data of the operator, the reference sample library, the corresponding continuous contact center trajectory, and the contact area are written into the manipulation assessment database.

[0010] Furthermore, based on the preprocessed multimodal hand dynamics data, the thrust migration process is reconstructed. The principal direction extraction method is used to characterize the target force direction. The specific process for analyzing the consistency of velocity direction and the variation in jitter amplitude to determine the thrust flow stability is as follows: 1) Obtain the continuous contact center trajectory within the hand movement cycle; 2) Calculate the instantaneous velocity vector of the contact center using the finite difference method and calculate the instantaneous acceleration amplitude; 3) Read the corresponding continuous contact center trajectory from the reference sample library; 4) Extract the principal direction vector of the trajectory using principal component analysis and normalize the principal direction vector to obtain the unit vector of the target force direction; 5) Select the contact center trajectory from the reference sample library... The median of the instantaneous acceleration amplitude within the method cycle is used as the reference value of the contact center acceleration. For each sampling moment of the operator within the method cycle, the cosine of the angle between the instantaneous velocity vector of the contact center and the unit vector of the target force direction is calculated, and the cosine of the angle is non-negative to obtain the directional consistency value. The instantaneous acceleration amplitude is divided by the contact center acceleration reference value, and the opposite number is taken as the exponent for natural exponentiation to obtain the jitter penalty value. The directional consistency value is multiplied by the jitter penalty value, and the product of all sampling moments is accumulated point by point and divided by the total number of sampling points to obtain the force vector flow continuity value.

[0011] Furthermore, the specific process for generating force flow characteristic records is as follows: if the force vector flow continuity value is less than the smoothness threshold, it is determined that there is a smoothness anomaly in the current technique, and the abnormal period is marked; based on the directional consistency value at each sampling time within the technique cycle, segments where the thrust direction deviates from the target force direction are identified, and segments that continuously deviate for more than the allowable time threshold are marked as directional offset segments; at the same time, using the jitter penalty value sequence, the jitter intensity of each sampling point within the technique cycle is quantified to generate a jitter heat map; the force vector flow continuity value, abnormal period, and directional offset segment are written into the technique evaluation database.

[0012] Furthermore, the specific process of identifying the sequential relationship of force application in the wrist, palm, and fingers using preprocessed multimodal hand dynamics data, and forming a typical transmission rhythm by combining it with a reference sample library, to determine the temporal integrity of the current force chain is as follows: Amplitude threshold detection and rising edge search are performed on the electromyographic envelope signals of superficial and deep muscle groups within the hand technique cycle to identify the rising time points of wrist, palm, and fingers. Based on these three time points, the wrist-to-palm force application time difference and the palm-to-finger force application time difference are calculated. The wrist-to-palm force application time difference and the palm-to-finger force application time difference are calculated for N hand technique cycles in the reference sample library, and the median is selected to obtain the wrist-to-palm transmission time reference value and the palm-to-finger transmission time reference value. The following steps are performed: First, the wrist-to-palm force transfer reference value is subtracted from the wrist-to-palm force transfer reference value during the current technique cycle to obtain the wrist-to-palm directional time deviation. Second, the palm-to-finger force transfer reference value is subtracted from the palm-to-finger force transfer time difference during the current technique cycle to obtain the palm-to-finger directional time deviation. Third, the wrist-to-palm and palm-to-finger directional time deviations are squared, and the sum of the two squares is taken as the square root to obtain the composite time deviation. Fourth, the square root of the sum of the squares of the wrist-to-palm and palm-to-finger force transfer reference values ​​is taken as the reference time scale value. Finally, the ratio of the composite time deviation to the reference time scale value is subtracted from a constant to obtain the chain force application sequence consistency value.

[0013] Furthermore, the specific process for identifying the abnormal chain segment is as follows: If the consistency value of the chain force generation sequence within the current technique cycle is lower than the continuity threshold, it is determined that there is an abnormal force transmission, and the force generation chain segment where the abnormal link is located is identified: Compare the absolute value of the time deviation in the wrist-to-palm direction with the absolute value of the time deviation in the palm-to-finger direction. If the absolute value of the time deviation in the wrist-to-palm direction is larger, it is marked as a wrist-to-palm transmission abnormality; if the absolute value of the time deviation in the palm-to-finger direction is larger, it is marked as a palm-to-finger transmission abnormality; if the absolute values ​​of both time deviations are greater than the corresponding deviation thresholds, it is marked as a multi-segment chain transmission abnormality; write the chain force generation sequence consistency value, the wrist-to-palm direction time deviation, the palm-to-finger direction time deviation, and the corresponding abnormal markings into the technique evaluation database; if a continuous technique cycle is determined to have force transmission abnormalities exceeding the cycle allowable threshold, a chain transmission instability prompt is generated, and information on the abnormal force generation chain segment is pushed and targeted training reminders are executed.

[0014] Further, the specific process of comprehensively analyzing the preprocessed multi-modal hand dynamics data, thrust flow stability and timing integrity of the force generation chain is as follows: acquiring the surface muscle group electromyographic envelope signal and the deep muscle group electromyographic envelope signal within the manipulation cycle, and performing time integration respectively to obtain the surface muscle energy value and the deep muscle energy value, and dividing the deep muscle energy value by the surface muscle energy value to obtain the deep-surface muscle participation energy ratio; meanwhile, reading the reference sample library, calculating the deep-surface muscle participation energy ratios of N manipulation cycles in the reference sample library, and screening out the median as the standard reference ratio; dividing the operator's deep-surface muscle participation energy ratio within the manipulation cycle by the standard reference ratio, and taking the square root to obtain the normalized deep-surface muscle energy ratio, multiplying the normalized deep-surface muscle energy ratio by the chain force generation sequence consistency value and the force vector flow continuity value, and taking the cube root to obtain the comprehensive force assessment value.

[0015] Further, the specific process of grading the operator's force performance, generating targeted training prompts, and updating threshold parameters based on the distribution of historical excellent samples to form closed-loop optimization feedback is as follows: the comprehensive force assessment value is compared with the multi-level force thresholds T1 and T2: when <T1, it is determined that the force is insufficient, and the corresponding abnormal force generation chain segment, direction deviation segment and jitter heat peak area are displayed on the feedback interface; a training scheme focusing on basic force coherence is recommended, and prompts to increase exercises targeting hand force generation connection, thrust direction correction and deep-surface force generation coordination; when T1≤ <T2, it is determined that the force is good, a prompt indicating that the overall force is good is given on the feedback interface, and advanced training is recommended; when ≥T2, it is determined that the force is excellent, a prompt indicating that the force is excellent is given on the feedback interface, the current manipulation cycle is marked as an excellent sample, and the corresponding multi-modal hand dynamics data, chain force generation sequence consistency value, force vector flow continuity value and deep-surface muscle participation energy ratio are written into the individual excellent sample library, and complex manipulation combination training is recommended; the comprehensive force assessment value, force grading result, corresponding prompt information and individual excellent sample library are written into the manipulation assessment database, and the operator's force training feedback report is generated; meanwhile, according to the sample distribution recorded in the individual excellent sample library, an adaptive update algorithm based on quantile statistics is regularly adopted to adjust the standard reference ratio and the coherence threshold.

[0016] The second aspect of this invention provides an evaluation system for Qing Dynasty-style hand manipulation techniques based on multimodal dynamics fusion, comprising: a multimodal data acquisition and preprocessing module for acquiring multimodal hand dynamics data from operators and professional hand manipulation technicians, constructing a reference sample library using the multimodal hand dynamics data from professional hand manipulation technicians, and preprocessing the multimodal hand dynamics data; and a three-dimensional motion and force vector fusion reconstruction module for reconstructing the thrust migration process based on the preprocessed multimodal hand dynamics data, characterizing the target force direction using a principal direction extraction method, and analyzing the consistency of velocity direction and the change in jitter amplitude to determine the stability of thrust flow. The system analyzes and generates force flow characteristic records; the force chain dynamics analysis module uses preprocessed multimodal hand dynamics data to identify the sequential relationship of force application in the wrist, palm, and fingers, and combines it with a reference sample library to form a typical transmission rhythm, determine the temporal integrity of the current force chain, and identify the chain segment to which anomalies belong; the technique comparison and force evaluation module comprehensively analyzes the preprocessed multimodal hand dynamics data, thrust flow stability, and temporal integrity of the force chain, classifies the operator's force performance, and generates targeted training prompts; at the same time, it updates threshold parameters based on the distribution of historical excellent samples to form a closed-loop optimization feedback.

[0017] Beneficial effects The present invention has the following beneficial effects: (1) This invention, by synchronously collecting multimodal hand dynamics data and combining it with professional technician samples to construct a reference sample library, enables the key dynamic characteristics of the Qing Palace Li Jin Technique to be recorded in a unified and structured manner, providing a quantifiable basis for subsequent action reproduction, comparison and evaluation.

[0018] (2) This invention extracts the target force direction based on the continuous contact center trajectory, and analyzes the consistency of velocity direction and acceleration disturbance mode to objectively determine the continuity, directional deviation and jitter of the thrust on the force path, thereby quantitatively reflecting the flow stability of the force transmission process.

[0019] (3) This invention utilizes the rising edge of electromyography to detect the sequence of force exertion in the wrist, palm, and fingers, and combines it with typical rhythms obtained from the reference sample library to construct a temporal integrity index. This index identifies cases of premature, delayed, or broken chains in the force exertion chain, accurately locating abnormal transmission links and supporting targeted training improvements.

[0020] (4) This invention achieves multi-level grading of force performance by comprehensively quantifying the stability of force flow, the consistency of chain force generation sequence and the energy relationship between deep and superficial muscles; and adaptively updates key thresholds based on the distribution characteristics of excellent samples, so that the evaluation criteria are continuously optimized as the usage process is carried out, forming a training feedback loop that can be continuously improved.

[0021] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0022] Figure 1 The flowchart is for the evaluation method of Qing Dynasty uterine muscle manipulation based on multimodal dynamic fusion. Figure 2 This is a structural diagram of the Qing Dynasty Palace Meridian Lisi Manipulation Evaluation System based on Multimodal Dynamics Fusion; Figure 3 This is a jitter heat map based on jitter penalty values; Figure 4 This is a trend comparison chart of the comprehensive strength assessment values. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. As those skilled in the art will understand, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figures 1-4 This invention provides a technical solution: a method and system for evaluating the Qinggong Lijin manipulation technique based on multimodal dynamics fusion, such as... Figure 1 As shown, the process includes the following steps: S1, collecting multimodal hand dynamics data from operators and professional muscle-relaxing technicians, constructing a reference sample library using the multimodal hand dynamics data from professional muscle-relaxing technicians, and preprocessing the multimodal hand dynamics data; S2, reconstructing the thrust migration process based on the preprocessed multimodal hand dynamics data, characterizing the target force direction using the main direction extraction method, analyzing the consistency of velocity direction and the change in jitter amplitude to determine the thrust flow stability, and generating force flow characteristic records; S3, identifying the sequential relationship of force application of the wrist, palm, and fingers using the preprocessed multimodal hand dynamics data, forming a typical transmission rhythm in combination with the reference sample library, judging the temporal integrity of the current force chain, and identifying the chain segment to which the abnormality belongs; S4, comprehensively analyzing the preprocessed multimodal hand dynamics data, thrust flow stability, and temporal integrity of the force chain, classifying the operator's strength performance, and generating targeted training prompts; at the same time, updating the threshold parameters based on the distribution of historical excellent samples to form a closed-loop optimization feedback.

[0025] Specifically, the process of collecting multimodal hand dynamics data from operators and professional muscle relaxants, and constructing a reference sample library using the multimodal hand dynamics data from professional muscle relaxants, is as follows: The process from the start to the end of a movement is defined as a manipulation cycle. Within each manipulation cycle, multimodal hand dynamics data of the operator is collected in real time using a smart sensing glove. This multimodal hand dynamics data includes: triaxial hand acceleration, triaxial hand angular velocity, pressure values ​​from each sampling unit of the tactile array, electromyographic envelope signals of superficial and deep muscle groups; among these, the triaxial hand acceleration data is collected by a miniature inertial measurement unit deployed on the back of the hand. IMU acquisition; hand triaxial angular velocity data are synchronously output from the gyroscope channel within the same IMU to reflect changes in wrist and hand posture; pressure values ​​of each sampling unit of the tactile array are sensed in real time by flexible pressure sensor arrays integrated on the palm, fingertips, and finger sides, and each pixel-level pressure value is output in a point-by-point scanning manner; superficial muscle group electromyographic envelope signals are acquired by surface electromyographic electrodes placed on the surface of superficial forearm muscles, and the envelope signals are obtained after pre-amplification, rectification, and smoothing; deep muscle group electromyographic envelope signals are acquired by high-impedance surface electromyographic electrodes attached to the near-deep muscle area of ​​the forearm, and the envelope signals of deep muscle activation are obtained through differential amplification and filtering. Simultaneously, multimodal hand dynamics data from professional muscle-relaxing technicians were collected during manipulation cycles to construct a reference sample library. Each data set in the reference sample library was associated with the professional technician's manipulation technique tag information and demonstration manipulation cycle number. The reference sample library preferably contains at least 30 manipulation cycles to cover the natural fluctuation range of professional technicians' techniques and obtain a stable statistical benchmark. Specifically, the start of a movement was determined when the triaxial acceleration of the hand exceeded the motion threshold and was accompanied by activation of the electromyographic envelope signal; the end of a movement was determined when the triaxial acceleration, angular velocity, pressure value, and electromyographic envelope signal all fell back to the resting threshold and remained there for more than the resting time window. The motion threshold was determined based on the multimodal noise characteristics under resting conditions. The statistical method is adaptively set, preferably the mean of resting acceleration plus three times the standard deviation, and verified during equipment calibration. The electromyographic activation state is determined by the proportion threshold of the envelope signal exceeding the baseline noise level, ensuring that the determination of the start of the movement is not affected by instantaneous noise. The resting time window is used to avoid misjudgment caused by small oscillations at the end of the movement, and is preferably set to 100 to 300 ms. Based on the unified system clock of the sampling device, a timestamp is added to each sampling point. When the start and end of the movement are determined, the corresponding timestamps are recorded to determine the start and end times of the manual cycle. After the end of the movement is determined, all multimodal hand dynamics data within the cycle are automatically extracted to ensure that the cycle boundary is unique and reproducible.

[0026] This implementation plan comprehensively and meticulously characterizes the acceleration changes, posture changes, pressure contact patterns, and synergistic activation of superficial and deep muscle groups of the operator and professional muscle-manipulating technician during the application of force by simultaneously collecting and fusing multimodal hand dynamics data within the manipulation cycle. A reference sample library is constructed using multi-cycle demonstration data from professional technicians, and combined with an automated motion start and end detection mechanism and a unified system clock recording, ensuring the uniqueness and reproducibility of manipulation cycle divisions while guaranteeing robust and reliable motion recognition, avoiding misjudgments caused by noise disturbances. Based on the statistical benchmark formed by professional manipulation cycles, a stable reference model is provided, laying an accurate and reliable data foundation for subsequent force chain analysis, force determination, and training feedback, thereby improving feasibility and the consistency of evaluation results.

[0027] Specifically, the data preprocessing process for multimodal hand dynamics data is as follows: multimodal time synchronization is performed based on the sampling timestamps of each channel, and the multimodal hand dynamics data are uniformly mapped to the same time axis using hardware clock alignment. All sensing channels are timed by the same main control chip, with a timestamp accuracy preferably not less than 1ms, to ensure accurate and unbiased timing correspondence between different modes. To avoid cross-modal phase misalignment caused by buffer delay, the system employs a circular buffer polling mechanism to correct the channel input order in real time. Low-pass filtering and drift compensation are applied to the three-axis acceleration and three-axis angular velocity of the hand. The low-pass filtering preferably uses a Butterworth filter with a cutoff frequency of 10 to 20Hz to suppress high-frequency noise. Drift compensation is achieved by automatically identifying and estimating the zero bias during the resting segment to ensure that the integration and differentiation processes are not affected by sensor drift. Spatial convolution smoothing is used to suppress isolated noise in the pressure values ​​of each sampling unit of the tactile array. Simultaneously, the contact center coordinates and contact area are calculated using a two-dimensional centroid algorithm based on the pressure values, and linear interpolation is performed on the contact center coordinates to obtain a continuous contact center trajectory. The spatial convolution kernel is preferably a 3×3 or 5×5 structure to enhance pressure consistency in local areas. The two-dimensional centroid algorithm calculates the contact center based on a weighted average of pressure pixels, avoiding offsets caused by local high-pressure noise. Linear interpolation is used to compensate for trajectory points when the sampling interval is large, ensuring the temporal continuity and smoothness of the trajectory. Effective electromyographic envelopes for superficial and deep muscle groups are obtained through bandpass filtering, rectification, and smoothing. Bandpass filtering preferably acquires effective electromyographic frequencies from 20 to 500 Hz to remove power frequency interference and motion artifacts. Full-wave rectification is used for rectification. Smoothing preferably uses a moving average window, such as 20 to 50 ms, to enhance envelope stability. Z-score normalization is performed on hand dynamics data to ensure comparability of data between different personnel and different equipment; a manipulation assessment database is established, using a time-series structured storage method, to write the original and pre-processed multimodal hand dynamics data of operators, reference sample library, corresponding continuous contact center trajectory and contact area into the manipulation assessment database, supporting index retrieval by manipulation cycle number, and binding a unique cycle ID and timestamp to each set of data when writing.

[0028] In this implementation scheme, multimodal time synchronization, denoising, and trajectory reconstruction enable stable fusion of heterogeneous data from IMU, pressure array, and electromyography (EMG) sensors on a unified time axis. Targeted filtering and compensation of acceleration, angular velocity, pressure, and EMG signals significantly improve the resolvability of dynamic thrust trajectory, contact center changes, and EMG activation patterns. Furthermore, contact center trajectory interpolation, two-dimensional centroid reconstruction, and standardization ensure that the input data for subsequent thrust direction extraction, chain force identification, and force assessment are unified, stable, and physiologically accurate, thus providing a reliable data foundation and reproducible processing flow for overall technique assessment.

[0029] Specifically, the thrust migration process is reconstructed based on preprocessed multimodal hand dynamics data. The target force direction is characterized using a principal direction extraction method. The stability of the thrust flow is determined by analyzing the consistency of velocity direction and the variation in jitter amplitude. The process involves: acquiring continuous contact center trajectories within the manipulation cycle; calculating the instantaneous velocity vector of the contact center using the finite difference method; and calculating the instantaneous acceleration amplitude. The contact center trajectory is continuously generated from the two-dimensional centroid coordinates of the pressure distribution combined with the tactile array, and the velocity is calculated point-by-point to ensure the temporal accuracy of the instantaneous velocity calculation. The instantaneous acceleration is obtained again from the instantaneous velocity sequence through differential calculation and synchronized with a unified time axis. The continuous contact center trajectories corresponding to the reference sample library are read, and the principal direction vector of the trajectory is extracted using principal component analysis. This principal direction vector is then normalized to obtain the target force direction unit vector. Principal component analysis uses the covariance matrix of the trajectory point cloud to obtain the maximum eigenvector, which can robustly represent the principal force direction of the professional hand manipulation technician. Vector normalization ensures that the direction representation is not affected by length. The median of instantaneous acceleration amplitude within the manipulation cycle in the reference sample library is selected as the contact center acceleration reference value to avoid interference from individual peak values ​​on the overall force application rhythm. For each sampling moment within the manipulation cycle, the cosine of the angle between the instantaneous velocity vector at the contact center and the unit vector of the target force direction is calculated. The cosine is then non-negated to obtain a directional consistency value, reflecting the degree to which the operator's thrust direction deviates from the professional force application direction. Non-negation involves truncating the negative cosine value to 0 to avoid evaluation ambiguity caused by reverse thrust. The instantaneous acceleration amplitude is divided by the contact center acceleration reference value, and the opposite number is used as the exponent for natural exponential calculation to obtain a jitter penalty value, which is used to quantify the degree of abnormal fluctuation in instantaneous acceleration. When the instantaneous acceleration is significantly higher than the reference value, the exponent term decreases rapidly, thereby enhancing the sensitivity to high-frequency jitter. The directional consistency value is multiplied by the jitter penalty value, and the product result is accumulated point by point and divided by the total number of sampling points to obtain the force vector flow continuity value. The average processing after point-by-point accumulation is used to obtain the global stability index of the entire technique cycle, which can effectively avoid the distortion of the overall evaluation due to local anomalies, and finally form a quantifiable feature value that integrates the thrust direction, jitter amplitude and force application rhythm.

[0030] The specific formula for the continuity value of the force vector flow is as follows: ; In the formula, It represents the force vector flow continuity value, which measures the degree to which thrust flows continuously in the correct direction; It represents the unit vector of the target force direction, indicating the direction in which the thrust should be output in the Qinggong Lijin technique, which is the ideal force direction; Represents the instantaneous velocity vector at the center of contact, and indicates the direction of the operator's actual thrust at the current moment; This represents the magnitude of a unit vector in the direction of the applied force to the target. This represents the magnitude of the instantaneous velocity vector at the contact center; It represents the instantaneous acceleration amplitude, indicating whether there is jitter or sudden surge in thrust at the current moment; It represents the reference value of the contact center acceleration, which is used as a benchmark for thrust stability.

[0031] In this implementation plan, through detailed analysis of the contact center trajectory, velocity vector, and acceleration changes, a unified quantitative expression of thrust direction, force application rhythm, and local jitter is achieved. Principal component analysis is used to extract the main force application direction of the professional technician, making the target direction statistically robust. The force vector flow continuity index, constructed by combining the direction consistency value and jitter penalty value, can simultaneously reflect whether the thrust path is continuous, whether the direction is stable, and whether there are abnormal acceleration fluctuations. The overall method can not only accurately characterize the subtle deviations in the operator's thrust transfer process, but also provide quantifiable and comparable process characteristics for force training, improving the objectivity and operability of the evaluation results.

[0032] Specifically, the process of generating force flow characteristic records is as follows: If the force vector flow continuity value is less than the smoothness threshold, it is determined that there is a smoothness anomaly in the current technique, and the abnormal period is marked; according to the directional consistency value at each sampling time within the technique cycle, the directional consistency value is scanned sequentially. When the directional consistency value is continuously lower than the directional offset threshold, segments in which the thrust direction deviates from the target force direction are identified. Segments in which the continuous deviation exceeds the allowable time threshold are marked as directional offset segments, and the corresponding time range is recorded; at the same time, the jitter intensity of each sampling point within the technique cycle is quantified using the jitter penalty value sequence, and the jitter penalty value is mapped to color intensity according to the sampling time sequence to generate a jitter heat distribution map arranged along the time axis; the force vector flow continuity value, abnormal period, and directional offset segment are all written into the technique evaluation database in the form of structured fields, including the timestamp range of the abnormal period, the index interval of the directional offset segment, the matrix value of the jitter heat distribution, and the corresponding force vector flow continuity value.

[0033] like Figure 3 The image shows a heat map of jitter intensity based on jitter penalty values. It visually demonstrates the stability of thrust output within a single manipulation cycle. The horizontal axis represents the normalized time of the manipulation cycle, while the vertical axis uses bolded bars to represent the heat map encoding of jitter intensity at continuous sampling points, with colors ranging from cold to hot corresponding to increasing jitter intensity. The overall manipulation cycle in the image has a light yellow background, indicating that the instantaneous acceleration amplitude at most sampling points is close to the reference value, and the overall thrust flow is relatively stable. However, bright red and dark red segments appear in the time intervals of 0.18 to 0.30, 0.55 to 0.67, and 0.78 to 0.88, indicating significantly higher instantaneous accelerations within these time segments and correspondingly lower jitter penalty values, representing unstable stages of thrust output. Unstable stages typically reflect deviations in the operator's control of thrust direction, joint chain force application, or local force release rhythm, and are important factors affecting the overall continuity of force. Visualized images enable operators to quickly pinpoint the specific time period of thrust fluctuations and quantify the spatial distribution intensity of the jitter, thus providing an intuitive basis for subsequent force correction training.

[0034] This implementation scheme, by uniformly processing the consistency of force vector direction, jitter amplitude, and temporal continuity, can accurately characterize thrust flow features within a single manipulation cycle, enabling precise identification and visualization of smoothness anomalies, directional offset segments, and local jitter peaks. By structurally recording abnormal time periods, offset intervals, and jitter intensity distributions into the database, the traceability of evaluation results and quantifiable support for subsequent training feedback are achieved, thereby significantly improving the precision and feasibility of thrust flow analysis and providing a stable data foundation for subsequent force transmission analysis and comprehensive evaluation.

[0035] Specifically, the process of identifying the sequential relationship of force application in the wrist, palm, and fingers using preprocessed multimodal hand dynamics data, and forming typical transmission rhythms by combining them with a reference sample library, to determine the temporal integrity of the current force chain, involves: performing amplitude threshold detection and rising edge search on the electromyographic envelope signals of superficial and deep muscle groups within the hand technique cycle to identify the rising time points of wrist, palm, and finger electromyographic signals. The rising edge search is preferably based on first-order difference and envelope signal... The composite criterion for triggering the signal exceeding the activation threshold is executed to avoid false triggering due to transient noise. The detection window for the rising edge is preferably set to 20 to 50 ms to ensure robust adaptability to different individual EMG activation characteristics. The wrist-to-palm force application time difference and the palm-to-finger force application time difference are calculated based on three time points. The wrist-to-palm force application time difference is obtained by calculating the difference between the palm EMG rising time point and the wrist EMG rising time point, and the palm-to-finger force application time difference is obtained by calculating the difference between the finger EMG rising time point and the palm EMG rising time point. The wrist-to-palm force application time difference and the palm-to-finger force application time difference are calculated for N manipulation cycles in the reference sample library, and the median is selected to obtain the wrist-to-palm transmission time reference value and the palm-to-finger transmission time reference value. The median as a reference value can reduce the interference of abnormal samples on the statistical benchmark and ensure that the natural fluctuations between different technicians will not cause deviation. N is preferably data from no less than 30 manipulation cycles to cover the natural manipulation fluctuation range of professional muscle-relaxing technicians and ensure the stability and robustness of the median reference value. The wrist-to-palm force application time difference in the current technique cycle is subtracted from the wrist-to-palm transmission time reference value to obtain the wrist-to-palm directional time deviation; the palm-to-finger force application time difference in the current technique cycle is subtracted from the palm-to-finger transmission time reference value to obtain the palm-to-finger directional time deviation. The wrist-to-palm and palm-to-finger directional time deviations are squared respectively, and the sum of the two squares is taken as the square root to obtain the composite time deviation. The squaring operation suppresses the cancellation effect of positive and negative offsets, ensuring that both advance and lag are quantified as effective errors. Taking the square root restores the dimension consistent with the actual time deviation scale, making the indicator easier to normalize later. The square root of the sum of the squares of the wrist-to-palm and palm-to-finger transmission time reference values ​​yields the reference time scale value, which is used to normalize different cycles or rhythms, ensuring that the chain force application sequence consistency value stably expresses the temporal continuity within the 0-1 range. The chain force application sequence consistency value is obtained by subtracting the ratio of the composite time deviation to the reference time scale value from a constant.

[0036] The specific formula for the consistency value of the chain-like force application sequence is as follows: ; In the formula, It represents the consistency value of the chain force application sequence, quantifying whether the three segments of force transmission from the wrist to the palm to the fingers are connected in the correct sequence, whether they are too early or too late, and whether there is a break in the force. This indicates the time point of rise in palmar electromyography, representing the moment when the palmar muscle group begins to exert force, and is used to calculate the transmission time difference from wrist to palm. This indicates the time point of rise in wrist electromyography, which represents the moment when the wrist muscles begin to exert force in the current manipulation cycle, and is used to determine the starting point of chain force exertion; It indicates the time point of rise of electromyography in the fingers, which is the moment when the finger muscles begin to exert force, and is used to calculate the time difference of transmission from the palm to the fingers; This indicates a reference value for wrist-to-palm transfer time; This represents the reference value for the hand-to-finger transfer time, serving as the benchmark for deviation calculation.

[0037] In this implementation scheme, a quantifiable expression of the force chain temporal structure is constructed by accurately extracting the electromyographic activation sequence of the wrist, palm, and fingers. By utilizing rising edge search, difference calculation, and the median benchmark of the reference sample library, the typical transmission rhythm of professional technicians can be effectively characterized, and the influence of individual differences and abnormal samples on the statistical results can be suppressed. Through the normalization processing of the time deviation synthesis amount and the reference time scale value, the rhythm differences of different cycles, different speeds, and different operators can be mapped to a unified interval, so that the consistency value of the chain force sequence can stably and intuitively reflect the degree of advance, lag, or disorder of the force chain. Overall, the objective quantification of the force transmission rhythm is achieved, which helps to reliably identify force connection abnormalities and provide interpretable and practical dynamic basis for subsequent evaluation and training.

[0038] Specifically, the process for identifying the abnormal chain segment is as follows: If the consistency value of the chain force application sequence within the current technique cycle is lower than the continuity threshold, an abnormal force transmission is determined. Threshold comparison is automatically performed after each technique cycle, and the force chain segment where the abnormal link is located is identified: The absolute value of the time deviation in the wrist-to-palm direction is compared with the absolute value of the time deviation in the palm-to-finger direction. If the absolute value of the time deviation in the wrist-to-palm direction is larger, it is marked as a wrist-to-palm transmission abnormality; if the absolute value of the time deviation in the palm-to-finger direction is larger, it is marked as a palm-to-finger transmission abnormality; if the absolute values ​​of both time deviations are greater than the corresponding deviation thresholds, it is marked as a multi-segment chain transmission abnormality. When determining the abnormal chain segment, the two deviations are independently compared with the corresponding deviation thresholds. The deviation threshold is preferably set as the mean of the corresponding time deviation distribution in the reference sample library plus twice the standard deviation to ensure that the abnormal marking is statistically significant. The consistency value of the chain-like force application sequence, the time deviation from wrist to palm, the time deviation from palm to finger, and the corresponding abnormal markers are written into the technique evaluation database. If a continuous technique cycle is judged to have an abnormal force transmission exceeding the cycle tolerance threshold, an unstable chain transmission prompt is generated, and information on the abnormal force application chain segment is pushed along with targeted training reminders. The cycle tolerance threshold is preferably set to 2 to 3 consecutive cycles to avoid false prompts due to occasional noise or individual abnormalities. Targeted training reminders automatically match training templates based on the abnormal chain segment category. For example, an abnormal wrist-to-palm transmission will trigger wrist force application connection training suggestions, an abnormal palm-to-finger transmission will trigger fingertip force application delay coordination training suggestions, and multiple abnormal chain transmissions will trigger a full-chain force application rhythm training program. All prompts are presented as text prompts on the feedback interface.

[0039] This implementation plan, based on the quantification of time deviation, can further distinguish specific abnormal links in the force transmission process, achieving precise positioning of the wrist-to-palm, palm-to-finger, and multi-segment chains. This avoids the technical shortcomings of traditional methods that can only provide an overall abnormality without pinpointing the source of the problem. By comparing the absolute value of the deviation and determining a threshold, this method makes anomaly identification quantifiable, repeatable, and interpretable. Combined with continuous cycle tolerance judgments, it effectively filters out sporadic noise, ensuring stable and reliable prompts. Simultaneously, anomaly information is written into a structured database and linked to targeted training reminders, enabling not only accurate detection of force transmission problems but also closed-loop guidance and continuous optimization, significantly improving the scientific rigor and operability of the Qinggong Lijin manipulation technique training.

[0040] Specifically, the process of comprehensively analyzing the preprocessed multimodal hand dynamics data, thrust flow stability, and the temporal integrity of the force chain is as follows: Electromyographic envelope signals of superficial and deep muscle groups are acquired within the manipulation cycle, and time integration is performed to obtain superficial and deep muscle energy values. The ratio of superficial to deep muscle participation energy is obtained by dividing the deep muscle energy value by the superficial muscle energy value. The time integration preferably employs a trapezoidal method-based numerical integration approach, using the sampling timestamp as the integration step size to accumulate the continuous electromyographic envelope curve point by point, thereby obtaining the energy value representing the overall muscle participation in the cycle. This ensures that the integration process has consistent physical meaning at different sampling frequencies. Simultaneously, the reference sample library is read, and the energy ratio of deep and superficial muscle involvement in N manipulation cycles is calculated. The median is selected as the standard benchmark ratio. Using the median can reduce the impact of occasional abnormalities on the assessment benchmark, making the standard benchmark ratio more robust. The operator's energy ratio of deep and superficial muscle involvement in the manipulation cycle is divided by the standard benchmark ratio, and the square root is taken to obtain the normalized ratio of deep and superficial muscle energy. The normalized ratio of deep and superficial muscle energy is used to reflect the operator's degree of coordination of deep and superficial muscle involvement relative to the professional benchmark. The square root operation can weaken the amplification effect of extreme samples on subsequent comprehensive indicators, so that the indicators maintain a stable dynamic range. The normalized ratio of deep and superficial muscle energy is multiplied by the consistency value of chain force application sequence and the continuity value of force vector flow, and the cube root is taken to obtain the comprehensive strength assessment value. The cube root is used to balance abnormal items when integrating multiple sub-indicators, so that the contribution of the three sub-indicators is more balanced and the overall assessment result is not distorted due to the low value of one indicator.

[0041] The specific formula for the comprehensive strength assessment value is as follows: ; In the formula, This represents a comprehensive strength assessment value, which integrates the three dimensions of strength in a structured and quantitative manner to obtain an objective indicator that can be used for level determination and training feedback. It represents the consistency value of the chain-like force generation sequence, reflecting whether the power transmission of the wrist, palm, and fingers is connected in a professional sequence, and is the temporal basis for the formation of power. It represents the continuity value of force vector flow, and measures whether the thrust continues to advance along the target direction during the technique cycle, and whether there is directional deviation or jitter. It is a key indicator of force stability. Indicates the length of the technique cycle; It represents the electromyographic envelope signal of deep muscle groups, which is used to calculate the energy involved by deep muscles throughout the entire manipulation cycle and is a quantitative reflection of the main source of force. It represents the electromyographic envelope signal of superficial muscle groups, used to calculate the energy involved in superficial muscles, and is a source of auxiliary force generation; This represents the standard baseline ratio, used for normalization to ensure comparability of the deep and superficial muscle involvement ratios between different operators.

[0042] In this embodiment, Table 1 is a comprehensive power assessment value data table. The standard baseline ratio is set to 1.10. The table details the consistency value of the chain-like force application sequence, the continuity value of the force vector flow, the deep muscle energy value, the superficial muscle energy value, and the comprehensive power assessment value for five technique cycles. Specifically, cycle 1 has a chain-like force application sequence consistency value of 0.55, a force vector flow continuity value of 0.50, a deep muscle energy value of 60, a superficial muscle energy value of 70, and a comprehensive power assessment value of 0.6238; cycle 2 has a chain-like force application sequence consistency value of 0.78, a force vector flow continuity value of 0.72, a deep muscle energy value of 95, a superficial muscle energy value of 80, and a comprehensive power assessment value of 0.835. 6; The chain-like force generation sequence consistency value for cycle 3 is 0.86, the force vector flow continuity value is 0.81, the deep muscle energy value is 110, the superficial muscle energy value is 75, and the comprehensive power assessment value is 0.9300; The chain-like force generation sequence consistency value for cycle 4 is 0.90, the force vector flow continuity value is 0.85, the deep muscle energy value is 130, the superficial muscle energy value is 70, and the comprehensive power assessment value is 0.9980; The chain-like force generation sequence consistency value for cycle 5 is 0.85, the force vector flow continuity value is 0.79, the deep muscle energy value is 85, the superficial muscle energy value is 90, and the comprehensive power assessment value is 0.8537.

[0043] Table 1. Comprehensive Strength Assessment Value Data Table

[0044] like Figure 4 As shown, this is a trend comparison chart of the comprehensive strength assessment value. It displays the changing trend of the comprehensive strength assessment value over time for five technique cycles. Two strength threshold lines, T1 and T2, are also shown in the chart. The line graph reflects the overall trajectory of the operator's strength performance in different technique cycles, visually presenting the evolution of strength from insufficient to excellent. Combined with Table 1 and... Figure 4It can be seen that the operator's force performance presents a continuous improvement process from low to high: the evaluation value of cycle 1 is significantly lower than T1, which is in the insufficient force interval, manifested as unstable chain connection and unsmooth thrust flow; after entering cycle 2, the evaluation value rises into the good force interval between T1 and T2, indicating that the chain transmission rhythm and thrust direction control are significantly improved; cycle 3 further crosses to the excellent force interval higher than T2, showing relatively professional mastery of force rhythm and thrust path; cycle 4 reaches the highest value in the whole process, with the most superior force performance, and the energy matching of deep and superficial muscle groups, the timing of the force generation chain and the thrust flow all reach an ideal state; although the evaluation value of cycle 5 drops slightly compared with cycle 4, it is still in the good interval above T1, indicating that the force performance remains stable with only normal fluctuations. On the whole, the five cycles present a force evolution law of continuous progress, prominent peak and subsequent stable maintenance.

[0045] In this embodiment, multi-dimensional objective quantification of force performance is realized through integral calculation of superficial and deep electromyographic energy, ratio normalization and joint fusion with thrust flow and chain transmission indicators; a standard reference ratio is constructed by using a reference sample library to ensure the comparability and robustness of evaluation among different operators; through the cube root integration method of multiple indicators, the evaluation result remains stable even under abnormal fluctuations; the overall method can uniformly characterize force coordination, force layer participation and force generation rhythm completeness, providing a reliable quantitative basis for subsequent force grading, training guidance and individualized advancement.

[0046] Specifically, the operator's force performance is graded, and targeted training prompts are generated; meanwhile, the threshold parameters are updated based on the distribution of historical excellent samples to form a closed-loop optimization feedback, and the specific process is: the comprehensive force evaluation value is compared with the multi-level force thresholds T1 and T2: when <T1, it is determined as insufficient force, and the corresponding abnormal force generation chain segment, direction deviation segment and jitter heat peak area are displayed on the feedback interface; a training scheme focusing on basic force coherence is recommended, prompting to increase exercises targeting hand force generation connection, thrust direction correction and deep-superficial force generation coordination; when T1≤ <T2, it is determined as good force, a prompt indicating that the overall force is good is given on the feedback interface, and advanced training is recommended; when When the value is ≥T2, it is judged as having excellent power. A prompt indicating excellent power is given in the feedback interface, and the current technique cycle is marked as an excellent sample. The corresponding multimodal hand dynamics data, chain force sequence consistency value, force vector flow continuity value, and deep and superficial muscle participation energy ratio are written into the individual excellent sample database, and complex technique combination training is recommended. Among them, the training prompts can automatically match the corresponding practice templates based on the grading results when they are generated, such as rhythm connection training, force direction control training, and muscle group coordination training, to ensure the relevance of the training program. The comprehensive strength assessment value, strength grading results, corresponding prompts, and an individual excellent sample library are written into the technique assessment database, and a strength training feedback report for the operator is generated. Simultaneously, based on the sample distribution recorded in the individual excellent sample library, an adaptive update algorithm based on quantile statistics is used periodically to adjust the standard baseline ratio and consistency threshold: the latest deep and superficial muscle participation energy ratio sequence and chain force application sequence consistency value sequence are extracted from the excellent sample library and accumulated in chronological order to form an update window; quantile statistics are calculated for each sequence, preferably using the 50th percentile as the new standard baseline ratio and the 30th percentile as the new consistency threshold, using quantile characteristics to ensure that the update results are not affected by individual extreme samples; when the number of newly added excellent samples reaches the quantity threshold, the threshold parameters are automatically updated, the updated standard baseline ratio and consistency threshold are written into the database, and applied to the strength assessment of subsequent technique cycles, achieving adaptive adjustment of the thresholds that dynamically stabilizes with individual performance.

[0047] In this implementation plan, by constructing a power grading and training feedback mechanism and combining it with an individual high-performing sample library to achieve adaptive updates of threshold parameters, this method can not only automatically generate targeted training prompts that match the operator's level based on the comprehensive power assessment value, but also dynamically adjust the standard benchmark ratio and consistency threshold based on the real performance of high-performing samples during continuous training. This allows the assessment to gradually stabilize as individual abilities change, avoiding the risk of misjudgment or lag, thereby achieving personalized, self-evolving, and closed-loop optimization of power assessment standards, and significantly improving the accuracy and long-term effectiveness of training feedback.

[0048] Reference Figure 2As shown, the second aspect of this invention provides an evaluation system for Qinggong muscle manipulation techniques based on multimodal dynamics fusion, applied to the aforementioned evaluation method for Qinggong muscle manipulation techniques based on multimodal dynamics fusion. The system includes: a multimodal data acquisition and preprocessing module for acquiring multimodal hand dynamics data from operators and professional muscle manipulation technicians, constructing a reference sample library using the multimodal hand dynamics data from professional technicians, and preprocessing the multimodal hand dynamics data; and a three-dimensional motion and force vector fusion reconstruction module for reconstructing the thrust migration process based on the preprocessed multimodal hand dynamics data, characterizing the target force direction using a principal direction extraction method, and analyzing the consistency of velocity direction. The system uses vibration amplitude changes to determine thrust flow stability and generate force flow characteristic records. The force chain dynamics analysis module uses pre-processed multimodal hand dynamics data to identify the sequential relationship of force application from the wrist, palm, and fingers, and combines this with a reference sample library to form typical transmission rhythms, determine the temporal integrity of the current force chain, and identify the chain segment to which anomalies belong. The technique comparison and force evaluation module comprehensively analyzes pre-processed multimodal hand dynamics data, thrust flow stability, and the temporal integrity of the force chain, classifies the operator's force performance, and generates targeted training prompts. Simultaneously, it updates threshold parameters based on the distribution of historical excellent samples, forming a closed-loop optimization feedback.

[0049] This implementation plan achieves closed-loop optimization of the entire process from data collection and analysis to feedback by constructing a multi-level, multi-modal quantitative evaluation chain for the Qinggong Lijin manipulation technique. On the one hand, by introducing a reference sample library of professional technicians, the direction of the push, the amplitude of the shaking, and the timing rhythm all have objective benchmarks, thus effectively avoiding the subjectivity of traditional experience-based judgments. On the other hand, a synergistic evaluation is formed in three key dimensions: push flow, chain-like force sequence, and involvement of deep and superficial muscle groups, so that the core elements of the force performance can be comprehensively and granularly quantitatively expressed. Finally, combined with an adaptive threshold update strategy based on historical excellent samples, the evaluation criteria can be dynamically adjusted as the operator's skills improve, ensuring the continuous effectiveness and individualized adaptability of training feedback, thereby significantly improving the learning efficiency and movement quality control ability of massage force.

[0050] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0051] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. As those skilled in the art will understand, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An evaluation method for Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion, characterized in that, Includes the following steps: S1. Collect multimodal hand dynamics data from operators and professional muscle relaxers, construct a reference sample library using the multimodal hand dynamics data from professional muscle relaxers, and perform data preprocessing on the multimodal hand dynamics data; S2, based on the preprocessed multimodal hand dynamics data, reconstruct the thrust migration process, use the main direction extraction method to characterize the target force direction, analyze the consistency of velocity direction and the change of jitter amplitude to determine the thrust flow stability, and generate force flow characteristic records; S3 uses preprocessed multimodal hand dynamics data to identify the sequential relationship of force application in the wrist, palm, and fingers, and combines it with a reference sample library to form a typical transmission rhythm, determine the temporal integrity of the current force chain, and identify the chain segment to which the abnormality belongs. S4 comprehensively analyzes the preprocessed multimodal hand dynamics data, thrust flow stability, and force chain timing integrity to classify the operator's strength performance and generate targeted training prompts; at the same time, it updates the threshold parameters based on the distribution of historical excellent samples to form a closed-loop optimization feedback.

2. The method for evaluating Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process of collecting multimodal hand dynamics data from both the operator and the professional muscle relaxant, and constructing a reference sample library using the multimodal hand dynamics data from the professional muscle relaxant is as follows: The process from the start to the end of a movement is defined as a manipulation cycle. Within each manipulation cycle, multimodal hand dynamics data of the operator is collected in real time using a smart sensing glove. The multimodal hand dynamics data includes: triaxial acceleration of the hand, triaxial angular velocity of the hand, pressure values ​​of each sampling unit of the tactile array, electromyographic envelope signals of superficial muscle groups, and electromyographic envelope signals of deep muscle groups. At the same time, multimodal hand dynamics data of professional muscle manipulation technicians are collected within the manipulation cycle to construct a reference sample library. Specifically, the start of a movement is determined when the triaxial acceleration of the hand exceeds the motion threshold and is accompanied by activation of the electromyographic envelope signal; the end of a movement is determined when the triaxial acceleration, angular velocity, pressure value, and electromyographic envelope signal of the hand all fall back to the resting threshold and continue to exceed the resting time window.

3. The method for evaluating Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process for preprocessing the multimodal hand dynamics data is as follows: Multimodal time synchronization is performed based on the sampling timestamps of each channel, and hardware clock alignment is used to uniformly map multimodal hand dynamics data to the same time axis. Low-pass filtering and drift compensation are applied to the three-axis acceleration and three-axis angular velocity of the hand. Spatial convolution smoothing is used to suppress isolated noise for the pressure values ​​of each sampling unit of the tactile array. At the same time, the contact center coordinates and contact area are calculated using a two-dimensional centroid algorithm based on the pressure values, and the contact center coordinates are linearly interpolated to obtain the continuous contact center trajectory. The electromyographic envelope signals of superficial muscle groups and deep muscle groups are obtained by bandpass filtering, rectification and smoothing to obtain effective electromyographic envelopes. Z-score normalization is performed on the hand dynamics data. A manipulation assessment database is established, in which the original and preprocessed multimodal hand dynamics data of the operator, the reference sample library, the corresponding continuous contact center trajectory and contact area are written into the manipulation assessment database.

4. The method for evaluating the Qing Dynasty uterine muscle manipulation technique based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process of reconstructing the thrust migration process based on preprocessed multimodal hand dynamics data, using the principal direction extraction method to characterize the target force direction, and analyzing the consistency of velocity direction and the change in jitter amplitude to determine the thrust flow stability is as follows: Obtain the continuous contact center trajectory within the technique cycle, calculate the instantaneous velocity vector of the contact center using the finite difference method, and calculate the instantaneous acceleration amplitude; Read the continuous contact center trajectory corresponding to the reference sample library, extract the principal direction vector of the trajectory using principal component analysis, and normalize the principal direction vector to obtain the unit vector of the target force direction; select the median of the instantaneous acceleration amplitude within the operation cycle in the reference sample library as the contact center acceleration reference value; For each sampling moment during the operator's technique cycle, the cosine of the angle between the instantaneous velocity vector at the contact center and the unit vector of the target force direction is calculated, and the cosine is non-negative to obtain the directional consistency value. The instantaneous acceleration amplitude is divided by the reference value of the acceleration at the contact center, and the opposite number is used as the exponent for natural exponential operation to obtain the jitter penalty value. The directional consistency value and the jitter penalty value are multiplied together, and the product of all sampling moments is accumulated point by point and divided by the total number of sampling points to obtain the force vector flow continuity value.

5. The method for evaluating the Qing Dynasty uterine muscle manipulation technique based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process for generating the force flow characteristic record is as follows: If the force vector flow continuity value is less than the smoothness threshold, the current technique is judged to have a smoothness anomaly, and the abnormal period is marked. Based on the directional consistency value of each sampling time within the technique cycle, segments where the thrust direction deviates from the target force direction are identified, and segments that continuously deviate for more than the allowable time threshold are marked as directional offset segments. At the same time, the jitter intensity of each sampling point within the technique cycle is quantified using the jitter penalty value sequence to generate a jitter heat map. The force vector flow continuity value, abnormal period, and directional offset segment are written into the technique evaluation database.

6. The method for evaluating Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process of identifying the sequential relationship of force application in the wrist, palm, and fingers using preprocessed multimodal hand dynamics data, and combining this with a reference sample library to form a typical transmission rhythm, and determining the temporal integrity of the current force chain, is as follows: Amplitude threshold detection and rising edge search are performed on the electromyographic envelope signals of superficial and deep muscle groups within the hand technique cycle to identify the rising time points of wrist, palm, and fingers. The wrist-to-palm force application time difference and palm-to-finger force application time difference are calculated based on the three time points. Calculate the wrist-to-palm force application time difference and palm-to-finger force application time difference for N manual cycle in the reference sample library, and select the median to obtain the wrist-to-palm transmission time reference value and palm-to-finger transmission time reference value respectively. The wrist-to-palm force application time difference of the operator in the current technique cycle is subtracted from the wrist-to-palm transmission time reference value to obtain the wrist-to-palm directional time deviation; the palm-to-finger force application time difference of the operator in the current technique cycle is subtracted from the palm-to-finger transmission time reference value to obtain the palm-to-finger directional time deviation; the wrist-to-palm directional time deviation and the palm-to-finger directional time deviation are squared respectively, the two squared results are added together and the square root is taken to obtain the composite time deviation; The reference time scale value is obtained by adding the square of the wrist-to-palm transmission time reference value to the square of the palm-to-finger transmission time reference value and taking the square root. The consistency value of the chain-type force generation sequence is obtained by subtracting the ratio of the time deviation composite amount to the reference time scale value from the constant.

7. The method for evaluating Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process for identifying the chain segment to which the anomaly belongs is as follows: If the consistency value of the chain of force generation within the current technique cycle is lower than the continuity threshold, it is determined that there is an abnormality in the force transmission, and the force chain segment where the abnormal link is located is identified: compare the absolute value of the time deviation in the wrist to palm direction with the absolute value of the time deviation in the palm to finger direction. If the absolute value of the time deviation in the wrist to palm direction is larger, it is marked as an abnormality in wrist to palm transmission; if the absolute value of the time deviation in the palm to finger direction is larger, it is marked as an abnormality in palm to finger transmission. If the absolute values ​​of both time deviations are greater than the corresponding deviation thresholds, then it is marked as a multi-segment chain transmission anomaly. The consistency value of the chain force application sequence, the time deviation from wrist to palm, the time deviation from palm to finger, and the corresponding abnormal markers are written into the technique evaluation database. If a continuous technique cycle is judged to have abnormal force transmission exceeding the cycle allowable threshold, an unstable chain transmission prompt is generated, and information on the abnormal force application chain segment is pushed and targeted training reminders are executed.

8. The method for evaluating Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process of comprehensively analyzing and preprocessing the multimodal hand dynamics data, thrust flow stability, and power chain timing integrity is as follows: The electromyographic envelope signals of superficial and deep muscle groups were acquired during the manipulation cycle, and the superficial and deep muscle energy values ​​were obtained by time integration. The energy ratio of superficial and deep muscles was obtained by dividing the deep muscle energy value by the superficial muscle energy value. Simultaneously, the reference sample library is read, the energy ratio of deep and superficial muscles involved in N manual cycles in the reference sample library is calculated, and the median is selected as the standard benchmark ratio. Divide the ratio of deep to superficial muscle energy participation by the operator during the technique cycle by the standard baseline ratio, and take the square root to obtain the normalized ratio of deep to superficial muscle energy. Multiply the normalized ratio of deep to superficial muscle energy by the consistency value of the chain force generation sequence and the continuity value of the force vector flow, and take the cube root to obtain the comprehensive strength assessment value.

9. The method for evaluating Qing Dynasty uterine muscle manipulation based on multimodal dynamics fusion according to claim 1, characterized in that, The specific process of classifying the operator's strength performance and generating targeted training prompts, while updating the threshold parameters based on the distribution of historical excellent samples to form a closed-loop optimization feedback is as follows: Comprehensive strength assessment value Compare with the multi-level force thresholds T1 and T2: When <T1, it is determined that the force exertion is insufficient, and the corresponding abnormal force exertion chain segment, direction deviation segment and jitter heat peak area are displayed on the feedback interface; a training scheme focusing on basic force exertion consistency is recommended, and a prompt is given to increase the training targeting hand force exertion connection, thrust direction correction, and deep-shallow force exertion coordination; When T1 ≤ < T2, it is determined that the muscle strength and texture are good, a prompt indicating that the overall muscle strength and texture are good is provided on the feedback interface, and advanced training is recommended; when When the value is ≥T2, it is judged as having excellent power. A prompt indicating excellent power is given in the feedback interface, and the current technique cycle is marked as an excellent sample. The corresponding multimodal hand dynamics data, chain force sequence consistency value, force vector flow continuity value, and deep and superficial muscle energy ratio are written into the individual excellent sample library, and complex technique combination training is recommended. The comprehensive strength assessment value, strength grading results, corresponding prompts, and individual excellent sample library are written into the technique assessment database, and a strength training feedback report for the operator is generated. At the same time, based on the sample distribution recorded in the individual excellent sample library, the standard benchmark ratio and consistency threshold are periodically adjusted using an adaptive update algorithm based on quantile statistics.

10. An evaluation system for Qing Dynasty uterine muscle manipulation techniques based on multimodal dynamics fusion, characterized in that, include: The multimodal data acquisition and preprocessing module is used to collect multimodal hand dynamics data from operators and professional muscle relaxants, construct a reference sample library using the multimodal hand dynamics data from professional muscle relaxants, and perform data preprocessing on the multimodal hand dynamics data. The 3D motion and force vector fusion reconstruction module is used to reconstruct the thrust migration process based on preprocessed multimodal hand dynamics data. It uses the main direction extraction method to characterize the target force direction, analyzes the consistency of velocity direction and the change of jitter amplitude to determine the thrust flow stability, and generates force flow feature records. The dynamic analysis module of the power chain is used to identify the sequential relationship of force application of the wrist, palm and fingers using preprocessed multimodal hand dynamics data, and to form a typical transmission rhythm by combining it with the reference sample library, to determine the temporal integrity of the current force chain, and to identify the chain segment to which the abnormality belongs. The technique comparison and force assessment module is used to comprehensively analyze the preprocessed multimodal hand dynamics data, thrust flow stability and force chain timing integrity, classify the operator's force performance, and generate targeted training prompts; at the same time, it updates the threshold parameters based on the distribution of historical excellent samples to form a closed-loop optimization feedback.

Images