Heart rehabilitation action precision evaluation method and system of wearable sensor equipment

Abstract

The invention discloses a heart rehabilitation action precision assessment method and system of wearable sensor equipment, and relates to the technical field of heart rehabilitation assessment, and the method comprises the following steps: utilizing a multi-mode sensor array module integrated with an electrocardio sensor, an inertial measurement unit IMU, a myoelectricity sensor and a pressure sensor; synchronously collecting multi-dimensional data of heart activity, motion posture, muscle electric signals and limb pressure; according to the method, multiple sensors are integrated, multi-dimensional data such as heart activity, movement postures, muscle electric signals and limb pressure are comprehensively collected, accurate acquisition of related information of rehabilitation actions is achieved, heterogeneous data are deeply fused through an improved self-adaptive weighted fusion algorithm, a more comprehensive and accurate rehabilitation action characteristic spectrum is generated, and the rehabilitation action characteristic spectrum is more accurate. And the transfer learning and meta-learning mechanism is introduced to construct a personalized evaluation model, individual differences of patients are fully considered, and an accurate evaluation standard is customized for each patient.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of cardiac rehabilitation evaluation, in particular to a cardiac rehabilitation action precision evaluation method and system of a wearable sensor device. BACKGROUND

[0002] Cardiac rehabilitation is an important part of comprehensive management of cardiovascular diseases. It is through comprehensive and individualized intervention measures, including exercise training, medical evaluation, health education, psychological and behavioral intervention, etc., aiming to improve the physical function, psychological state and quality of life of patients with cardiovascular diseases, prevent the recurrence of cardiovascular diseases, and many patients have impaired cardiac function after the onset, and the self-care ability and exercise capacity decrease. Cardiac rehabilitation can help patients enhance cardiac function, improve exercise tolerance, improve vascular endothelial function, and reduce the risk of recurrence of cardiovascular diseases. At the same time, it can also alleviate the anxiety, depression and other negative emotions of patients caused by the disease, improve the psychological health level, and promote the patients to return to normal life and work. In the process of cardiac rehabilitation, the precision evaluation of rehabilitation action plays a key role in the rehabilitation effect. Precise rehabilitation action can more effectively stimulate the recovery of cardiac function and avoid excessive burden on the heart or poor rehabilitation effect caused by incorrect action.

[0003] However, the existing technology still has certain defects in the precision evaluation of cardiac rehabilitation action. The existing evaluation methods mostly rely on the subjective observation and experience judgment of medical staff, lack objective and accurate quantitative indicators, and the evaluation standards of different medical staff are quite different. Moreover, the traditional evaluation methods are difficult to comprehensively obtain the information of the patient's cardiac activity, exercise posture, muscle state and other aspects, and cannot comprehensively analyze the influence of rehabilitation action on cardiac function. In addition, since the individual differences of patients are not fully considered, the general evaluation model cannot provide precise rehabilitation guidance for different patients. Therefore, it is of great significance to develop a cardiac rehabilitation action precision evaluation method and system of a wearable sensor device. SUMMARY

[0004] The purpose of the present application is to make up for the shortcomings of the prior art, and to provide a cardiac rehabilitation action precision evaluation method and system of a wearable sensor device, which can realize accurate and comprehensive evaluation of the precision of cardiac rehabilitation action by integrating multi-modal data acquisition, deep data fusion, personalized model construction and intelligent rehabilitation scheme adjustment, etc. It provides a scientific and reliable basis for the dynamic optimization of clinical rehabilitation scheme, and improves the overall effect of cardiac rehabilitation treatment.

[0005] To solve the above technical problems, the present application provides the following technical scheme: a cardiac rehabilitation action precision evaluation method of a wearable sensor device, which comprises the following steps:

[0006] The multi-modal sensor array module integrated with an electrocardio sensor, an inertial measurement unit (IMU), an electromyography sensor and a pressure sensor is used to synchronously collect multi-dimensional data of cardiac activity, motion posture, muscle electrical signal and limb pressure.

[0007] The collected multi-dimensional data is transmitted to a data processing module, an improved adaptive weighted fusion algorithm is used to dynamically allocate weights according to different data type characteristics, to fuse heterogeneous data, and generate a rehabilitation action feature map, and the calculation formula of the fused data F is: Wherein n is the number of data types, w i is the dynamic weight of the i-th data type, D i is the preprocessed data of the i-th data type, and The dynamic weight w i is calculated by the formula S i represents the stability index of the i-th data type, and C i represents the correlation index of the i-th data type and the rehabilitation action.

[0008] The transfer learning and meta-learning mechanism is introduced, and a personalized cardiac rehabilitation action precision evaluation model is constructed based on the patient's historical rehabilitation data, physiological indicators and disease characteristics.

[0009] The personalized evaluation model is used to evaluate the precision of the cardiac rehabilitation action according to the fused rehabilitation action feature map.

[0010] Further, when the multi-modal sensor array module is used to collect data, the electrocardio sensor is attached to the chest position to collect cardiac electrical signals, the IMU is installed at the preset positions of the wrist, ankle and torso to measure the motion posture, the electromyography sensor is attached to the main muscle parts involved in the rehabilitation action to collect muscle electrical signals, and the pressure sensor is installed at the shoe sole and the ground contact area and the rehabilitation equipment handle holding place to collect limb pressure data.

[0011] Further, the improved adaptive weighted fusion algorithm is used to preprocess the multi-dimensional data before data fusion, remove noise by using a digital filtering algorithm, and identify and repair abnormal data by using an abnormal value detection method based on statistical analysis.

[0012] Further, the specific steps of constructing the personalized cardiac rehabilitation action precision evaluation model are as follows:

[0013] The transfer learning mechanism extracts key features K from the existing cardiac rehabilitation data and general models according to the preset feature extraction rules, and the calculation formula is Wherein M is the number of general models, β m is the feature contribution coefficient of the m-th general model, and fm is a feature extraction function for the mth general model, G m is the mth general model, and an initial model framework is constructed based on the extracted key features K;

[0014] The meta-learning mechanism calculates a comprehensive feature vector V for the patient's historical rehabilitation data H, physiological indicators P, and disease characteristics D according to the formula V = γ1 × H + γ2 × P + γ3 × D, where γ1, γ2, and γ3 are the weight coefficients of the historical rehabilitation data, physiological indicators, and disease characteristics, respectively, and γ1 + γ2 + γ3 = 1. The weight coefficients γ1, γ2, and γ3 of the historical rehabilitation data, physiological indicators, and disease characteristics are first determined by experts according to the analytic hierarchy process to obtain initial values, then optimized using a random forest algorithm to analyze a large number of patient sample data, and finally calibrated through cross-validation to determine the final values. Based on the comprehensive feature vector V, the gradient descent method is used to adjust the model parameters multiple times to complete the construction of the personalized model.

[0015] Further, after evaluating the accuracy of the heart rehabilitation action, an action deviation vector is calculated based on the evaluation results. The action deviation vector is obtained by comparing the differences between the actual action feature map and the corresponding feature points in the standard rehabilitation action feature map. The calculation formula is: wherein is the action deviation vector, s is the number of feature points, is the value of the kth feature point in the actual action feature map, is the value of the kth feature point in the standard rehabilitation action feature map, is the unit vector of the corresponding feature point, which is a standardized directional vector determined based on the coordinate system of the standard rehabilitation action feature map.

[0016] Further, after evaluating the accuracy of the heart rehabilitation action, time series analysis is performed on the evaluation results of different stages. By calculating the change rate of the action accuracy data at adjacent evaluation time points and the correlation degree of the accuracy between different rehabilitation actions, the action change trend and potential rules in the patient's rehabilitation process are mined, providing a basis for subsequent rehabilitation plan adjustment.

[0017] Further, the results of evaluating the accuracy of the heart rehabilitation action are presented in the form of a visual chart. The chart includes a line chart and a column chart. The line chart shows the change of action accuracy over time, and the column chart compares the accuracy of different rehabilitation actions.

[0018] The heart rehabilitation action accuracy evaluation system of the wearable sensor device is suitable for the heart rehabilitation action accuracy evaluation method of the wearable sensor device described above. The system includes the following components:

[0019] A multi-modal sensor array module is used to integrate multi-dimensional data of cardiac activity, motion posture, muscle electrical signal and limb pressure;

[0020] A data transmission module is used to transmit the data collected by the multi-modal sensor array module to the data processing module;

[0021] The data processing module runs an improved adaptive weighted fusion algorithm to fuse the received data and generate a rehabilitation action feature map;

[0022] A personalized evaluation model library stores personalized cardiac rehabilitation action precision evaluation models constructed based on transfer learning and meta-learning mechanisms;

[0023] An evaluation module uses the personalized evaluation model to evaluate the precision of the cardiac rehabilitation action according to the rehabilitation action feature map.

[0024] Further, the multi-modal sensor array module adopts a modular design, each sensor module has high integration and low power consumption, and each module interacts and cooperates with each other through a standardized interface, and the sensor modules can be combined and replaced according to the rehabilitation needs of the patient.

[0025] Compared with the prior art, the wearable sensor device cardiac rehabilitation action precision evaluation method and system have the following beneficial effects:

[0026] The present application comprehensively collects multi-dimensional data such as cardiac activity, motion posture, muscle electrical signal and limb pressure by integrating multiple sensors, realizes accurate acquisition of rehabilitation action related information, deeply fuses heterogeneous data by using an improved adaptive weighted fusion algorithm, generates a more comprehensive and accurate rehabilitation action feature map, improves the accuracy and objectivity of evaluation, introduces a transfer learning and meta-learning mechanism to construct a personalized evaluation model, fully considers individual differences of patients, realizes customization of accurate evaluation standards for each patient, provides a scientific basis for dynamic adjustment of clinical rehabilitation programs, and achieves the effects of improving cardiac rehabilitation treatment effect, promoting recovery of cardiac function of patients and improving life quality of patients.

[0027] Other advantages, objects and features of the present application will be set forth in part in the following specification, and in part will become apparent to those skilled in the art upon examination of the following, or can be learned from practice of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0029] Figure 1 Flowchart of the heart rehabilitation action precision evaluation method for the wearable sensor device;

[0030] Figure 2 Structure diagram of the heart rehabilitation action precision evaluation system for the wearable sensor device. DETAILED DESCRIPTION

[0031] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purpose, the following will combine the drawings and preferred embodiments to specifically describe the specific embodiments, structures, features and effects according to the present application as follows.

[0032] Embodiment one

[0033] Referring to Figure 1 and Figure 2 In a professional heart rehabilitation medical institution, a large number of cardiovascular disease rehabilitation patients are received, and these patients have great differences in rehabilitation needs due to different conditions and physical conditions. In order to help patients obtain more accurate and effective rehabilitation treatment, the institution adopts the heart rehabilitation action precision evaluation method and system of the wearable sensor device of the present application.

[0034] Before the patient starts the rehabilitation training, the medical staff wears a multi-modal sensor array integrating an electrocardiogram sensor, an inertial measurement unit (IMU), an electromyography sensor and a pressure sensor for the patient. The electrocardiogram sensor is attached to the chest of the patient according to the specification to ensure that the cardiac electrical activity signal can be stably collected, and key data such as heart rate and heart rhythm can be obtained. The IMU is installed at the wrist, ankle and trunk pre-set positions, respectively, which can capture the acceleration and angular velocity of the body in three-dimensional space in all directions, and accurately monitor the motion posture. The electromyography sensor is attached to the surface of the main muscles participating in the rehabilitation action, such as the biceps brachii, triceps brachii of the upper limbs, quadriceps femoris and calf triceps of the lower limbs, etc., which is used to collect the electrical signals generated by muscle contraction and relaxation. The pressure sensor is installed at the key areas of the shoe sole in contact with the ground and the grip position of the rehabilitation equipment handle, which can collect limb pressure data in real time.

[0035] When the patient is performing a rehabilitation exercise such as walking, the multi-modal sensor array synchronously collects multi-dimensional data such as cardiac activity, motion posture, muscle electrical signal and limb pressure at a high frequency in real time. The collected data is transmitted to the data processing terminal through wireless transmission technology such as Bluetooth or Wi-Fi.

[0036] After receiving the multi-dimensional data, the data processing terminal first performs preprocessing, removes noise interference in the data using a digital filtering algorithm such as Kalman filtering, and identifies and repairs abnormal data using an outlier detection method based on statistical analysis.

[0037] Subsequently, an improved adaptive weighted fusion algorithm is started, and the weight of each data type is calculated according to the characteristics of different data types through the formula wherein S i represents the stability index of the i-th data type, which is determined by factors such as the fluctuation degree of the data, and C i represents the correlation index of the i-th data type with the rehabilitation action, which is pre-set according to medical research and clinical experience and adjusted in combination with real-time data. For example, in the evaluation of walking rehabilitation action, if the electrocardiogram data can well reflect the relationship between exercise intensity and cardiac load, the correlation index C i is relatively high.

[0038] The preprocessed data is then fused according to the formula wherein a graph reflecting the characteristics of the patient's walking rehabilitation action is generated, wherein D i is the preprocessed data of the i-th data type. Through this fusion method, different types of data are organically integrated to provide more accurate basis for subsequent evaluation.

[0039] Key features K are extracted from a large number of existing cardiac rehabilitation data and general models using a transfer learning mechanism, according to the formula wherein M is the number of general models, β m is the feature contribution coefficient of the m-th general model, which is determined according to the applicability of the model to the target patient group and the importance of the features, f m is a feature extraction function for the m-th general model, which is designed according to the model structure and data characteristics, and G m is the m-th general model, which constructs an initial model framework for the current patient based on the extracted key features K.

[0040] Then, by using the meta-learning mechanism, the historical rehabilitation data H, physiological indicators P (including age, gender, blood pressure, heart rate variability, etc.) and disease characteristics D (such as heart disease type, disease duration, treatment method, etc.) of the patient are calculated by the formula V = γ1 x H + γ2 x P + γ3 x D to obtain a comprehensive feature vector V, where γ1, γ2 and γ3 are weight coefficients of the historical rehabilitation data, physiological indicators and disease characteristics, respectively, which are determined through learning and analysis of a large amount of patient data. According to the comprehensive feature vector V, the model parameters are adjusted multiple times by using optimization algorithms such as gradient descent method to complete the construction of a personalized heart rehabilitation action precision evaluation model suitable for the patient.

[0041] Using the constructed personalized evaluation model, the walking rehabilitation action precision of the patient is evaluated according to the fused rehabilitation action feature map. After evaluation, the action deviation vector is calculated based on the evaluation results, and the formula is wherein, is the action deviation vector, and s is the number of feature points determined by action analysis; is the value of the kth feature point in the actual action feature map, is the value of the kth feature point in the standard rehabilitation action feature map, is the unit vector corresponding to the feature point. By analyzing the action deviation vector, the deviation direction and degree of the patient's rehabilitation action are determined.

[0042] At the same time, time series analysis is performed on the evaluation results of different stages to calculate the change rate of action precision data at adjacent evaluation time points and the correlation degree of precision between different rehabilitation actions, so as to mine the action change trend and potential law in the patient's rehabilitation process. Finally, the evaluation results are presented to medical staff and patients in the form of visual charts. The chart includes a line chart showing the change of action precision with time, which is convenient for observing the rehabilitation progress; a column chart comparing the precision of different rehabilitation actions, which helps patients and medical staff understand the mastery of different actions.

[0043] According to the evaluation results and analysis data, the medical staff dynamically adjusts the rehabilitation plan of the patient. If the evaluation finds that the patient's action precision is low when walking, and the action deviation vector shows that the step length and step frequency have problems, the medical staff will adjust the rehabilitation plan, such as adding a rhythm guiding link in the walking training, or using auxiliary equipment to help the patient standardize the step length. If the patient has good rehabilitation progress in a period of time and the action precision continues to improve, the medical staff will appropriately increase the training intensity according to the change trend obtained by time series analysis, such as prolonging the walking training time or increasing the walking speed requirement.

[0044] In summary, by applying the evaluation method and system of the present application in this embodiment, the accuracy, comprehensiveness and personalization of the evaluation of the patient's cardiac rehabilitation movement accuracy are realized. The data collected by the multi-modal sensor array is rich and accurate, and the data fusion algorithm ensures the effective integration of information, improves the objectivity and accuracy of the evaluation, and the personalized evaluation model fully considers the individual differences of patients, making the rehabilitation program more targeted. The rehabilitation program is dynamically adjusted in real time during the rehabilitation process, effectively improving the rehabilitation treatment effect, helping patients recover faster, improving exercise capacity and quality of life, and also improving the overall level and efficiency of cardiac rehabilitation treatment in medical institutions.

[0045] Embodiment Two

[0046] Referring to Figure 1 and Figure 2 In a community cardiac rehabilitation service center, rehabilitation services are provided for elderly people with chronic cardiovascular diseases in the surrounding area. These elderly people prefer to receive convenient rehabilitation treatment in the community due to their limited mobility. The community rehabilitation center has limited resources, but hopes to develop appropriate rehabilitation programs for the elderly through scientific methods to improve their quality of life and reduce the risk of cardiovascular disease recurrence. The cardiac rehabilitation movement accuracy evaluation method and system of the wearable sensor device of the present application provide strong support for the community rehabilitation center to solve this problem.

[0047] When the elderly come to the community rehabilitation center for rehabilitation training, the staff wears a multi-modal sensor array for them. The electrocardiogram sensor is carefully attached to the appropriate position on the chest of the elderly to ensure good contact for stable cardiac electrical signal collection. The IMU is installed on the wrist and ankle of the elderly to facilitate the monitoring of limb movement posture during daily activities. The electromyography sensor is attached to the main muscle parts of the arms and legs, such as the biceps and quadriceps, to obtain electrical signals during muscle activity. The pressure sensor is installed on the handle of the walking aid and the sole of the shoe used by the elderly in daily life to collect hand holding force and plantar pressure data during walking.

[0048] When the elderly perform simple stretching and balance training, such as upper limb stretching in standing position and standing balance exercise, the multi-modal sensor array starts to work, collecting real-time multi-dimensional data such as cardiac activity, movement posture, muscle electrical signal and limb pressure. The collected data is transmitted to the data processing terminal equipped in the community rehabilitation center through low-power Bluetooth technology, ensuring the stability and low energy consumption of data transmission.

[0049] After the data processing terminal receives the data, it first performs preprocessing, removes noise in the data using the median filtering algorithm, and finds and corrects abnormal data points using the outlier detection method based on clustering analysis. Then, the improved adaptive weighted fusion algorithm is started, and the formula is as follows: The weights of different data types are calculated, for example, in the balance training evaluation, if the pressure sensor data can well reflect the change of the old person's body center of gravity, its stability index S i and the correlation index C i will be given appropriate values according to the actual situation, and then according to the formula The preprocessed data is fused, wherein The atlas that can accurately reflect the characteristics of the old person's rehabilitation action is generated, which provides a comprehensive basis for subsequent evaluation.

[0050] With the help of transfer learning mechanism, key features K are extracted from existing heart rehabilitation data and general models, and the formula is Wherein, the number of general models M, the feature contribution coefficient β m of each general model, the feature extraction function f m and the general model G m are determined based on a large number of rehabilitation data of similar old patients with cardiovascular diseases and related research, and the initial model framework is constructed based on the extracted key features K.

[0051] Then, using the meta-learning mechanism, combining the old person's historical rehabilitation data H (if any), physiological indicators P (including age, blood pressure, heart rate, etc.) and disease characteristics D (such as cardiovascular disease type, disease duration, etc.), the comprehensive feature vector V is calculated through the formula V = γ1 × H + γ2 × P + γ3 × D, wherein γ1, γ2, γ3 are weight coefficients determined according to the characteristics of the old patient group, and according to the comprehensive feature vector V, the model parameters are adjusted multiple times by using the stochastic gradient descent method to construct a personalized heart rehabilitation action precision evaluation model suitable for the old person.

[0052] Using the constructed personalized evaluation model, the stretching and balance rehabilitation action precision of the old person is evaluated according to the fused rehabilitation action feature atlas, and after the evaluation is completed, the action deviation vector is calculated through the formula Through the analysis of the action deviation vector, the deviation direction and degree of the old person's rehabilitation action are determined, and at the same time, the time series analysis of the evaluation results at different stages is carried out, the change rate of the action precision data at adjacent evaluation time points is calculated, and the correlation degree between stretching and balance action precision is calculated, and the action change trend and potential law in the old person's rehabilitation process are mined.

[0053] Finally, the evaluation results are presented to the old person and the community rehabilitation staff in simple and easy-to-understand visual charts, including action completion condition comparison chart marked with different colors, and line chart showing the change of action precision with time, so that the old person and the staff can intuitively understand the rehabilitation progress.

[0054] The community rehabilitation worker adjusts the rehabilitation program for the old person according to the evaluation result and the analysis data, if it is found that the old person has low precision when performing the upper limb stretching action, and the action deviation vector shows that the stretching angle and speed of the arm have problems, the worker will simplify the action difficulty, such as reducing the target angle of stretching, or increasing the auxiliary tool to help the old person to complete the action, if the old person gradually improves the precision in the balance training, according to the result of time series analysis, the worker will appropriately increase the training difficulty, such as shortening the standing time interval or increasing some simple hand actions when standing.

[0055] In summary, by applying the evaluation method and system of the present application in the community heart rehabilitation service center, the precision of the rehabilitation action of the old cardiovascular disease patient is effectively evaluated, the use of the multi-modal sensor array can collect comprehensive data when the old people perform daily simple rehabilitation training, and the data fusion and personalized model construction fully consider the physical characteristics and individual differences of the old people, the rehabilitation program adjusted based on the precise evaluation result is more in line with the actual rehabilitation needs of the old people, and the safety and effectiveness of the rehabilitation training are improved.

[0056] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the present application, and any simplification, modification, equivalent change and modification of the above embodiments made according to the technical essence of the present application are within the scope of the present application.

Claims

1. A method for evaluating cardiac rehabilitation movement accuracy using a wearable sensor device, characterized in that: The method comprises the following steps: Utilize a multimodal sensor array module that integrates electrocardiogram (ECG) sensors, inertial measurement units (IMUs), myoelectric sensors, and pressure sensors to synchronously collect multi-dimensional data on cardiac activity, movement posture, muscle electrical signals, and limb pressure. The collected multi-dimensional data is transmitted to the data processing module. The improved adaptive weighted fusion algorithm is used to dynamically assign weights according to the characteristics of different data types, fuse the heterogeneous data, and generate the rehabilitation movement feature map. The calculation formula of the fused data F is: Where n is the number of data types, w i is the dynamic weight of the i-th data type, D i is the preprocessed data of the i-th data type, and Dynamic weight w i By formula Calculation, S i represents the stability index of the i-th data type, C i represents the correlation index between the i-th data type and rehabilitation action; Introducing transfer learning and meta-learning mechanisms to build a personalized cardiac rehabilitation movement accuracy assessment model based on patients' historical rehabilitation data, physiological indicators, and disease characteristics; The personalized assessment model is used to evaluate the accuracy of cardiac rehabilitation movements based on the fused rehabilitation movement feature map.

2. The cardiac rehabilitation movement accuracy assessment method of a wearable sensor device according to claim 1, characterized in that: When using the multimodal sensor array module to collect data, the electrocardiogram sensor is attached to the chest to collect cardiac electrical signals, the IMU is installed at preset positions on the wrist, ankle and torso to measure movement posture, the electromyography sensor is attached to the main muscle parts involved in rehabilitation movements to collect electromyography signals, and the pressure sensor is installed in the contact area between the sole and the ground and the handle of the rehabilitation equipment to collect limb pressure data.

3. The cardiac rehabilitation movement accuracy assessment method of a wearable sensor device according to claim 1, characterized in that: The improved adaptive weighted fusion algorithm preprocesses the multi-dimensional data before data fusion, uses a digital filtering algorithm to remove noise, and utilizes an outlier detection method based on statistical analysis to identify and repair abnormal data.

4. The cardiac rehabilitation movement accuracy assessment method of a wearable sensor device according to claim 1, characterized in that: The specific steps for building a personalized cardiac rehabilitation movement accuracy assessment model are as follows: The transfer learning mechanism extracts key features K from existing cardiac rehabilitation data and general models according to the preset feature extraction rules. The calculation formula is: Where M is the number of general models, β m is the characteristic contribution coefficient of the mth general model, f m is the feature extraction function for the mth general model, G m It is the mth general model, which builds the initial model framework based on the extracted key features K; The meta-learning mechanism calculates the comprehensive feature vector V based on the patient's historical rehabilitation data H, physiological indicators P and disease characteristics D through the formula V = γ1×H+γ2×P+γ3×D, where γ1, γ2, and γ3 are the weight coefficients of historical rehabilitation data, physiological indicators and disease characteristics, respectively. Based on the comprehensive feature vector V, the gradient descent method is used to iteratively adjust the model parameters multiple times to complete the construction of the personalized model.

5. The cardiac rehabilitation movement accuracy assessment method of a wearable sensor device according to claim 1, characterized in that: After evaluating the accuracy of cardiac rehabilitation movements, the movement deviation vector is calculated based on the evaluation results. The movement deviation vector is obtained by comparing the differences between the patient's actual movement feature map and the corresponding feature points in the standard rehabilitation movement feature map. The calculation formula is: in is the action deviation vector, s is the number of feature points, is the value of the kth feature point in the actual action feature map, is the value of the kth feature point in the standard rehabilitation action feature map, is the unit vector corresponding to the feature point.

6. The cardiac rehabilitation movement accuracy assessment method of a wearable sensor device according to claim 1, characterized in that: After evaluating the accuracy of cardiac rehabilitation movements, a time series analysis was conducted on the evaluation results at different stages. By calculating the rate of change of movement accuracy data at adjacent evaluation time points and the correlation between the accuracy of different rehabilitation movements, the trend of movement changes and potential patterns during the patient's rehabilitation process were explored.

7. The cardiac rehabilitation movement accuracy assessment method of a wearable sensor device according to claim 1, characterized in that: After the cardiac rehabilitation movement accuracy is evaluated, the results of the evaluation of the cardiac rehabilitation movement accuracy are presented in the form of a visual chart. The chart includes a line chart and a bar chart. The line chart shows the change of movement accuracy over time, and the bar chart compares the accuracy of different rehabilitation movements.

8. A cardiac rehabilitation movement accuracy assessment system for a wearable sensor device, applicable to the cardiac rehabilitation movement accuracy assessment method for a wearable sensor device according to claims 1-7, characterized in that: The system includes the following components: Multimodal sensor array module, used to integrate and collect multi-dimensional data of cardiac activity, movement posture, muscle electrical signals and limb pressure; A data transmission module, used to transmit the data collected by the multimodal sensor array module to the data processing module; The data processing module runs the improved adaptive weighted fusion algorithm to fuse the received data and generate a rehabilitation movement feature map; Personalized evaluation model library, which stores personalized cardiac rehabilitation movement accuracy evaluation models built based on transfer learning and meta-learning mechanisms; The evaluation module uses a personalized evaluation model to evaluate the accuracy of cardiac rehabilitation movements based on the rehabilitation movement characteristic map. The cardiac rehabilitation movement accuracy assessment system of a wearable sensor device according to claim 8 is characterized in that: The multimodal sensor array module adopts a modular design, each sensor module has high integration and low power consumption, and the modules exchange data and work together through standardized interfaces.

Images