Whole body exercise quality evaluation and suggestion generation system based on automatic identification

Abstract

The invention relates to the technical field of medicine and rehabilitation, and discloses an automatic recognition-based whole body exercise quality evaluation and suggestion generation system, which comprises a data acquisition module, a data processing module, an exercise recognition module, an evaluation analysis module, a suggestion generation module and a user interaction module which are in communication connection in sequence. According to the method, data are collected through a camera, human body key point labeling and association are performed in combination with growth data, motion features are extracted by applying a growth association learning algorithm, a motion quality score is obtained based on a GMS standard evaluation method and growth data analysis, personalized rehabilitation suggestions are generated based on a big data expert knowledge base, and the personalized rehabilitation suggestions are displayed through a graphical interface. The defects that a traditional manual evaluation mode consumes manpower and time and is high in subjectivity are effectively overcome, the efficiency, accuracy and individuation degree of whole-body exercise quality evaluation and suggestion generation are improved, and the urgent requirement for exercise function evaluation in the field of medical treatment and rehabilitation is met.

Description

Technical Field

[0001] The present application relates to the field of medicine and rehabilitation technology, and specifically to a whole-body movement quality assessment and suggestion generation system based on automatic recognition. Background Art

[0002] In the medical and rehabilitation fields, motor function assessment is crucial for disease diagnosis, monitoring rehabilitation progress, and developing intervention plans. Traditional motor function assessment methods, particularly whole-body movement quality assessment, rely heavily on manual video observation and analysis by professionals. This approach has numerous drawbacks. Firstly, it is labor-intensive and time-consuming, requiring professionals to focus on video details for extended periods, resulting in high workload and low efficiency. Secondly, subjectivity is inevitable, and different professionals may differ in their assessments of the same exercise performance, making it difficult to ensure the consistency and accuracy of assessment results.

[0003] With the rapid development of science and technology, artificial intelligence and computer vision technologies have gradually emerged and demonstrated strong application potential in many fields. In terms of motor function assessment, the assessment system based on automatic recognition and analysis has become a research hotspot. Chinese patent number CN118039068A discloses a comprehensive assessment method and system for human motor function. Its application is not very targeted and does not fully consider the special needs of specific groups (such as children with motor dysfunction). Furthermore, it is also lacking in the depth of technical integration, and has failed to give full play to the advantages of deep learning algorithms to achieve efficient and accurate motion feature extraction. In addition, it lacks personalized considerations in the generation of rehabilitation suggestions, making it difficult to meet the differentiated rehabilitation needs of individuals.

[0004] In summary, there is an urgent need for a new technical solution for whole-body movement quality assessment and suggestion generation based on automatic recognition to meet the urgent needs of motor function assessment in the medical and rehabilitation fields. Summary of the Invention

[0005] The purpose of this application is to provide a whole-body motion quality assessment and suggestion generation system based on automatic recognition to solve the technical problems raised in the above background technology.

[0006] To achieve the above objectives, the present application discloses the following technical solutions: a whole-body motion quality assessment and suggestion generation system based on automatic recognition, comprising a data acquisition module, a data processing module, a motion recognition module, an assessment and analysis module, a suggestion generation module, and a user interaction module, which are sequentially communicatively connected;

[0007] The data acquisition module is configured to: collect a video data set of the subject's movements based on a preset high-definition camera; wherein the video data set stores video data of the subject at different growth stages;

[0008] The data processing module is configured to: pre-process the video data, the pre-processing including denoising, frame extraction, and body key point labeling; wherein the body key point labeling is to associate the body key points of the video data set in a time series based on the growth data of the subject to obtain associated body key points, the growth data being the physiological changes of the subject, the physiological changes including at least height and weight, and the associated body key points being used to characterize the changes in the body key points of the subject at different growth stages;

[0009] The motion recognition module is configured to extract motion features from the associated human body key points using a preset growth association learning algorithm; wherein the growth association learning algorithm is used to extract features from the associated human body key points based on deep learning technology and the growth data to obtain the motion features;

[0010] The evaluation and analysis module is configured to analyze the movement characteristics based on the GMS standard evaluation method and the growth data, and output a movement quality score; wherein the movement quality score is used to characterize the normality of the subject's movement level, and the normality is used to determine the subject's movement pattern, and the movement pattern includes a normal movement pattern and an abnormal movement pattern;

[0011] The suggestion generation module is configured to generate rehabilitation suggestions based on the analysis results of the exercise quality score and the exercise characteristics and based on a preset big data expert knowledge base; wherein the rehabilitation suggestions are used for rehabilitation of the subject's exercise level;

[0012] The user interaction module is configured to display the rehabilitation advice to the subject using a preset graphical interface.

[0013] Preferably, the video data is preprocessed, and the preprocessing includes denoising, frame extraction, and human key point labeling, specifically:

[0014] Obtaining a preset subject database; wherein the subject database stores pre-processed records of the physiological data and video history data of multiple historical subjects;

[0015] Performing similarity analysis on the physiological data corresponding to the video data of the current subject and the physiological data of historical subjects in the subject database, outputting preprocessed records of the video history data of historical subjects that meet a preset physiological data similarity threshold, and performing feature extraction on the preprocessed records including denoising features, frame extraction features, and human body key point annotation features;

[0016] The video data of the current subject is preprocessed using the extracted denoising features, the frame extraction features, and the human key point labeling features.

[0017] Preferably, the associated human body key points are specifically:

[0018] Get the change amplitude ΔK of the subject in the i-th growth stage i and the frequency of change f i , calculate the comprehensive value S of the change in each growth stage of the subject i =ΔK i Based on the size of the comprehensive value of the change, the human body key points are divided into n categories and then classified and extracted, and the human body key points of the same category are associated to obtain associated human body key points.

[0019] Preferably, the growth association learning algorithm is specifically:

[0020] Extracting spatial features of the associated human body key points using a convolutional neural network to obtain spatial features;

[0021] Extracting dynamic features of the time series of the associated human body key points using a recurrent neural network to obtain dynamic features;

[0022] The motion feature is calculated using the spatial feature, the dynamic feature, and the rate of change of the growth data. The calculation is:

[0023]

[0024] Among them, e is the natural base, F s is the spatial feature, F t is a dynamic feature, ΔH is the rate of change of the subject's height within the preset monitoring period T, ΔW is the rate of change of the subject's weight within the preset monitoring period T, a, b, c and d are preset weight parameters, and F is the calculated motion feature.

[0025] Preferably, the exercise quality score is specifically:

[0026] Obtain the characteristic value of each dimension of the motion feature and the corresponding importance weight at the growth stage corresponding to the growth data, and calculate the motion quality score based on the characteristic value and the importance weight. The calculation is:

[0027]

[0028] Among them, v j is the eigenvalue of the jth dimension of the motion feature, ω jis the importance weight corresponding to the eigenvalue of the j-th dimension of the motion feature, m is the total number of dimensions of the motion feature, and Q is the calculated motion quality score; wherein, the importance weight is updated based on a preset growth dynamic mechanism, and the growth dynamic mechanism is used to dynamically weight the eigenvalues ​​of each dimension based on the importance of different growth stages and the criticality of the motion feature.

[0029] Preferably, the growth dynamic mechanism is specifically:

[0030] Get the weight of the jth dimension in the kth growth stage as ω j,k and the proportion of this growth stage in the current assessment k , calculate the importance weight ω based on the weight and the proportion j , the calculation is:

[0031]

[0032] Here, p is the number of growth stages.

[0033] Preferably, the rehabilitation suggestions are:

[0034] Obtaining a preset standard motion quality score, and calculating a deviation value between the motion quality score and the standard motion quality score;

[0035] The daily activity level and exercise preference of the subject are collected, and a rehabilitation index is calculated based on a rehabilitation index calculation formula. The rehabilitation index is used to screen the big data expert knowledge base and generate rehabilitation suggestions; wherein the rehabilitation index calculation formula is:

[0036] R=γ*ΔQ+δ*A+ε*P

[0037] Wherein, ΔQ is the deviation value, A is the daily activity amount, P is the degree of exercise preference, and γ, δ and ε are preset influence coefficients.

[0038] Preferably, the growth association learning algorithm further includes:

[0039] Based on the GMS standard evaluation method, the standard motion feature template is extracted, and the deviation between the extracted motion feature and the standard motion feature template is analyzed to obtain the feature deviation ΔF. When ΔF>ΔF τ When the weight parameters a, b, c and d preset by the growth association learning algorithm are readjusted, and the motion features are re-extracted.

[0040] Preferably, the rehabilitation suggestion further includes:

[0041] Collecting the exercise quality scores of subjects that meet a preset physiological data similarity threshold to obtain a similar exercise quality score set, and calculating the average value and standard deviation of the similar exercise quality score set;

[0042] A special rehabilitation index is calculated using a preset special rehabilitation index calculation formula. The special rehabilitation index is used to add special training content when generating the rehabilitation suggestion. The special training content is obtained by screening the special rehabilitation index in the big data expert knowledge base; wherein the special rehabilitation index calculation formula is:

[0043]

[0044] in, is the average value of the same type of motion quality score set, | | is the absolute value operator, σ is the variance of the same type of motion quality score set, g is the preset deviation multiple, c is the special method index, and R s is the calculated special rehabilitation index.

[0045] Preferably, the display of the rehabilitation suggestion is specifically:

[0046] The rehabilitation advice also includes a rehabilitation instruction video, which stores a plurality of rehabilitation training movements of different complexity and importance. When the rehabilitation training movements are presented in the rehabilitation advice, the corresponding video playback duration is calculated based on the complexity and importance. The video playback duration is calculated as follows:

[0047] T=φ*CO+ρ*IM

[0048] Among them, CO is the complexity, IM is the importance, φ and ρ are the preset time allocation coefficients, and T is the calculated video playback time.

[0049] Beneficial Effects: This application's system for whole-body exercise quality assessment and recommendation generation based on automatic recognition automates the entire process of video data collection, preprocessing, feature extraction, quality scoring, and the generation and presentation of rehabilitation recommendations for subjects at different growth stages. Data is collected via a high-definition camera, combined with growth data to annotate and associate key points on the human body. A growth-related learning algorithm is used to extract movement features. An exercise quality score is derived based on the GMS standard evaluation method and growth data analysis. Personalized rehabilitation recommendations are then generated based on a big data expert knowledge base and presented through a graphical interface. This system effectively overcomes the time-consuming and subjective drawbacks of traditional manual assessment methods, improving the efficiency, accuracy, and personalization of whole-body exercise quality assessment and recommendation generation, thus meeting the urgent need for motor function assessment in the medical and rehabilitation fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 This is a structural block diagram of the system for whole-body motion quality assessment and suggestion generation based on automatic recognition provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] The following is a clear and complete description of the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0053] In this document, the term "comprising" is intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0054] The first aspect of this embodiment discloses Figure 1 The system for evaluating and generating whole-body motion quality based on automatic identification includes a data acquisition module, a data processing module, a motion recognition module, an evaluation and analysis module, a suggestion generation module, and a user interaction module, which are sequentially communicatively connected;

[0055] The data acquisition module is configured to: collect a video data set of the subject's movements based on a preset high-definition camera; wherein the video data set stores video data of the subject at different growth stages;

[0056] The data processing module is configured to: pre-process the video data, the pre-processing including denoising, frame extraction, and body key point labeling; wherein the body key point labeling is based on the growth data of the subject, correlating the body key points of the video data set in a time series to obtain associated body key points, the growth data being the physiological changes of the subject, the physiological changes including at least height and weight, and the associated body key points being used to characterize the changes in the body key points of the subject at different growth stages;

[0057] The motion recognition module is configured to: extract motion features from associated human key points using a preset growth association learning algorithm; wherein the growth association learning algorithm is used to extract features from associated human key points based on deep learning technology and growth data to obtain motion features;

[0058] The evaluation and analysis module is configured to analyze movement characteristics based on the GMS standard evaluation method and growth data and output a movement quality score; wherein the movement quality score is used to characterize the normality of the subject's movement level, and the normality is used to determine the subject's movement pattern, which includes a normal movement pattern and an abnormal movement pattern;

[0059] The suggestion generation module is configured to: generate rehabilitation suggestions based on the analysis results of the exercise quality score and exercise characteristics and based on a preset big data expert knowledge base; wherein the rehabilitation suggestions are used for the rehabilitation of the subject's exercise level;

[0060] The user interaction module is configured to display rehabilitation suggestions to the subject using a preset graphical interface.

[0061] Through the above, this embodiment realizes the full-process automation of video data collection, preprocessing, feature extraction, quality scoring, and rehabilitation suggestion generation and display of subjects' movements at different growth stages; data is collected by high-definition cameras, and key points of the human body are labeled and associated in combination with growth data. The growth association learning algorithm is used to extract movement features, and the movement quality score is obtained based on the GMS standard evaluation method and growth data analysis. Personalized rehabilitation suggestions are generated based on the big data expert knowledge base and displayed through a graphical interface, which effectively overcomes the shortcomings of traditional manual evaluation methods that are time-consuming and highly subjective, improves the efficiency, accuracy and personalization of whole-body movement quality assessment and suggestion generation, and meets the urgent needs of the medical and rehabilitation fields for movement function assessment.

[0062] Specifically, the video data is preprocessed, which includes denoising, frame extraction and human key point annotation, specifically:

[0063] Obtaining a preset subject database; wherein the subject database stores pre-processed records of physiological data and historical video data of multiple historical subjects;

[0064] Perform similarity analysis on the physiological data corresponding to the video data of the current subject and the physiological data of historical subjects in the subject database, output the preprocessed records of the video data of the historical subjects that meet the preset physiological data similarity threshold, and perform feature extraction of denoising features, frame extraction features, and human key point annotation features on the preprocessed records; it should be noted that the feature extraction in this embodiment includes at least the angle, speed, acceleration, amplitude, and smoothness of the key point movement;

[0065] The extracted denoising features, frame extraction features and human key point annotation features are used to preprocess the video data of the current subject.

[0066] Through the above, this embodiment achieves more accurate preprocessing of the current subject's video data. Based on the preprocessing experience of historical subjects, it extracts denoising, frame extraction and human key point labeling features to process the current data. Compared with conventional preprocessing methods, it can better adapt to different individual differences and reduce the preprocessing errors caused by individual differences, thereby improving the accuracy of subsequent motion feature extraction and analysis, providing a more reliable data basis for whole-body motion quality assessment and suggestion generation, and enhancing the stability and reliability of the entire system.

[0067] Specifically, the key points of the human body are:

[0068] Get the change amplitude ΔK of the subject in the i-th growth stage i and the frequency of change f i , calculate the comprehensive value S of the change in each growth stage of the subject i =ΔK i , based on the size of the comprehensive value of the change, the human body key points are divided into n categories and then classified and extracted, and the human body key points of the same category are associated to obtain associated human body key points.

[0069] Through the above, this embodiment realizes the effective organization and utilization of key points of the human body, classifies the comprehensive change values ​​determined based on the change amplitude and frequency, highlights the key change information of key points of the human body at different growth stages, and makes the associated key points of the human body more representative and targeted. In the subsequent motion feature extraction process, it can more accurately reflect the changes in the motion characteristics of the subjects with the growth stage, improve the accuracy and effectiveness of the motion recognition module in extracting motion features, and thus improve the accuracy of the whole-body motion quality assessment.

[0070] Specifically, the growth association learning algorithm is as follows:

[0071] The convolutional neural network is used to extract the spatial features of the key points of the human body to obtain spatial features;

[0072] The dynamic features of the time series of key points of the human body are extracted using a recurrent neural network to obtain dynamic features;

[0073] The motion characteristics are calculated using the spatial characteristics, dynamic characteristics and the rate of change of growth data. The calculation is:

[0074]

[0075] Among them, e is the natural base, F sis the spatial feature, F t is a dynamic feature, ΔH is the rate of change of the subject's height within the preset monitoring period T, ΔW is the rate of change of the subject's weight within the preset monitoring period T, a, b, c and d are preset weight parameters, and F is the calculated motion feature.

[0076] It can be understood that in a specific application of this embodiment, when applied to the assessment of the quality of whole-body movement in early childhood, the preset monitoring period T can be the child's age in months. The change in the child's height is more reflected in the spatial characteristics than the change in weight. Similarly, the change in the child's weight is more reflected in the movement characteristics than the change in height. Therefore, this embodiment designs the use of ΔH and F s Binding, ΔW, and F t The combined approach improves the representativeness of sports features.

[0077] Based on the above, this embodiment utilizes existing convolutional neural networks and recurrent neural networks, as well as growth data change rate calculations to achieve comprehensive and accurate extraction of motion features. This combines spatial features, dynamic features, and growth data change rates, and through calculations, fully considers the impact of body structure changes on movement and the spatiotemporal characteristics of movement. Compared with traditional feature extraction methods, this method can more deeply explore motion information, providing richer and more accurate feature data for the evaluation and analysis module, thereby improving the reliability of movement quality scores and laying the foundation for generating reasonable rehabilitation recommendations.

[0078] Specifically, the movement quality score is as follows:

[0079] Obtain the eigenvalues ​​of each dimension of the motion feature and the corresponding importance weights of the growth stage corresponding to the growth data, and calculate the motion quality score based on the eigenvalues ​​and importance weights. The calculation is:

[0080]

[0081] Among them, v j is the eigenvalue of the jth dimension of the motion feature, ω j is the importance weight corresponding to the eigenvalue of the j-th dimension of the motion feature, m is the total number of dimensions of the motion feature, and Q is the calculated motion quality score; wherein the importance weight is updated based on a preset growth dynamic mechanism, and the growth dynamic mechanism is used to dynamically weight the eigenvalues ​​of each dimension based on the importance of different growth stages and the criticality of the motion feature.

[0082] Based on the above, this embodiment utilizes a growth-based dynamic mechanism to calculate the exercise quality score, achieving dynamic optimization of the score. Dynamically weighting the characteristic values ​​of each dimension based on the importance of different growth stages and the criticality of movement characteristics makes the exercise quality score more tailored to the subject's actual situation. Compared to a scoring method with fixed weights, this method can more accurately reflect the normal level of exercise level of the subject at different growth stages, providing a more scientific basis for the subsequent generation of rehabilitation recommendations, and improving the accuracy and adaptability of the entire system in assessing the subject's motor function.

[0083] Specifically, the growth dynamic mechanism is as follows:

[0084] Get the weight of the jth dimension in the kth growth stage as ω j,k and the proportion of this growth stage in the current assessment k , calculate the importance weight ω based on the weight and the proportion j , the calculation is:

[0085]

[0086] Here, p is the number of growth stages.

[0087] Based on the above, this embodiment utilizes a dynamic growth mechanism that calculates importance weights based on dimension weights and growth stage proportions to achieve scientific and reasonable weight distribution. By comprehensively considering multiple factors to determine weights, the one-sidedness of weight determination by a single factor is avoided. When calculating the exercise quality score, the actual contribution of the characteristic values ​​of each dimension can be more accurately reflected, improving the accuracy of the score, making the evaluation results more reflective of the subject's actual exercise status, and providing a more reliable basis for the generation of rehabilitation recommendations.

[0088] Specific rehabilitation recommendations include:

[0089] Obtaining a preset standard motion quality score, and calculating a deviation value between the motion quality score and the standard motion quality score;

[0090] The subjects' daily activity levels and exercise preferences were collected, and rehabilitation indicators were calculated based on the rehabilitation indicator calculation formula. The rehabilitation indicators were used to screen the big data expert knowledge base and generate rehabilitation suggestions. The rehabilitation indicator calculation formula is:

[0091] R=γ*ΔQ+δ*A+ε*P

[0092] Among them, ΔQ is the deviation value, A is the daily activity amount, P is the degree of exercise preference, and γ, δ, and ε are the preset influence coefficients.

[0093] Based on the above, this embodiment uses the calculation of the deviation value of the exercise quality score and the combination of the subject's daily activity level and exercise preference to generate rehabilitation indicators, thereby realizing personalized customization of rehabilitation suggestions. Comprehensive consideration of multiple factors to determine rehabilitation indicators can better meet the individual differences and actual needs of different subjects compared to the method of generating suggestions based solely on exercise quality scores. Rehabilitation suggestions that are more suitable for the subjects are screened out from the big data expert knowledge base, improving the pertinence and effectiveness of rehabilitation suggestions and promoting the recovery of the subjects' exercise levels.

[0094] Specifically, the growth association learning algorithm also includes:

[0095] Based on the GMS standard evaluation method, the standard motion feature template is extracted and the deviation between the extracted motion features and the standard motion feature template is analyzed to obtain the feature deviation ΔF. When ΔF>ΔF τ When the weight parameters a, b, c and d preset by the growth association learning algorithm are readjusted, and the motion features are re-extracted.

[0096] It should be noted that the GMS standard evaluation method of this embodiment is an existing movement evaluation method used for premature infants and newborns with high-risk factors during pregnancy, delivery, and after delivery. It is an ultra-early screening and evaluation method. During the evaluation process, the newborn lies on his back in an awake state and is filmed continuously for 5-10 minutes to provide data for analyzing movement characteristics. In the journal "Advanced Science", the GMS standard evaluation method is described in detail in the paper "Intelligence Sparse Sensor Network for Automatic Early Evaluation of General Movements in Infants" published by Dr. Bao Benkun and Dr. Zhang Senhao as co-first authors on March 6, 2024. It can be understood that as an existing technology, the standard motion video of the GMS standard evaluation method is extracted using existing feature extraction technology to obtain the corresponding standard motion feature template, thereby providing accurate data reference for the deviation analysis of motion features.

[0097] As described above, this embodiment utilizes deviation analysis and a weight parameter adjustment mechanism to achieve self-optimization of motion feature extraction. By comparing with a standard motion feature template, potential deviations are promptly detected and corrected. When the deviation exceeds a threshold, the weight parameters of the growth association learning algorithm are readjusted and the motion features are re-extracted. This ensures the accuracy and reliability of the extracted motion features, improves the performance of the motion recognition module, and ultimately enhances the assessment accuracy of the entire system.

[0098] Specific rehabilitation recommendations also include:

[0099] Collecting the exercise quality scores of subjects that meet a preset physiological data similarity threshold to obtain a similar exercise quality score set, and calculating the mean and standard deviation of the similar exercise quality score set;

[0100] The special rehabilitation index is calculated using the preset special rehabilitation index calculation formula. The special rehabilitation index is used to add special training content when generating rehabilitation suggestions. The special training content is obtained by screening the special rehabilitation index in the big data expert knowledge base. The special rehabilitation index calculation formula is:

[0101]

[0102] in, is the average value of the same type of sports quality score set, | | is the absolute value operator, σ is the variance of the same type of sports quality score set, g is the preset deviation multiple, c is the special method index, and R s is the calculated special rehabilitation index.

[0103] As described above, this embodiment utilizes a similar set of exercise quality scores to calculate specialized rehabilitation indicators, enabling targeted reinforcement of rehabilitation recommendations. By analyzing the exercise quality scores of similar subjects, specialized rehabilitation indicators are determined, and specialized training content is added when generating rehabilitation recommendations. Compared to standard rehabilitation recommendations, this allows for more precise training reinforcement tailored to the subject's specific situation, improving rehabilitation outcomes and accelerating the recovery of the subject's motor function.

[0104] Specifically, show rehabilitation suggestions, specifically:

[0105] Rehabilitation advice also includes rehabilitation instruction videos, which contain multiple rehabilitation exercises of varying complexity and importance. When presenting rehabilitation advice, the corresponding video playback duration is calculated based on the complexity and importance of each exercise. The video playback duration is calculated as follows:

[0106] T=φ*CO+ρ*IM

[0107] Among them, CO is the complexity, IM is the importance, φ and ρ are the preset time allocation coefficients, and T is the calculated video playback time.

[0108] Based on the above, this embodiment calculates the video playback duration based on the complexity and importance of the rehabilitation training movements, achieving the rationality and effectiveness of the rehabilitation advice display. The display time is reasonably allocated according to the characteristics of different movements, enabling the subjects and their assistants to more clearly understand and learn the rehabilitation training methods. Compared with the display method with a uniform playback duration, this improves the practicality of the rehabilitation guidance video, helps subjects better perform rehabilitation training, and improves the rehabilitation effect.

[0109] In summary, the automatic recognition-based whole-body exercise quality assessment and recommendation generation system of this embodiment automates the entire process of video data acquisition, preprocessing, feature extraction, quality scoring, and the generation and presentation of rehabilitation recommendations for subjects at different growth stages. Data is collected using a high-definition camera, combined with growth data to annotate and associate key points on the human body. A growth-related learning algorithm is used to extract movement features. An exercise quality score is derived based on the GMS standard evaluation method and growth data analysis. Personalized rehabilitation recommendations are then generated based on a big data expert knowledge base and presented through a graphical interface. This effectively overcomes the time-consuming and subjective nature of traditional manual assessment methods, improving the efficiency, accuracy, and personalization of whole-body exercise quality assessment and recommendation generation, thus meeting the urgent need for motor function assessment in the medical and rehabilitation fields.

[0110] In the embodiments provided herein, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any appropriate combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a processor, a controller, a microcontroller, a microprocessor, other electronic units designed to implement the functions described herein, or a combination thereof. For software implementation, part or all of the processes of the embodiments can be completed by instructing the relevant hardware through a computer program. When implemented, the above program can be stored in a computer-readable storage medium or transmitted as one or more instructions or codes on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein the communication media include any medium that facilitates the transmission of a computer program from one place to another. The storage medium can be any available medium that a computer can access. The computer-readable storage medium can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer.

[0111] Finally, it should be noted that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent replacements for some of the technical features therein. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A whole-body exercise quality assessment and suggestion generation system based on automatic recognition, characterized in that: It includes a data acquisition module, a data processing module, a motion recognition module, an evaluation and analysis module, a suggestion generation module and a user interaction module which are sequentially connected in communication; The data acquisition module is configured to: collect a video data set of the subject's movements based on a preset high-definition camera; wherein the video data set stores video data of the subject at different growth stages; The data processing module is configured to: pre-process the video data, the pre-processing including denoising, frame extraction, and body key point labeling; wherein the body key point labeling is to associate the body key points of the video data set in a time series based on the growth data of the subject to obtain associated body key points, the growth data being the physiological changes of the subject, the physiological changes including at least height and weight, and the associated body key points being used to characterize the changes in the body key points of the subject at different growth stages; The motion recognition module is configured to extract motion features from the associated human body key points using a preset growth association learning algorithm; wherein the growth association learning algorithm is used to extract features from the associated human body key points based on deep learning technology and the growth data to obtain the motion features; The evaluation and analysis module is configured to analyze the movement characteristics based on the GMS standard evaluation method and the growth data, and output a movement quality score; wherein the movement quality score is used to characterize the normality of the subject's movement level, and the normality is used to determine the subject's movement pattern, and the movement pattern includes a normal movement pattern and an abnormal movement pattern; The suggestion generation module is configured to generate rehabilitation suggestions based on the analysis results of the exercise quality score and the exercise characteristics and based on a preset big data expert knowledge base; wherein the rehabilitation suggestions are used for rehabilitation of the subject's exercise level; The user interaction module is configured to display the rehabilitation advice to the subject using a preset graphical interface.

2. The system for whole-body exercise quality assessment and suggestion generation based on automatic recognition according to claim 1, characterized in that: The video data is preprocessed, including denoising, frame extraction, and human key point annotation, specifically: Obtaining a preset subject database; wherein the subject database stores pre-processed records of the physiological data and video history data of multiple historical subjects; Performing similarity analysis on the physiological data corresponding to the video data of the current subject and the physiological data of historical subjects in the subject database, outputting preprocessed records of the video history data of historical subjects that meet a preset physiological data similarity threshold, and performing feature extraction on the preprocessed records including denoising features, frame extraction features, and human body key point annotation features; The video data of the current subject is preprocessed using the extracted denoising features, the frame extraction features, and the human key point labeling features.

3. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 1, characterized in that: The associated human body key points are specifically: Get the change amplitude ΔK of the subject in the i-th growth stage i and the frequency of change f i , calculate the comprehensive value S of the change in each growth stage of the subject i =ΔK i Based on the size of the comprehensive value of the change, the human body key points are divided into n categories and then classified and extracted, and the human body key points of the same category are associated to obtain associated human body key points.

4. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 1, characterized in that: The growth association learning algorithm is specifically: Extracting spatial features of the associated human body key points using a convolutional neural network to obtain spatial features; Extracting dynamic features of the time series of the associated human body key points using a recurrent neural network to obtain dynamic features; The motion feature is calculated using the spatial feature, the dynamic feature, and the rate of change of the growth data. The calculation is: Among them, e is the natural base, F s is the spatial feature, F t is a dynamic feature, ΔH is the rate of change of the subject's height within the preset monitoring period T, ΔW is the rate of change of the subject's weight within the preset monitoring period T, a, b, c and d are preset weight parameters, and F is the calculated motion feature.

5. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 1, characterized in that: The movement quality score is specifically: Obtain the characteristic value of each dimension of the motion feature and the corresponding importance weight at the growth stage corresponding to the growth data, and calculate the motion quality score based on the characteristic value and the importance weight. The calculation is: Among them, v j is the eigenvalue of the jth dimension of the motion feature, ω j is the importance weight corresponding to the eigenvalue of the j-th dimension of the motion feature, m is the total number of dimensions of the motion feature, and Q is the calculated motion quality score; wherein, the importance weight is updated based on a preset growth dynamic mechanism, and the growth dynamic mechanism is used to dynamically weight the eigenvalues ​​of each dimension based on the importance of different growth stages and the criticality of the motion feature.

6. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 5, characterized in that: The growth dynamic mechanism is specifically: Get the weight of the jth dimension in the kth growth stage as ω j,k and the proportion of this growth stage in the current assessment k , calculate the importance weight ω based on the weight and the proportion j , the calculation is: Here, p is the number of growth stages.

7. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 1, characterized in that: The rehabilitation recommendations are as follows: Obtaining a preset standard motion quality score, and calculating a deviation value between the motion quality score and the standard motion quality score; The daily activity level and exercise preference of the subject are collected, and a rehabilitation index is calculated based on a rehabilitation index calculation formula. The rehabilitation index is used to screen the big data expert knowledge base and generate rehabilitation suggestions; wherein the rehabilitation index calculation formula is: R=γ*ΔQ+δ*A+ε*P Wherein, ΔQ is the deviation value, A is the daily activity amount, P is the degree of exercise preference, and γ, δ and ε are preset influence coefficients.

8. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 4, characterized in that: The growth association learning algorithm further includes: Based on the GMS standard evaluation method, the standard motion feature template is extracted, and the deviation between the extracted motion feature and the standard motion feature template is analyzed to obtain the feature deviation ΔF. When ΔF>ΔF τ When the weight parameters a, b, c and d preset by the growth association learning algorithm are readjusted, and the motion features are re-extracted.

9. The system for assessing whole-body exercise quality and generating suggestions based on automatic recognition according to claim 7, characterized in that: The rehabilitation recommendations also include: Collecting the exercise quality scores of subjects that meet a preset physiological data similarity threshold to obtain a similar exercise quality score set, and calculating the average value and standard deviation of the similar exercise quality score set; A special rehabilitation index is calculated using a preset special rehabilitation index calculation formula. The special rehabilitation index is used to add special training content when generating the rehabilitation suggestion. The special training content is obtained by screening the special rehabilitation index in the big data expert knowledge base; wherein the special rehabilitation index calculation formula is: in, is the average value of the same type of motion quality score set, | | is the absolute value operator, σ is the variance of the same type of motion quality score set, g is the preset deviation multiple, c is the special method index, and R s is the calculated special rehabilitation index.

10. The system for whole-body exercise quality assessment and suggestion generation based on automatic recognition according to claim 1, characterized in that: The display of the rehabilitation suggestion is specifically: The rehabilitation advice also includes a rehabilitation instruction video, which stores a plurality of rehabilitation training movements of different complexity and importance. When the rehabilitation training movements are presented in the rehabilitation advice, the corresponding video playback duration is calculated based on the complexity and importance. The video playback duration is calculated as follows: T=φ*CO+ρ*IM Among them, CO is the complexity, IM is the importance, φ and ρ are the preset time allocation coefficients, and T is the calculated video playback time.

Images