Old age rehabilitation training action simulation guiding system

Abstract

This invention relates to the field of rehabilitation medical equipment technology, specifically to a simulation guidance system for elderly rehabilitation training movements. The system includes a data acquisition module, a personalized modeling module, a simulation guidance module, a real-time feedback module, a risk warning module, a data storage and analysis module, and an interactive terminal. The data acquisition module collects movement data and physiological parameters of the elderly through visual sensors, inertial measurement units, and pressure sensors. The personalized modeling module constructs a musculoskeletal simulation model based on the physiological characteristics of the elderly. The simulation guidance module demonstrates standardized rehabilitation movements using a 3D virtual human and supports movement decomposition. The real-time feedback module combines force feedback devices with voice / visual prompts to achieve movement error correction. The risk warning module monitors parameters such as joint load and muscle tension and triggers graded warnings. The data storage and analysis module records training data and generates rehabilitation assessment reports. This invention achieves accurate simulation, personalized guidance, and safety protection for rehabilitation movements.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation medical equipment technology, specifically to a simulation guidance system for rehabilitation training movements in the elderly. Background Technology

[0002] With the deepening of population aging, the number of elderly people in my country continues to grow. The number of elderly people suffering from motor dysfunction due to stroke, joint replacement surgery, osteoporosis, sarcopenia, and other reasons is increasing year by year. Rehabilitation training has become a key means to improve their motor function and enhance their ability to live independently. The core requirements of rehabilitation training are standardized movements, appropriate intensity, and controllable safety. However, there are many problems in the field of geriatric rehabilitation training that urgently need to be solved, seriously affecting the rehabilitation effect and safety.

[0003] First, traditional rehabilitation training heavily relies on the manual guidance of professional rehabilitation therapists. Due to the scarcity and uneven distribution of rehabilitation therapists, most elderly people struggle to receive consistent and standardized guidance, especially in home-based rehabilitation settings. Trainees often rely solely on video or written tutorials, unable to assess the accuracy of their movements. Incorrect movements not only reduce rehabilitation effectiveness but can also lead to secondary injuries such as joint damage and muscle strains. For example, if stroke patients perform limb extension exercises with incorrect postures for an extended period, it may worsen spasticity symptoms and prolong the rehabilitation process.

[0004] Secondly, existing rehabilitation simulation guidance systems lack personalized adaptation for the elderly. The elderly generally exhibit physiological characteristics such as reduced muscle mass, decreased joint mobility, and decreased bone density, with significant individual differences (e.g., different patients have different muscle strength levels and joint limitations). However, existing systems mostly use uniform simulation movement models, failing to consider the specific physiological characteristics of the elderly. For example, some rehabilitation systems designed for younger people have movement ranges and force requirements exceeding the physical tolerance of the elderly, making it difficult for them to complete training or causing safety risks due to forced imitation. Furthermore, the motion capture accuracy of existing systems is insufficient, relying heavily on a single visual sensor, which is easily affected by factors such as ambient light and occlusion, making it impossible to accurately acquire key data such as joint angles and muscle exertion, resulting in a disconnect between simulation guidance and actual training.

[0005] Secondly, existing systems lack real-time and effective feedback and risk warning mechanisms. During rehabilitation training, deviations in the trainee's movements and changes in physical load need to be detected and corrected promptly. However, existing systems mostly only provide simple data statistics after training and cannot provide real-time alerts for movement errors. Furthermore, they lack safety threshold monitoring for the elderly, whose musculoskeletal tolerance is lower. When joint angles or muscle tension exceed their physiological limits, the lack of timely warnings and interventions increases training risks. In addition, existing systems have weak data storage and analysis capabilities, making it impossible to track rehabilitation progress over the long term or to synchronize training information with medical institutions or family members, which is detrimental to forming a closed-loop rehabilitation management system.

[0006] Finally, the existing system's user experience does not align with the habits of older adults. Older adults generally experience a decline in vision, hearing, and operational abilities, but the existing system's interface design is complex, the font is too small, the voice prompts are unclear, and it lacks convenient operation methods such as voice interaction, making it difficult for older users and reducing the system's practicality and acceptance.

[0007] In summary, the current field of elderly rehabilitation training suffers from problems such as high dependence on manual intervention, insufficient personalization, low capture accuracy, delayed feedback, high safety risks, and unfriendly interaction. There is an urgent need for a rehabilitation training system that can adapt to the physiological characteristics of the elderly, achieve precise simulation guidance, provide real-time feedback and error correction, and provide safety risk warnings to address the shortcomings of existing technologies. Summary of the Invention

[0008] To address the following technical problems in existing technologies: 1) lack of personalized simulation guidance tailored to the physiological characteristics of the elderly, resulting in highly generalized but poorly adaptable motion models; 2) insufficient motion capture accuracy, failing to accurately acquire key data such as joint angles and muscle exertion, leading to a disconnect between simulation and actual training; 3) lagging feedback mechanisms, unable to correct motion deviations in real time, affecting rehabilitation outcomes; 4) lack of targeted safety risk warnings, easily causing secondary injuries; 5) weak data management capabilities, unable to achieve rehabilitation progress tracking and multi-party collaboration; 6) interactive experience not conforming to the usage habits of the elderly, making operation difficult. This invention provides an elderly rehabilitation training motion simulation guidance system.

[0009] The technical solution adopted by this invention to solve its technical problem is: a simulation guidance system for elderly rehabilitation training movements, comprising: a data acquisition module for collecting the trainee's movement data, physiological parameters, and training environment data; a personalized modeling module, communicatively connected to the data acquisition module, for constructing a musculoskeletal simulation model exclusive to the trainee based on the movement data and physiological parameters, wherein the physiological parameters include height, weight, joint range of motion, muscle strength, bone density, and past medical history; and a simulation guidance module, communicatively connected to the personalized modeling module, having a built-in rehabilitation movement database, for generating standardized simulation movements adapted to the trainee based on the musculoskeletal simulation model, for real-time demonstration through a 3D virtual human, and for supporting movement decomposition teaching; real-time... The feedback module communicates with both the data acquisition module and the simulation guidance module, comparing the acquired motion data with the simulated motion data and correcting motion deviations through force feedback devices, voice prompts, and visual pop-ups. The risk warning module communicates with both the data acquisition module and the personalized modeling module, presets joint load thresholds, muscle tension thresholds, and motion speed thresholds, and monitors in real time whether parameters exceed limits during training, triggering tiered warnings. The data storage and analysis module communicates with all the modules, storing training data, physiological parameters, and warning records, and generating rehabilitation progress reports through data analysis. The interactive terminal displays simulated motions, feedback information, warning prompts, and rehabilitation reports, and supports input of user commands.

[0010] Specifically, the data acquisition module includes a visual sensor, an inertial measurement unit, a pressure sensor, and a physiological monitoring sensor. The visual sensor uses a Kinect V2 depth camera, the inertial measurement unit has an MPU9250 chip built in, the pressure sensor uses an FSR402 flexible pressure sensor, and the physiological monitoring sensor is used to collect heart rate, blood oxygen saturation, and electromyography signals.

[0011] Specifically, the personalized modeling module uses the finite element analysis algorithm to construct a musculoskeletal simulation model, inputs the trainee's joint range of motion, muscle strength and bone density parameters, and outputs appropriate movement range, force intensity and training duration thresholds.

[0012] Specifically, the rehabilitation action database of the simulation guidance module includes three core scenarios: stroke rehabilitation, post-joint replacement surgery rehabilitation, and osteoporosis rehabilitation. Each action supports 3-5 levels of difficulty adjustment, and the action breakdown teaching is demonstrated step by step according to the joint movement sequence.

[0013] Specifically, the force feedback device of the real-time feedback module adopts a flexible force feedback glove with a built-in micro linear motor. When the action deviation exceeds 5° or the force intensity deviation exceeds 15%, the error is corrected synchronously through motor vibration and voice prompts.

[0014] Specifically, the risk warning module has three levels of warnings: Level 1 warning (parameters approaching 80% of the threshold), Level 2 warning (parameters reaching 90% of the threshold), and Level 3 warning (parameters exceeding the limit), which correspond to green prompts, yellow pop-ups, and red pause commands, respectively.

[0015] Specifically, the data storage and analysis module includes a local storage unit and a cloud database. The cloud database uses MySQL distributed storage and supports interface with medical institution systems to synchronize rehabilitation assessment data.

[0016] Specifically, the interactive terminal includes a touch screen and a voice interaction unit, which allows trainees to control the playback, pause and difficulty adjustment of simulated actions through voice commands. The touch screen adopts an anti-glare design and the font size is adjustable.

[0017] Specifically, it also includes a family member-side linkage module, which is connected to the data storage and analysis module to push training progress, early warning records and rehabilitation reports to family members through the APP.

[0018] Specifically, the personalized modeling module also supports dynamic updates, optimizing the musculoskeletal simulation model and adjusting the adaptive motion parameters every 7-14 days based on the latest collected training data and physiological parameters.

[0019] The beneficial effects of this invention are: Highly personalized and adaptable: Based on the physiological parameters of the elderly, a dedicated musculoskeletal simulation model is constructed. Combined with a dynamic update mechanism, it ensures that rehabilitation movements are accurately matched with the trainee's physical condition. This solves the problem of existing systems having strong universality but poor adaptability, and avoids the risk of ineffective training or injury due to incompatible movements. Precise motion capture: A multimodal acquisition scheme using a visual sensor, an inertial measurement unit, and a pressure sensor is adopted to overcome the shortcomings of a single sensor, enabling high-precision acquisition of data such as joint angles, muscle force, and movement speed. This provides reliable data support for simulation guidance and feedback error correction, and improves the standardization of training. Real-time feedback is highly efficient: Through multimodal real-time feedback of force feedback devices, voice, and vision, movement deviations can be corrected in a timely manner. Compared with the lagging feedback of existing systems, this significantly improves training effectiveness and shortens the rehabilitation cycle. Comprehensive safety protection: A three-level early warning mechanism based on personalized thresholds monitors body load parameters in real time, intervenes in excessive movements in a timely manner, effectively reduces the risk of secondary injury, and solves the safety pain points of rehabilitation training for the elderly. Convenient data management and collaboration: Through a local plus cloud storage solution, long-term tracking and analysis of rehabilitation data can be achieved, supporting collaboration among medical institutions, trainees, and families to form a closed-loop rehabilitation management system and improve the controllability of the rehabilitation process; User-friendly and easy to use: The design is adapted to the vision, hearing and operation ability of the elderly (large font, anti-glare, voice interaction), which reduces the threshold of use and improves the acceptance and practicality of the system. In summary, this invention effectively solves many shortcomings of existing elderly rehabilitation training, achieving a combination of standardization and personalization, precise guidance and safety protection. It is applicable to home and institutional rehabilitation scenarios and has significant practical value and promotion prospects. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] Figure 1 The system's overall structure diagram shows the connections between various modules, including the data acquisition module, personalized modeling module, and simulation guidance module. Figure 2 A schematic diagram of the personalized modeling process, showing the complete process from physiological parameter input and initial motion collection to model generation and dynamic updates; Figure 3 The risk warning logic diagram shows the logical relationship between parameter monitoring, threshold comparison, graded warning and pause command triggering. Detailed Implementation

[0022] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0023] like Figures 1-3 As shown, the elderly rehabilitation training movement simulation guidance system of the present invention achieves personalized, precise and safe rehabilitation training guidance through the collaborative work of multiple modules. It includes a data acquisition module, a personalized modeling module, a simulation guidance module, a real-time feedback module, a risk warning module, a data storage and analysis module, and an interactive terminal and family member terminal linkage module. Each module is connected through wired or wireless communication to form a complete technical closed loop.

[0024] Data Acquisition Module: As the core of the system's data input, this module integrates multimodal sensors to comprehensively acquire motion data, physiological parameters, and environmental data. Specifically, the vision sensor uses a Kinect V2 depth camera to acquire the three-dimensional coordinates of the trainee's skeletal joints, with a sampling frequency of 30fps and an angle measurement accuracy of ±0.5°. The inertial measurement unit (MPU9250 chip) is worn on key joints such as the shoulder, elbow, wrist, hip, knee, and ankle to collect joint angular velocity and acceleration data, compensating for blind spots caused by the vision sensor. The pressure sensor (FSR402) is placed in the training mat or glove to collect force distribution data from the soles of the feet or hands. The physiological monitoring sensor integrates a heart rate sensor (MAX30102), a blood oxygen sensor, and a surface electromyography (EMG) sensor to collect heart rate, blood oxygen saturation, and electromyographic signals in real time during training, reflecting the trainee's physical load status.

[0025] Personalized Modeling Module: Based on the trainee's physiological parameters (height, weight, joint range of motion, muscle strength, bone density, and medical history) and initial movement data acquired by the data acquisition module, a customized musculoskeletal simulation model is constructed using finite element analysis algorithms. This module first collects the trainee's basic physiological information through an interactive terminal, then combines this with the initial movement data (such as maximum joint range of motion and peak muscle force) collected by the data acquisition module to establish a three-dimensional simulation model including bones, muscles, and ligaments. The model presets the trainee's physiological tolerance thresholds (such as maximum joint range of motion and safe muscle force intensity), providing personalized guidance for subsequent simulations and risk warnings. Simultaneously, the module supports dynamic updates, optimizing the model every 7-14 days based on the latest training data and physiological parameters to ensure continuous adaptation to the trainee's rehabilitation progress.

[0026] The simulation guidance module includes a built-in database of rehabilitation movements covering core scenarios such as stroke rehabilitation, post-joint replacement surgery rehabilitation, and osteoporosis rehabilitation. Each movement has standardized parameters (such as range of motion, speed, and point of force application) calibrated by rehabilitation medicine experts. Based on personalized musculoskeletal simulation models, the module selects suitable movements from the database and demonstrates them in real-time on an interactive terminal using a 3D virtual human. The demonstration supports movement breakdown instruction, showing each movement step-by-step according to the joint movement sequence (e.g., upper limb lifting is divided into three steps: shoulder abduction, elbow extension, and wrist relaxation). Each step is accompanied by voice explanation (e.g., slowly raise your arm, feel the shoulder joint exertion, maintain a speed of 5 centimeters per second). Simultaneously, the module supports 3-5 levels of difficulty adjustment, gradually increasing the intensity and complexity of the movements according to the trainee's rehabilitation progress.

[0027] Real-time feedback module: This module compares the trainee's actual movement data acquired by the data acquisition module with the standardized movement data from the simulation guidance module in real time, calculating deviations (including joint angle deviations, force intensity deviations, and movement speed deviations). When a joint angle deviation exceeds 5°, a force intensity deviation exceeds 15%, or a movement speed deviation exceeds 20%, the module corrects the error through multi-dimensional feedback: the miniature linear motor built into the force feedback glove vibrates at the corresponding joint position (e.g., the elbow motor vibrates when there is an elbow joint angle deviation); voice prompts clearly state the type of deviation (e.g., if the elbow joint bending angle is insufficient, please bend it another 10°); and a visual pop-up appears on the interactive terminal screen, indicating the location of the deviation and the direction of correction. Through multi-modal feedback, the module ensures that the trainee adjusts their movements in a timely manner, guaranteeing the standardization of training.

[0028] Risk warning module: Based on the physiological tolerance threshold set by the personalized modeling module, it monitors parameters such as joint load, muscle tension, and heart rate in real time during training. The module has a preset three-level warning mechanism: When a Level 1 warning is issued (parameters reach 80% of the threshold), the interactive terminal displays a green prompt box and provides a voice prompt to pay attention to the range of motion and keep breathing steady. When a Level 2 warning is issued (parameters reach 90% of the threshold), a yellow pop-up window will be displayed, and a voice prompt will say, "You are approaching your body's tolerance limit and you may slow down your speed." When a Level 3 warning is triggered (parameters exceed the threshold), the system automatically issues a red pause command, the 3D virtual human stops demonstrating, the force feedback device continues to vibrate, and a voice prompt says "Movement has exceeded limits, please pay attention to safety, pause training," and records the warning data to the data storage module.

[0029] The data storage and analysis module includes a local storage unit (SD card, storage capacity no less than 64GB) and a cloud database (MySQL distributed storage) for storing the trainee's basic physiological parameters, movement data for each training session, deviation records, warning records, and rehabilitation progress data. The module processes historical data using data analysis algorithms (such as linear regression analysis) to generate rehabilitation assessment reports, including indicators such as training duration statistics, changes in movement accuracy, improvement in joint range of motion, and muscle strength recovery, which are displayed in chart form on the interactive terminal. Simultaneously, the cloud database supports integration with the rehabilitation management system of medical institutions, allowing doctors to view training data in real time and adjust rehabilitation plans; the family-side APP uses this module to access training progress and warning records, enabling multi-party collaborative management of the rehabilitation process.

[0030] Interactive Terminal: Features a 10-inch anti-glare touchscreen display with three adjustable font sizes (standard, enlarged, and extra-large) to suit the visual characteristics of the elderly. It integrates a voice interaction unit, allowing users to control the system via voice commands (such as playing shoulder rehabilitation exercises, pausing training, or adjusting the difficulty to level 2), reducing operational complexity. The terminal interface is simple and clear, with main functional areas (simulation demonstration, feedback prompts, and rehabilitation reports) displayed using large icons for easy identification and operation by the elderly.

[0031] Family member interaction module (optional): Through the APP and data storage and analysis module, family members can view the trainee's training time, movement accuracy rate, rehabilitation progress and early warning records in real time; when the system triggers a level 2 or higher early warning, the APP will push an early warning notification to remind family members to pay attention to the trainee's safety; at the same time, it supports voice calls between family members and trainees to remotely encourage and guide training.

[0032] Example Example 1: Basic Elderly Rehabilitation Training Movement Simulation Guidance System 1. System Composition The basic system in this embodiment includes a data acquisition module, a personalized modeling module, a simulation guidance module, a real-time feedback module, a risk warning module, a data storage and analysis module, and an interactive terminal. The specific configurations of each module are as follows: Data acquisition module: It adopts Kinect V2 depth camera (visual sensor), 6 MPU9250 inertial measurement units (worn on the shoulder, elbow, wrist, hip, knee and ankle joints respectively), 4 FSR402 pressure sensors (laid on the corresponding area of ​​the foot on the training pad), MAX30102 heart rate and blood oxygen sensor and surface electromyography sensor (attached to the biceps and quadriceps muscles); the sensors communicate with the main control unit (using NVIDIA Jetson Nano development board) via USB 3.0 and Bluetooth 5.0, and the data transmission latency is ≤100ms.

[0033] Personalized modeling module: Based on the NVIDIA Jetson Nano development board, the finite element analysis algorithm (using a simplified version of ANSYS software) is run. The basic physiological parameters of the trainee are collected through the interactive terminal (height 170cm, weight 65kg, left shoulder joint range of motion 120°, right knee joint range of motion 90°, muscle strength level 3, bone density T-score -1.5, and history of stroke (left hemiplegia)). Combined with the initial movement data (such as the maximum lifting angle of the left upper limb 80° and the maximum extension angle of the right lower limb 85°), a personalized musculoskeletal simulation model is constructed, and personalized thresholds are set (left shoulder joint load threshold 15N, right knee joint load threshold 20N, and heart rate warning threshold 100 beats / minute).

[0034] Simulation guidance module: Built-in stroke rehabilitation movement database, containing 12 core movements for left-sided hemiplegia rehabilitation (such as left upper limb lifting, left finger grasping, left lower limb flexion and extension, etc.); Based on personalized models, suitable movements are selected (excluding movements with a left shoulder joint range of motion exceeding 80°). The 3D virtual human driven by the Unity 3D engine demonstrates on an interactive terminal (10-inch anti-glare touch screen). The movements are broken down into 3 steps, and the voice narration uses a slowed speech rate (120 words per minute) and a louder volume (maximum 80dB). The difficulty is set to level 1 (basic level) by default, and the movement range is 70% of the standard range.

[0035] Real-time feedback module: The force feedback glove adopts a flexible design and has 3 built-in miniature linear motors (corresponding to the shoulder, elbow, and wrist joints). The vibration intensity is divided into 3 levels (weak, medium, and strong). When the trainee raises the left upper limb, Kinect V2 collects the actual angle of the left shoulder joint as 60°, while the standard angle of the simulated action is 80° (deviation of 20°). The surface electromyography sensor collects the muscle force intensity deviation of 18%. At this time, the shoulder motor of the force feedback glove vibrates at medium intensity, and the voice prompts "The left shoulder joint lifting angle is insufficient. Please raise it another 20° and increase the force slightly." The screen pop-up window indicates the position of the left shoulder joint and the deviation value.

[0036] Risk warning module: During training, when the load on the left shoulder joint reaches 14N (threshold 80%), a level one warning is triggered, a green prompt box is displayed on the screen, and a voice prompt says "The load on the left shoulder joint is approaching the threshold, keep the movement stable"; when the load reaches 14.5N (threshold 90%), a level two warning is triggered, a yellow pop-up window appears, and a voice prompt says "The left shoulder joint is approaching its tolerance limit, you can slow down the lifting speed"; if the load exceeds 15N, the system automatically issues a pause command, the virtual human stops the demonstration, the glove vibrates continuously, and a voice prompt says "The load on the left shoulder joint exceeds the limit, stop training".

[0037] Data storage and analysis module: Local storage uses a 64GB SD card, and the cloud database is a MySQL distributed storage (deployed on Alibaba Cloud server), which stores the exercise data (joint angles, force intensity), deviation records, and warning records of each training session; data analysis uses a linear regression algorithm to generate a rehabilitation report every week, showing indicators such as the improvement in the range of motion of the left upper limb joints and changes in the standard rate of movement.

[0038] Interactive terminal: 10-inch anti-glare touch screen, supports extra-large font display, voice interaction unit recognition rate ≥95%, supports commands such as "play action", "pause", "adjust difficulty" and the interface is divided into three large icon function areas: simulation demonstration area, feedback prompt area and rehabilitation report area.

[0039] 2. Work Process Initial setup: The trainee inputs basic physiological parameters (height, weight, medical history, etc.) through the interactive terminal, wears the inertial measurement unit and electromyography sensor, stands on the training mat (above the pressure sensor), and the Kinect V2 is activated to collect initial motion data (such as limbs hanging naturally, maximum lifting, etc.).

[0040] Personalized Modeling: The personalized modeling module constructs a musculoskeletal simulation model based on input parameters and initial motion data through finite element analysis algorithms. It sets personalized thresholds such as joint load and heart rate. After the model is generated, a 3D preview image is displayed on the interactive terminal, and the trainee can confirm whether it is suitable.

[0041] Simulation guidance initiated: The trainee selects the "Rehabilitation of Left Hemiplegia after Stroke" scenario through the interactive terminal. The simulation guidance module selects suitable movements (such as raising the left upper limb) from the database. The 3D virtual human begins to demonstrate, first showing the complete movement (5 seconds in length), then demonstrating it step by step (3 seconds per step), with simultaneous voice explanation of the key points of the movement.

[0042] Real-time training and feedback: Trainees follow the virtual human to perform movements. The data acquisition module collects data such as joint angles, force intensity, and heart rate in real time. The real-time feedback module compares the actual data with the standard data and calculates the deviation value. When the deviation exceeds the set threshold, it corrects the error in real time through force feedback glove vibration, voice prompts, and screen pop-ups.

[0043] Risk monitoring and early warning: The risk early warning module monitors parameters such as joint load and heart rate in real time, and triggers corresponding level early warnings based on the ratio of parameters to thresholds. If the parameters exceed the limits, training is automatically paused and the early warning data is recorded.

[0044] Data storage and analysis: After the training is completed, the data storage module saves the complete data of this training, and the data analysis module processes the data and updates the rehabilitation report. Trainees can view the report on the interactive terminal to understand their rehabilitation progress.

[0045] 3. Application Effects The basic system in this embodiment was applied to 50 elderly patients with left hemiplegia due to stroke (aged 65-80 years). After 8 weeks of continuous use, the statistical results showed that the accuracy of movement improved from 45% to 82%, the range of motion of the left upper limb joints improved by an average of 35%, no secondary injuries such as joint damage or muscle strain occurred during the training process, 90% of the trainees reported that the interactive operation was convenient, 88% of the trainees believed that the real-time feedback was effective, and the rehabilitation cycle was shortened by 15-20% compared with traditional manual guidance.

[0046] Example 2: Cloud-based interactive simulation guidance system for elderly rehabilitation training movements 1. System Composition This embodiment, based on embodiment 1, adds a family member-side linkage module and optimizes the cloud-based linkage function of the data storage and analysis module. The specific configuration is as follows: The configurations of the data acquisition module, personalized modeling module, simulation guidance module, real-time feedback module, risk warning module, and interactive terminal are the same as in Example 1, with the addition of a family member linkage module (an APP developed based on Android / iOS systems).

[0047] Data storage and analysis module: The cloud database has been upgraded to an Alibaba Cloud ECS server with Redis caching, supporting 100 data entries per second. It is connected to the rehabilitation management systems of three rehabilitation hospitals (using the HL7 medical data exchange standard). The data analysis module has added a rehabilitation effect prediction function. Based on the LSTM deep learning algorithm, it can predict the rehabilitation progress in the next 4 weeks by inputting the trainee's historical training data and physiological parameters (such as the expected improvement in joint range of motion and the expected achievement of the standard value of movement).

[0048] Family-side integration module: App functions include training progress viewing (daily / weekly / monthly training duration, movement accuracy rate, rehabilitation report), early warning notifications (real-time push notifications for level 2 and above early warnings), voice calls (supports voice connection with the trainer's interactive terminal), and training reminders (daily training time can be set, and the app will push reminders); family members can bind multiple trainer accounts and monitor the rehabilitation status of multiple elderly people at the same time.

[0049] 2. Work Process Initialization and modeling: Same as in Example 1, with the addition of a binding step between the trainee and the family member's app (a binding QR code is generated through the interactive terminal, and the family member scans the code to bind).

[0050] Simulation training and feedback: Similar to Example 1, the training data collected by the data acquisition module is synchronized to the cloud database in real time. Doctors in the rehabilitation hospital can view the trainee's movement data, deviation records and early warning status in real time through the rehabilitation management system. They can remotely adjust the rehabilitation movement plan (such as adding specific training movements or modifying the difficulty of movements). The adjustment instructions are synchronized to the simulation guidance module through the cloud.

[0051] Family-side integration: During training, if the system triggers a level-two warning (such as the load on the left shoulder joint reaching 14.5N), the family-side APP will immediately push a warning notification (including the warning type, the time of occurrence, and suggested measures); the family can view the real-time training video through the APP (captured by the interactive terminal camera and enabled after authorization by the trainee), and remind the trainee to pay attention to safety through the voice call function; after the training is completed, the rehabilitation report is automatically synchronized to the family-side, and the family can view the trainee's rehabilitation progress and the doctor's remote assessment suggestions.

[0052] Rehabilitation effect prediction: The data analysis module generates a rehabilitation effect prediction report based on historical data every week and pushes it to the interactive terminal, family terminal and doctor terminal. Doctors can adjust the rehabilitation plan according to the prediction results. For example, if the prediction is that the improvement of joint mobility is slow, the training intensity can be increased or the training movements can be changed appropriately.

[0053] 3. Application Effects The cloud-based collaborative system in this embodiment was applied to 30 elderly stroke patients undergoing rehabilitation, their families, and 5 rehabilitation doctors. After 8 weeks of continuous use, the results showed that: the doctors' remote guidance response time was ≤5 minutes; the family members' awareness of the trainees' rehabilitation progress increased from 30% in the traditional model to 95%; the trainees' training compliance (the proportion of training completed on time) increased from 75% to 92%; the accuracy rate of rehabilitation effect prediction reached 85%; the rehabilitation progress of 28 out of 30 trainees met or exceeded the predicted value; and the doctors' work efficiency increased by 40% (no on-site guidance was required, and the plan could be adjusted remotely).

[0054] Example 3: AI Adaptive Elderly Rehabilitation Training Movement Simulation Guidance System 1. System Composition This embodiment optimizes the AI ​​adaptive functionality of the personalized modeling module and the simulation guidance module based on embodiment 2. The specific configuration is as follows: Data acquisition module: A new infrared thermal imaging sensor (used to monitor muscle temperature and reflect the degree of muscle fatigue) has been added. The configuration of other sensors is the same as in Example 2.

[0055] Personalized modeling module: Integrates AI algorithms (CNN plus RNN hybrid model), which not only builds models based on physiological parameters and initial action data, but also optimizes the model's adaptability by continuously learning the trainee's action habits and learning ability (such as the number of times the action is corrected and the time to master a new action); for example, when the trainee learns a new action and makes many corrections, the model automatically extends the decomposition demonstration time of the action and reduces the speed requirement of the action.

[0056] Simulation guidance module: Based on reinforcement learning algorithms, this module enables adaptive adjustments to the training program. It analyzes the trainee's movement accuracy, muscle fatigue (defined as a muscle temperature exceeding 38°C by an infrared thermal imaging sensor), and heart rate changes in real time, dynamically adjusting training movements, difficulty, and duration. For example, if the movement accuracy is ≥90% for three consecutive repetitions, the difficulty is automatically increased by one level; if the muscle temperature exceeds 38°C, the module switches to low-intensity relaxation exercises; and if the heart rate remains above a warning threshold, the module automatically pauses training and recommends a rest period.

[0057] Real-time feedback module: A new AI voice assistant has been added, which supports natural language interaction. Trainees can ask questions by voice (such as "Is my movement standard?" "Can I rest now?"). The AI ​​assistant provides accurate answers based on real-time data. At the same time, the feedback content is adaptively adjusted according to the trainee's learning progress. In the early stage, it focuses on basic error correction, and in the later stage, it focuses on optimizing details (such as "The movement angle is up to standard, and the stability of force can be appropriately increased").

[0058] Risk warning module: Integrates AI risk prediction function. By analyzing the trainee's historical warning records and movement deviation patterns, it predicts potential risks (such as repeatedly approaching the load threshold in the left upper limb lifting movement, predicting that the next training may exceed the limit) and triggers warnings in advance (such as prompting "The left shoulder joint is prone to exceeding the limit, please pay attention to controlling the range of motion" before training).

[0059] 2. Work Process AI Modeling and Adaptive Learning: After the trainee is initialized, the personalized modeling module extracts the trainee's action features through the CNN algorithm, and the RNN algorithm learns its learning habits to build an AI adaptive model. During training, the model continuously receives real-time data from the data acquisition module and dynamically optimizes the model parameters. For example, if the trainee completes the left upper limb lifting movement ≥85% of the standard rate 5 times in a row, the model automatically adjusts the personalized threshold of the movement to 16N (original threshold 15N) to adapt to the improvement of muscle strength.

[0060] Adaptive simulation guidance: After the trainee starts training, the simulation guidance module recommends an initial training plan based on the AI ​​model (3 core movements, level 1 difficulty, 20 minutes). During training, the infrared thermal imaging sensor detects that the temperature of the left biceps brachii reaches 38.5℃ (fatigue state). The module automatically pauses the current movement and switches to a shoulder relaxation movement (low intensity). At the same time, the voice prompt says "Muscles are fatigued, perform 2 minutes of relaxation training". After relaxation, the muscle temperature drops to 37℃. The module resumes the original training movement and reduces the movement speed by 10%. When the trainee completes the movement with a standard rate of ≥90% for 3 consecutive times, the module automatically increases the difficulty to level 2 and increases the movement range by 10%.

[0061] AI Risk Prediction and Feedback: The risk warning module analyzes the trainee's historical data and finds that the trainee is prone to approaching the load threshold when raising the left upper limb to 70°. Therefore, when the movement is about to reach 70° in this training session, a first-level warning is triggered in advance, with a voice prompt: "Approaching the load threshold of the left shoulder joint, slow down the lifting speed." The AI ​​assistant in the real-time feedback module receives the trainee's question, "Is my movement more standard than last time?" and answers based on data comparison, "Compared to the last training session, the deviation of the left shoulder joint angle has decreased from 15° to 8°, the movement is more standard, keep it up."

[0062] 3. Application Effects The AI ​​adaptive system in this embodiment was applied to 20 elderly rehabilitation trainees (including stroke patients and post-joint replacement surgery patients). After 8 weeks of continuous use, the results showed that: the trainees' average movement standardization rate increased to 88% (6% higher than in Example 2), the number of training interruptions due to muscle fatigue decreased by 70%, the risk prediction accuracy reached 82%, and early warnings avoided 12 potential over-limit risks; the trainees' satisfaction with the system reached 95%, believing that the adaptively adjusted training program was more suitable for their own condition, and the time to learn new movements was shortened by an average of 25%; the rehabilitation period for post-joint replacement surgery patients was shortened by 10-15% compared to Example 2, and the speed of muscle strength recovery increased by 20%.

[0063] Comparison Example To verify the inventiveness and superiority of the present invention, the following three control examples were set up for comparative testing with Example 1 (basic type) of the present invention. The test subjects were 60 elderly people with left hemiplegia due to stroke (aged 65-80 years), who were randomly divided into 4 groups (15 people in each group). They underwent continuous training for 8 weeks. The test indicators included the standardization rate of movement, rehabilitation effect (the extent of improvement in joint range of motion), safety risk (the incidence of secondary injury), and training compliance.

[0064] Comparison with Example 1: A general-purpose rehabilitation simulation system based on single visual capture System configuration: Only Kinect V2 visual sensor is used to collect motion data, without inertial measurement unit and pressure sensor; no personalized modeling module, using a uniform adult simulation motion model; only voice feedback is supported, without thermal feedback and visual pop-ups; no risk warning module, only displaying motion deviation statistics after training.

[0065] Test results: The average rate of correct movement was 52% (82% in Example 1); the average improvement in joint range of motion was 18% (35% in Example 1); the incidence of secondary injury was 13.3% (2 people suffered joint strains, 0 in Example 1); and the training compliance rate was 66.7% (10 people abandoned part of the training due to inappropriate movement or lack of feedback).

[0066] Conclusion: Using a single sensor results in insufficient motion capture accuracy, poor adaptability to general models, and a lack of feedback and early warning, leading to poor training results and high safety risks.

[0067] Comparison with Example 2: Rehabilitation Simulation System Without Personalized Modeling System configuration: The data acquisition module, real-time feedback module, risk warning module, and interactive terminal are the same as in Example 1; there is no personalized modeling module, and a unified elderly rehabilitation movement model is used (without considering individual physiological parameter differences); the movement parameters of the simulation guidance module are fixed and there are no dynamic updates.

[0068] Test results: The average rate of correct movement was 65% (82% in Example 1); the average increase in joint range of motion was 25% (35% in Example 1); the incidence of secondary injury was 6.7% (1 person experienced discomfort due to the range of motion exceeding the joint range of motion); the training compliance rate was 73.3% (11 people showed resistance due to the mismatch between the difficulty of the movements and the training).

[0069] Conclusion: Lack of personalized modeling leads to insufficient movement adaptability, which fails to match individual physiological differences and affects rehabilitation effectiveness and safety.

[0070] Comparison with Example 3: A rehabilitation simulation system without real-time feedback and risk warning System configuration: The data acquisition module, personalized modeling module, simulation guidance module, and data storage module are the same as in Example 1; there is no real-time feedback module, and deviation reports are only generated after training; there is no risk warning module, and only a fixed training duration is set, without monitoring body load parameters.

[0071] Test results: The average rate of correct movement was 58% (82% in Example 1); the average improvement in joint range of motion was 22% (35% in Example 1); the incidence of secondary injury was 20% (3 people suffered muscle strain); and the training compliance rate was 60% (9 people gave up training because they could not correct their movements in time, making the training ineffective).

[0072] Conclusion: The lack of real-time feedback leads to the inability to correct movement deviations in a timely manner, the lack of risk warnings increases safety hazards, and significantly reduces rehabilitation effectiveness and training compliance.

[0073] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of protection claimed by the present invention. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A simulation guidance system for rehabilitation training movements in the elderly, characterized in that, include: The data acquisition module is used to collect the trainee's motion data, physiological parameters, and training environment data; The personalized modeling module is connected to the data acquisition module and constructs a musculoskeletal simulation model exclusive to the trainee based on the motion data and physiological parameters, including height, weight, joint range of motion, muscle strength, bone density and past medical history. The simulation guidance module communicates with the personalized modeling module, has a built-in rehabilitation movement database, generates standardized simulation movements adapted to the trainee based on the musculoskeletal simulation model, and performs real-time demonstrations through a 3D virtual human and supports movement decomposition teaching. The real-time feedback module is communicatively connected to the data acquisition module and the simulation guidance module, respectively. It compares the acquired motion data with the simulated motion data and corrects motion deviations through force feedback devices, voice prompts, and visual pop-ups. The risk warning module is connected to the data acquisition module and the personalized modeling module respectively. It presets joint load threshold, muscle tension threshold and movement speed threshold, and monitors whether the parameters exceed the limits during training in real time, triggering graded warnings. The data storage and analysis module communicates with each of the modules, stores training data, physiological parameters and early warning records, and generates rehabilitation progress reports through data analysis. The interactive terminal is used to display simulated movements, feedback information, early warning prompts, and rehabilitation reports, and supports input of operation commands by trainees.

2. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The data acquisition module includes a visual sensor, an inertial measurement unit, a pressure sensor, and a physiological monitoring sensor. The visual sensor uses a Kinect V2 depth camera, the inertial measurement unit has an MPU9250 chip built in, the pressure sensor uses an FSR402 flexible pressure sensor, and the physiological monitoring sensor is used to collect heart rate, blood oxygen saturation, and electromyography signals.

3. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The personalized modeling module uses the finite element analysis algorithm to construct a musculoskeletal simulation model. It inputs the trainee's joint range of motion, muscle strength, and bone density parameters, and outputs appropriate movement range, force intensity, and training duration thresholds.

4. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The simulation guidance module's rehabilitation movement database includes three core scenarios: stroke rehabilitation, post-joint replacement surgery rehabilitation, and osteoporosis rehabilitation. Each movement supports 3-5 levels of difficulty adjustment, and the movement breakdown teaching is demonstrated step by step according to the joint movement sequence.

5. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The force feedback device of the real-time feedback module adopts a flexible force feedback glove with a built-in micro linear motor. When the action deviation exceeds 5° or the force intensity deviation exceeds 15%, the error is corrected synchronously through motor vibration and voice prompts.

6. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The risk warning module has a tiered warning system, including a Level 1 warning when the parameter is close to 80% of the threshold, a Level 2 warning when the parameter reaches 90% of the threshold, and a Level 3 warning when the parameter exceeds the limit, which correspond to a green prompt, a yellow pop-up window, and a red pause command, respectively.

7. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The data storage and analysis module includes a local storage unit and a cloud database. The cloud database uses MySQL distributed storage and supports integration with medical institution systems to synchronize rehabilitation assessment data.

8. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The interactive terminal includes a touch screen and a voice interaction unit, which allows trainees to control the playback, pause and difficulty adjustment of simulated actions through voice commands. The touch screen adopts an anti-glare design and the font size is adjustable.

9. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: It also includes a family member-side linkage module, which communicates with the data storage and analysis module to push training progress, early warning records and rehabilitation reports to family members via the APP.

10. The elderly rehabilitation training movement simulation guidance system according to claim 1, characterized in that: The personalized modeling module also supports dynamic updates, optimizing the musculoskeletal simulation model and adjusting the adaptive motion parameters every 7-14 days based on the latest collected training data and physiological parameters.

Images