Real-time rehabilitation information feedback system based on mobile terminal

Abstract

The application is suitable for the cross field of rehabilitation medical treatment and mobile internet, and provides a real-time rehabilitation information feedback system based on a mobile terminal, which comprises a user mobile terminal, an integrated data acquisition module, a data processing module, a display module and a communication module; the data acquisition module acquires physiological index data, motion data and rehabilitation training data; the data processing module performs cleaning, conversion and normalization preprocessing on the collected data, and the normalization formula is: y i = (x i - x min) / (x max- x min), wherein x is a data set, and x is a single data; the display module outputs rehabilitation feedback in the form of augmented reality, voice and vibration multi-modal; in the application, real-time closed loop: patient end data is transmitted to a server in millisecond level after edge preprocessing (delay < 50 ms under 5G network); medical staff end real-time analysis, and a decision module generates feedback in second level (for example, when training action deviation occurs, AR immediately superimposes a correction prompt), so that the intervention opportunity is compressed from a "recheck cycle" to "action execution moment".

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of rehabilitation medicine and mobile Internet, and in particular relates to a real-time rehabilitation information feedback system based on mobile terminals. Background Technology

[0002] In the field of rehabilitation medicine, precision rehabilitation relies on a continuous, multi-dimensional, and real-time information loop, but traditional models and existing technologies have core flaws:

[0003] I. Limitations of Traditional Rehabilitation Models:

[0004] Information lag: Patients rely on regular follow-up visits (e.g., once a week). During the rehabilitation process, key information such as physiological fluctuations (e.g., abnormal heart rate) and movement deviations (e.g., incorrect joint angles) cannot be transmitted in a timely manner. For example, in gait training for stroke patients, monitoring only the heart rate cannot detect the movement abnormality of "shortened step length on the affected side". Doctors can only adjust the plan based on static data, resulting in delayed intervention (e.g., muscle compensatory damage is only discovered after it has already occurred).

[0005] The data dimensions are limited: follow-up visits mainly rely on subjective descriptions ("feeling pain") and basic physiological indicators (blood pressure, heart rate), lacking quantitative collection of kinematic data (gait symmetry, joint range of motion) and training execution data (movement standard, force sequence), resulting in insufficient targeting of the treatment plan;

[0006] II. Deficiencies of existing mobile systems:

[0007] Insufficient real-time performance: Data relies on scheduled uploads (such as daily evening synchronization), and no edge preprocessing is performed on the collected data (such as removing sensor drift data), resulting in the server receiving "delayed and noisy" data, which cannot support real-time decision-making (such as training action deviations needing to be corrected immediately, rather than being reminded the next day).

[0008] Weak analytical capabilities: They mostly rely on simple statistics (such as calculating the number of training sessions), lack data mining (such as linking training movements with complications) and predictive modeling (such as predicting the risk of muscle fatigue within 3 days), and the feedback is limited to "reminding training" without forming personalized intervention plans;

[0009] The interaction methods are limited: mainly text prompts, without incorporating interactive technologies such as AR (motion correction), voice (real-time guidance), and haptic (vibration feedback), making it difficult for patients to obtain immersive and immediate guidance when performing rehabilitation movements.

[0010] In summary, traditional models and existing systems cannot meet the needs of precise rehabilitation due to "incomplete data collection, non-real-time transmission and processing, unintelligent analysis and feedback, and limited interaction methods." Therefore, building a real-time rehabilitation system covering the entire process of "collection-transmission-analysis-decision-feedback-storage" has become an urgent need for the industry.

[0011] Therefore, a real-time rehabilitation information feedback system based on mobile devices is needed to solve the above problems. Summary of the Invention

[0012] The purpose of this invention is to provide a real-time rehabilitation information feedback system based on a mobile terminal to solve the problems mentioned in the background art.

[0013] To achieve the above objectives, the present invention provides the following technical solution:

[0014] A mobile-based real-time rehabilitation information feedback system includes:

[0015] The user's mobile terminal integrates a data acquisition module, a data processing module, a display module, and a communication module, supporting multi-sensor access and edge preprocessing; the data acquisition module collects physiological index data, exercise data, and rehabilitation training data.

[0016] The data processing module performs cleaning, transformation, and normalization preprocessing on the collected data (normalization formula: ,in For a data set, (for a single data entry)

[0017] The display module outputs rehabilitation feedback in augmented reality (AR), voice, and vibration multimodal formats;

[0018] The communication module uses the SSL / TLS encryption protocol to implement data transmission;

[0019] Healthcare server: Deploys data receiving, data analysis, decision-making, and communication modules to enable real-time data processing and intelligent decision-making;

[0020] The data receiving module receives user-end data and verifies its integrity.

[0021] The data analysis module performs statistical analysis (mean, standard deviation, pass rate formula: ...) Data mining (Apriori association rules, isolated forest anomaly detection) and prediction models (LSTM neural network formula: Linear regression formula: Processing data;

[0022] The decision-making module is based on medical rules and patient profiles, and uses a multi-objective optimization algorithm (...). ,in For the goal of rehabilitation, Generate personalized solutions (based on actual values);

[0023] The communication module enables interaction with the user terminal and the cloud database;

[0024] Cloud-based database: Stores rehabilitation data, plans, solutions, and suggestions, and uses blockchain technology to achieve data traceability and access management, ensuring data security and traceability;

[0025] The system achieves real-time interaction and intelligent control of rehabilitation information through a closed-loop process of "data acquisition - edge preprocessing - encrypted transmission - intelligent analysis - decision feedback - blockchain storage".

[0026] This system enables real-time closed-loop interaction of rehabilitation information, replacing the lagging "regular follow-up visits" model; it constructs a multi-dimensional data collection system (physiological, motor, and training) to fill data dimension gaps; it integrates edge computing and cloud-based intelligent analysis to overcome bottlenecks in real-time performance and analytical capabilities; and it innovates multi-modal immersive interaction to enhance the intuitiveness and effectiveness of rehabilitation guidance.

[0027] The system is compatible with various terminals, including smartphones, tablets, and smart wearable devices (wristbands, glasses, insoles); the medical staff server supports multi-terminal access from PC workstations and mobile medical staff terminals (Pads, mobile phones).

[0028] Data security: The cloud database uses blockchain technology to achieve full-process traceability and tamper-proofing of rehabilitation data, combined with role-based access control (tiered authorization for doctors, patients, and family members) to meet medical data compliance requirements;

[0029] In a further technical solution, the data acquisition module includes:

[0030] Physiological data acquisition submodule: compatible with heart rate belt (sampling rate ≥50Hz), blood pressure monitor, electromyography patch (amplitude range 0-5mV), and body temperature sensor to acquire heart rate, blood pressure, electromyography signals, and body temperature data;

[0031] Motion data acquisition submodule: Adapts to IMU inertial module, GPS module, pressure sensor, joint angle sensor (accuracy ±0.5°), extracts gait parameters (stride length, cadence, center of gravity trajectory) and joint range of motion data;

[0032] Training data acquisition submodule: Integrates a camera pose estimation algorithm based on OpenPose, and rehabilitation equipment sensor interfaces (RS485, Bluetooth), and quantifies the training movement completion formula. .

[0033] In a further technical solution, the normalization process of the data processing module satisfies the following formula: ,in, Data sets The minimum and maximum values ​​are normalized and mapped to the interval [0,1], which serves as the input for subsequent analysis models.

[0034] In a further technical solution, the communication module uses SSL / TLS encryption protocol to ensure data transmission security and supports multiple network adaptations such as 5G, Wi-Fi, and Bluetooth;

[0035] Perform edge preprocessing on the user's mobile device: filter sensor drift data (such as invalid frames with pressure <5kPa), extract key features of gait abnormal frames before uploading, reducing transmission load by ≥60%.

[0036] In a further technical solution, the data analysis module employs the following technologies:

[0037] Predictive model: Predicts changes in physiological indicators over the next 3 days using an LSTM neural network (input is angle and electromyography data from the past 3 days).

[0038] Predicting recovery progress using a linear regression model (formula: );

[0039] Data mining: The Apriori algorithm was used to mine association rules between rehabilitation actions and effects (minimum support ≥ 10%, minimum confidence ≥ 80%).

[0040] The isolated forest algorithm was used to detect abnormal physiological data (such as heart rate variability HRV < 30ms).

[0041] Statistical analysis: Calculate the mean, standard deviation, and pass rate of the data (formula: ).

[0042] A further technical solution involves a built-in rehabilitation rule engine in the decision-making module. Based on medical guidelines and patient profiles (age, disease type, rehabilitation stage), it generates personalized rehabilitation plans through a multi-objective optimization algorithm. ;

[0043] in, For the first Rehabilitation goals (such as joint angles and exercise intensity). This is the patient's actual value;

[0044] The optimization objective is to minimize the sum of squared deviations between the target and the actual value, and the output includes decisions on training intensity adjustment, movement correction, and intervention timing.

[0045] A further technical solution is that the display module supports multimodal output:

[0046] Visual interaction: Dynamically display physiological / motor data trend graphs (such as heart rate-time curves and gait trajectory graphs), and overlay training movement correction prompts through AR technology;

[0047] Auditory interaction: Real-time broadcast of rehabilitation suggestions using text-to-speech (TTS) technology (such as "Knee flexion angle is less than 10°, please adjust your force").

[0048] Haptic interaction: Vibration feedback is triggered by the vibration module of the smart wearable device when the movement is incorrect, and stops after correction.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] This invention features a real-time closed-loop system: patient-side data is preprocessed at the edge and transmitted to the server in milliseconds (latency <50ms under 5G network); medical staff analyze the data in real time, and the decision-making module generates feedback in seconds (e.g., when training movements deviate, AR immediately overlays corrective prompts), compressing the intervention timing from the "follow-up visit cycle" to the "moment of action execution";

[0051] This invention collects data from multiple dimensions: covering physiological (heart rate, electromyography, body temperature), movement (gait parameters, joint angles, center of gravity trajectory), and training (movement completion, error type, load-fatigue curve), constructing a three-dimensional rehabilitation model of "physiology-movement-training". For example, by using electromyography signals and joint angle data, it can accurately determine the "degree of muscle compensation on the affected side" in stroke patients, providing a quantitative basis for adjusting the treatment plan.

[0052] This invention features edge + cloud collaboration: the mobile device first performs lightweight preprocessing (such as normalization and outlier removal) to reduce transmission load; the server processes massive amounts of data through a parallel computing framework, combining prediction models (LSTM, linear regression) and data mining algorithms (Apriori, Isolation Forest) to achieve an intelligent closed loop of "real-time monitoring → trend prediction → early intervention".

[0053] This invention, multimodal immersive interaction:

[0054] Visual: Dynamically display data trend charts (such as heart rate-training intensity curves), and overlay AR motion correction prompts (such as virtual arrows indicating joint trajectories).

[0055] Auditory: Real-time voice broadcast of rehabilitation suggestions (such as "Knee flexion angle is less than 10°, please adjust the force");

[0056] Tactile feedback: Vibration feedback from smart wearable devices (continuous vibration when movement is incorrect, stopping after correction), multimodal interaction improves the standardization of rehabilitation movements, indirectly improving rehabilitation efficiency;

[0057] This invention is compatible with multiple terminals: it supports mobile phones, tablets, and smart wearable devices (wristbands, glasses, insoles), covering hospital, home, and community scenarios;

[0058] Data security: Blockchain is used in the cloud to ensure that rehabilitation data is "traceable and tamper-proof" and meets the requirements for medical data compliance.

[0059] To more clearly illustrate the structural features and effects of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the system architecture of the present invention;

[0061] Figure 2 This is a schematic diagram of the user mobile terminal functional modules of the present invention;

[0062] Figure 3 This is a schematic diagram of the functional modules of the medical care server of the present invention;

[0063] Figure 4 This is a schematic diagram of the data processing flow of the present invention. Detailed Implementation

[0064] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0065] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0066] Example 1: Hospital Rehabilitation Department – ​​Precise Rehabilitation Monitoring After Knee Surgery

[0067] (1) Equipment configuration

[0068] User mobile device: Customized rehabilitation tablet (integrated camera, IMU inertial module), external electromyography patch (sampling rate 100Hz, amplitude 0-5mV), joint angle sensor (accuracy ±0.5°, range 0-120°), heart rate belt (sampling rate 100Hz).

[0069] Healthcare server: GPU servers are deployed on the hospital's intranet (to accelerate LSTM model computation);

[0070] Cloud-based database: Hospital private cloud, integrated with electronic medical record system.

[0071] (2) Process details

[0072] Data collection: Patients perform "straight leg raise" training:

[0073] Electromyography (EMG) patches collect quadriceps electromyography (EMG) signals, and joint angle sensors monitor knee flexion angles. ;

[0074] The camera uses the OpenPose algorithm to extract key points of the limbs and calculate the degree of motion completion. Simultaneously record the number of training repetitions N and the force exertion time for each movement. ;

[0075] Edge preprocessing (mobile):

[0076] Outlier removal: using Criteria for eliminating electromyographic signals exceeding [specific parameters] Noise points ( The mean, (Standard deviation)

[0077] Normalization: normalizing joint angles Mapped to [0,1]: ;

[0078] After preprocessing, the data is uploaded encrypted via a 5G network (latency <30ms).

[0079] Server-side analysis and decision-making:

[0080] Statistical analysis:

[0081] Average angle: ;

[0082] Standard deviation of force exertion time: ;

[0083] Predictive Model: Predicting the Risk of Muscle Fatigue the Next Day using an LSTM Network ,)( (Based on historical data from 3 days, including angles and electromyography data).

[0084] Decision-making rules:

[0085] like (The average angle is 80% lower than the standard), generating a solution of "increasing the range of knee flexion and wearing a joint assist device";

[0086] like (High risk of fatigue) triggers a warning of "30% reduction in training volume tomorrow".

[0087] Feedback presentation:

[0088] Visual: The tablet dynamically displays the "angle-time" curve (with the deviation range marked in red), and AR overlays virtual auxiliary lines (to indicate the correct knee flexion angle);

[0089] Auditory: Voice prompts "Current knee flexion angle is 60°, which is lower than the standard 80°. Please raise your knee by 5cm."

[0090] Tactile sensation: Flat plate vibration (continuous vibration when the movement is incorrect, stops after correction).

[0091] Example 2: Home Rehabilitation – Gait Training for Stroke Patients

[0092] (1) Equipment configuration

[0093] User mobile devices: Patient's mobile phone (with system APP installed), connected to smart insoles (pressure sensor, sampling rate 50Hz), smart glasses (AR function, camera frame rate 30fps), and heart rate bracelet (Bluetooth 5.0).

[0094] Healthcare server: A cloud-deployed SaaS service (with elastic scaling).

[0095] Cloud database: Public cloud (Alibaba Cloud), encrypted storage;

[0096] (2) Process details

[0097] Data collection: Patient walks 10 meters along the corridor (gait training):

[0098] Smart insoles collect foot pressure (Forefoot, Heel), calculate gait symmetry: (Target <10%)

[0099] Smart glasses extract step size using SLAM algorithm (distance between adjacent footprints), cadence (Steps / Time);

[0100] Heart rate bracelet uploads heart rate Calculate heart rate variability (SDNN, reflecting the state of the autonomic nervous system);

[0101] Edge preprocessing (mobile):

[0102] Outlier removal: Remove noise (sensor drift) with insole pressure <5kPa.

[0103] Feature extraction: Calculate the center of pressure trajectory (COP) for each step: ( (Use the x-coordinate of the pressure point) to determine the center of gravity offset (if the COP offset is >2cm, mark it as an abnormal frame);

[0104] Uploading only "abnormal frames" reduced traffic by 70%;

[0105] Server-side analysis and decision-making:

[0106] Trend Forecasting: Predicting the step length trend for the next 3 days using a linear regression model: (t represents the number of days of recovery);

[0107] Anomaly detection: The Isolation Forest algorithm is used to identify... (Normal >50ms indicates fatigue / abnormality);

[0108] Decision-making rules:

[0109] If the symmetry is greater than 15% and the COP offset is greater than 2cm, a plan of "wearing a foot orthosis and focusing on exerting force on the affected side" is generated.

[0110] like This triggers an alert to "pause training, drink 500ml of water and rest for 10 minutes".

[0111] Feedback presentation:

[0112] AR display: Smart glasses overlay a virtual "footprint guide line" (the line length automatically extends to indicate insufficient stride length).

[0113] Voice guidance: "Step length on the affected side is 3cm shorter, please increase the force."

[0114] Daily Report Push: Generate a daily rehabilitation report (gait parameter trend chart + text suggestions) on your mobile phone and push it to your family member's device (after authorization).

[0115] Example 3: Community Rehabilitation Center – Batch Management of Multiple Patients

[0116] (1) Equipment configuration

[0117] Mobile devices for users: The community is equipped with 10 smart tablets that connect to shared rehabilitation equipment (smart treadmills, upper limb trainers with built-in sensors) and public health monitoring stations (height, weight, body fat, blood oxygen).

[0118] Healthcare worker server: Edge server deployed at community workstation (local processing + cloud synchronization);

[0119] Cloud-based database: Regional medical cloud, connected to the Health Commission platform;

[0120] (2) Process details

[0121] Batch data collection:

[0122] COPD patients (treadmill training): Data collection speed ,slope Blood oxygen Heart rate Calculate the exercise intensity index: (Target range: 50-80);

[0123] Fracture patients (upper limb trainer): Collect joint angles Torque Calculate motion stability: (The larger the value, the more stable it is);

[0124] The public station collects all patients' basic physiological data (weight, body fat, etc.) and automatically synchronizes it to the tablet.

[0125] Edge server processing:

[0126] Parallel normalization: Heterogeneous data such as treadmill speed (0-12km / h) and trainer torque (0-50N・m) are normalized using a formula. Mapped to [0,1];

[0127] Anomaly marking: Using box plots, mark data exceeding 1.5 times the interquartile range (e.g., ...). (This triggers an emergency warning)

[0128] Server-side and cloud-based collaborative analytics:

[0129] Community Servers:

[0130] Calculate the training achievement rate: ;

[0131] Generate a patient leaderboard (for motivational training) and cluster patients using the K-means algorithm (e.g., "rapid progress group" and "slow progress group") to push differentiated treatment plans.

[0132] Cloud:

[0133] Based on regional data, a "exercise intensity-complication risk" model was constructed using the GBDT algorithm: (For patients at risk of acute exacerbation within 30 days), a "reduced intensity + oxygen therapy" plan is generated for high-risk patients.

[0134] Feedback and Management:

[0135] Patient-side: The tablet displays the "intensity-time" curve in real time (red warning zone: intensity >80 or <50), and AR prompts (such as "the speed is too fast, it is recommended to adjust to 5km / h").

[0136] Healthcare workers: The workstation's large screen displays multi-patient monitoring dashboards (target achievement rate, risk level, abnormal alerts), and clicking on a patient's avatar allows them to view detailed data and intelligent solutions;

[0137] Cloud synchronization: Community data is uploaded to the regional medical cloud, supporting remote access and guidance by experts from higher-level hospitals.

[0138] How it works (end-to-end process)

[0139] Data Acquisition and Preprocessing:

[0140] Multi-source sensors / devices collect physiological, motion, and training data in real time;

[0141] Cleaning is performed on mobile devices. Box plots, transformation (uniform format), normalization (formula), feature extraction (such as COP, completion rate), noise removal, and reduction of transmission load.

[0142] Real-time transmission and reception:

[0143] Employing SSL / TLS encryption protocols and combining 5G / Wi-Fi, millisecond-level data uploads are achieved between the patient and the server.

[0144] The server parses the protocol, verifies its integrity, temporarily stores it in the local cache, and synchronizes it to the cloud database.

[0145] Intelligent analysis and decision-making:

[0146] Statistical analysis (mean, standard deviation, achievement rate) quantifies rehabilitation progress;

[0147] Predictive models (LSTM, linear regression) predict physiological trends and rehabilitation progress;

[0148] Data mining (Apriori, Isolation Forest) is used to discover association rules and anomalies;

[0149] Multi-objective optimization algorithms (formulas) generate personalized solutions.

[0150] Multimodal feedback and storage:

[0151] Real-time patient guidance using AR, voice, and vibration multimodal methods;

[0152] Cloud-based blockchain storage enables data traceability and tamper-proofing, and supports access from multiple terminals.

[0153] The working principle and usage process of the invention: Multi-dimensional data collection: covering 3 major categories and 12 sub-dimensions of data, including physiological (heart rate, electromyography), movement (gait, joints) and training (movement, load);

[0154] Edge preprocessing: Perform cleaning (3σ criterion), transformation (uniform format), and normalization (formula: ) on mobile devices. This reduces transmission load.

[0155] Real-time encrypted transmission: Employs SSL / TLS protocol and achieves millisecond-level data upload (latency <50ms) via 5G / Wi-Fi.

[0156] Intelligent analysis and decision-making:

[0157] Statistical analysis (mean, standard deviation, pass rate);

[0158] Predictive modeling (LSTM predicts physiological trends, linear regression predicts rehabilitation progress).

[0159] Data mining (Apriori association rules, isolated forest anomaly detection);

[0160] Multi-objective optimization algorithms generate personalized solutions (formula: );

[0161] Multimodal feedback: Providing real-time guidance to patients through AR guidance, voice broadcasting, vibration feedback, and other methods;

[0162] Blockchain storage: Rehabilitation data is stored on the blockchain throughout the entire process, enabling traceability and preventing tampering.

[0163] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A mobile-based real-time rehabilitation information feedback system, characterized in that, include: User mobile terminal: integrates data acquisition module, data processing module, display module and communication module; The data acquisition module collects physiological indicator data, exercise data, and rehabilitation training data. The data processing module performs cleaning, transformation, and normalization preprocessing on the collected data. The normalization formula is as follows: ,in For a data set, For a single data entry; The display module outputs rehabilitation feedback in augmented reality, voice, and vibration multimodal formats. The communication module uses the SSL / TLS encryption protocol to implement data transmission; Healthcare server: Deploys data receiving module, data analysis module, decision-making module and communication module; The data receiving module receives user-end data and verifies its integrity. The data analysis module uses statistical analysis of the mean, standard deviation, and pass rate, using the following formula: Data mining and predictive model processing of data; The decision-making module generates personalized solutions based on medical rules and patient profiles through a multi-objective optimization algorithm. The communication module enables interaction with the user terminal and the cloud database; Cloud-based database: Stores rehabilitation data, plans, solutions, and recommendations, and uses blockchain technology to achieve data traceability and access control; The system achieves real-time interaction and intelligent control of rehabilitation information through a closed-loop process of "data acquisition - edge preprocessing - encrypted transmission - intelligent analysis - decision feedback - blockchain storage".

2. The real-time rehabilitation information feedback system based on a mobile terminal according to claim 1, characterized in that, The data acquisition module includes: Physiological data acquisition submodule: Adapts to heart rate belts, blood pressure monitors, electromyography patches, and body temperature sensors to acquire heart rate, blood pressure, electromyography signals, and body temperature data; Motion data acquisition submodule: Adapts to IMU inertial module, GPS module, pressure sensor, joint angle sensor, and extracts gait parameters and joint range of motion data; Training data acquisition submodule: Integrates a camera pose estimation algorithm based on OpenPose, a rehabilitation equipment sensor interface, and a formula for quantifying the completion rate of training movements. .

3. The real-time rehabilitation information feedback system based on a mobile terminal according to claim 1, characterized in that, The normalization process of the data processing module satisfies the following formula: ,in, Data sets The minimum and maximum values ​​are normalized and mapped to the interval [0,1], which serves as the input for subsequent analysis models.

4. The real-time rehabilitation information feedback system based on a mobile terminal according to claim 1, characterized in that, The communication module uses SSL / TLS encryption protocol to ensure data transmission security and supports multiple network adaptations such as 5G, Wi-Fi, and Bluetooth. Edge preprocessing is performed on the user's mobile device: after filtering sensor drift data and extracting key features of gait anomaly frames, the data is uploaded, reducing the transmission load by ≥60%.

5. The real-time rehabilitation information feedback system based on a mobile terminal according to claim 1, characterized in that, The data analysis module employs the following technologies: Predictive model: Predicts changes in physiological indicators over the next 3 days using an LSTM neural network; Formula for predicting recovery progress using a linear regression model: ; Data mining: The Apriori algorithm was used to discover association rules between rehabilitation actions and effects; The Isolation Forest algorithm was used to detect abnormal physiological data. Statistical analysis: Formulas for calculating the mean, standard deviation, and pass rate of the data: .

6. The real-time rehabilitation information feedback system based on a mobile terminal according to claim 1, characterized in that, The decision-making module incorporates a rehabilitation rule engine, which generates personalized rehabilitation plans based on medical guidelines and patient profiles through a multi-objective optimization algorithm. ; in, For the first A rehabilitation goal, This is the patient's actual value; The optimization objective is to minimize the sum of squared deviations between the target and the actual value, and the output includes decisions on training intensity adjustment, movement correction, and intervention timing.

7. The real-time rehabilitation information feedback system based on a mobile terminal according to claim 1, characterized in that, The display module supports multimodal output: Visual interaction: Dynamically display physiological / motor data trend charts, and overlay training movement correction prompts through AR technology; Auditory interaction: Real-time broadcast of rehabilitation suggestions using speech synthesis technology; Haptic interaction: Vibration feedback is triggered by the vibration module of the smart wearable device when the movement is incorrect, and stops after correction.

Images