97 results about "Gait phase" patented technology

The invention discloses a humanoid robot motion control method based on gait planning and reinforcement learning, and belongs to the technical field of humanoid robot motion control. According to the method, the problems of low convergence speed in a training process and poor stability and reliability of a trained model of an existing reinforcement learning method are solved. According to the method, a reinforcement learning model runs in parallel under multiple environment instances, in the running process, a target foothold is generated in real time based on feedback state information, the generated reference foothold and a mass center track serve as an action reference or reward target of a controller, and a continuous and dynamically-adjusted foot end track is constructed in combination with gait phase information. And after the reinforcement learning model is trained by utilizing all the collected trajectory data, motion control can be performed on the humanoid robot through the trained reinforcement learning model, so that the robot realizes stable walking similar to human beings under the condition of not depending on a predefined template. The method can be applied to motion control of the humanoid robot.

The invention discloses a pneumatic-biological signal feedback cooperative control method based on an anti-gravity rehabilitation treadmill, and the method comprises the steps: synchronously collecting sEMG, IMU and plantar pressure signals, and completing the time-space registration through clock synchronization, time warping and extended Kalman filtering; muscle activation degree, pressure center offset and joint angular velocity are extracted to construct multidimensional features, and gait phase switching is predicted through LSTM-CNN; the TS fuzzy system generates a preliminary control instruction according to expert rules, and Q-learning takes pressure balance and motion smoothing as reward functions to optimize the weight reduction proportion and the treadmill speed online; a waist annular air cavity is driven to be inflated to form a pressure supporting field so as to lift a pelvis and guide a lower limb gait mechanical mode, so that the muscle activation degree, the joint angular motion trail and the pressure center migration characteristic are dynamically influenced, a servo motor tracks phase nodes in a feedforward-feedback mode, and risk early warning and training parameter self-adaptive closed-loop rehabilitation are achieved.

The invention relates to the technical field of rehabilitation robot control, and particularly discloses a gait prediction compensation admittance control method and system for lower limb rehabilitation exoskeleton, and the method comprises the steps: inputting a multi-dimensional motion vector into a time sequence prediction model, and outputting a prediction result of a gait phase at a future moment; if the maximum probability output by the time sequence prediction model is lower than a preset threshold value, automatically switching to a pre-constructed rule criterion to carry out gait phase recognition; an admittance control parameter matched with the gait phase prediction result is called from an admittance parameter mapping table, and a main control torque is calculated based on the called admittance control parameter; calculating a compensation torque based on the deviation between the actual human-computer interaction force and the expected interaction force; and the main control torque and the compensation torque are superposed to serve as a final torque driving instruction of the exoskeleton motor. Real-time gait prediction, admittance parameter self-adaptive adjustment and interaction force error dynamic compensation can be fused, and high-flexibility and high-safety man-machine collaborative rehabilitation training is achieved.

The invention relates to an AR projection interaction treadmill virtual scene generation method and device, equipment and a medium. The method comprises the following steps: acquiring motion inertia data and running belt motion parameters of a user on a running machine; respectively carrying out motion state extraction and gait phase extraction according to the motion inertia data and the running belt motion parameters to obtain a current motion state vector and gait phase information of the user; inputting the motion state vector and the gait phase information into a pre-trained motion state prediction model to obtain a preset predicted motion state of a plurality of frames in the future; based on the predicted motion state, performing layered scene generation through a scene generator to obtain a corresponding virtual scene; and performing projection mapping transformation processing on the virtual scene, and obtaining real-time AR projection data in combination with low-delay compensation. By adopting the method, a highly immersive and interactive virtual training environment can be provided in an AR projection mode.

The invention relates to the technical field of intelligent control, and discloses a hip joint exoskeleton control method and system based on gait mode prediction and gait phase estimation, a terminal and a storage medium, and the method comprises the steps: constructing and training an initial unsupervised terrain detection model, and obtaining a target unsupervised terrain detection model; inputting the environment binary image into a target unsupervised terrain detection model to obtain a terrain detection result, and obtaining a gait mode prediction result based on a majority voting strategy; according to the hip joint angular velocity, target gait phase estimation is obtained based on a self-adaptive oscillator system; on the basis of the expected auxiliary torque curve, net auxiliary torque is obtained according to the gait mode prediction result and target gait phase estimation, and hip joint exoskeleton assistance modes are switched according to the net auxiliary torque. According to the method, marking of a large amount of data is avoided through the unsupervised terrain detection model, terrain detection and exoskeleton control are combined, and the terrain self-adaptive walking capacity of the exoskeleton is enhanced.

Various examples are provided related to gait kinematics. A methodology for human in the loop optimization (HILO) for use of a hip exoskeleton is presented. In one example, a method includes monitoring a gait phase of an exoskeleton and controlling switching time between admittance parameters associated with actuator control of the exoskeleton, where the switching time is controlled based upon the monitored gait phase. The admittance parameters can be predetermined and can be user specific. The time of the switching can be determined from use of the exoskeleton and can be determined using reinforcement learning.

The invention relates to the technical field of intelligent control, and discloses a power-assisted control method of a lower limb pneumatic flexible exoskeleton and a lower limb pneumatic flexible exoskeleton system.The method comprises the steps that healthy side plantar pressure data are obtained, healthy side key gait event nodes are recognized, and the power-assisted control method of the lower limb pneumatic flexible exoskeleton is obtained according to the healthy side key gait event nodes and ankle joint angle data; estimating an uninjured side continuous gait phase, and calculating an affected side target gait phase according to the uninjured side continuous gait phase; according to the target gait phase of the affected side, auxiliary torque and tension are determined; according to the auxiliary torque and the pulling force, target air pressure for supporting the airbag and shrinking the airbag is determined, an electromagnetic valve is controlled to execute deflation or inflation operation according to the target air pressure, and the internal pressure of the airbag is controlled within a target range. The invention aims to provide flexible assistance conforming to the biomechanical characteristics of the ankle joint of a human body for people with ankle joint dysfunction, improve the walking ability and improve the safety and comfort of rehabilitation training.

The invention relates to a stroke gait phase recognition method and system based on a multi-stage model, and the method specifically comprises the following steps: employing a self-adaptive sliding window algorithm based on index-standard deviation fusion, carrying out the dynamic fragment interception of a gait signal of a stroke patient, and obtaining the original stroke gait data; preprocessing the original stroke gait data to obtain original gait features; constructing an integrated feature enhancement unit (IFEU), inputting original gait features, and extracting time context features of the stroke gait signals; constructing a hierarchical gait feature aggregation unit HGFAU, inputting time context features, performing gait feature extraction of each level, and obtaining multiple fusion features after fusion; and inputting the multiple fusion features into a linear layer classifier, and classifying the input features to obtain a stroke gait phase recognition result. According to the invention, gait phase recognition can be realized, and objective walking state evaluation and abnormity monitoring can be carried out on rehabilitation treatment assistance of a patient.

The application relates to the technical field of intelligent control, and discloses a hip exoskeleton control method and system based on gait pattern prediction and gait phase estimation, a terminal and a storage medium, the method comprising the following steps: constructing an initial unsupervised terrain detection model and training the model to obtain a target unsupervised terrain detection model; inputting an environment binary image into the target unsupervised terrain detection model to obtain a terrain detection result, and obtaining a gait pattern prediction result based on a majority voting strategy; obtaining a target gait phase estimation based on an adaptive oscillator system according to a hip joint angular velocity; obtaining a net auxiliary torque based on an expected auxiliary torque curve, a gait pattern prediction result and the target gait phase estimation, and switching a hip exoskeleton assistance mode according to the net auxiliary torque. The unsupervised terrain detection model avoids labeling a large amount of data, the terrain detection is combined with the exoskeleton control, and the terrain self-adaptive walking capability of the exoskeleton is enhanced.

The invention relates to the technical field of intelligent walking assisting equipment, and discloses a blind person intelligent guiding walking assisting method and system based on deep learning. The intelligent guiding power-assisted walking method for the blind is applied to guiding power-assisted electronic equipment, and specifically comprises the following steps: S101, receiving a starting signal sent by a user terminal, starting a multi-source sensor array to collect environment data in parallel, executing a Kalman-Transform space-time alignment algorithm, establishing a multi-modal data space-time unified coordinate system, and establishing a multi-modal data space-time unified coordinate system; when gait phase changes are collected and detected, a multi-source sensor array is activated and dynamically triggered for high-precision scanning, and a threat evaluation model is generated to calculate an obstacle collision probability point cloud matrix; and S102, processing the obstacle collision probability point cloud matrix through Point VoxelNet to generate a three-dimensional environment skeleton. The cross-modal attention mechanism of the millimeter wave radar and the vision improves the rain and fog weather recognition accuracy, and in addition, the dynamic threat evaluation module is combined with the double-threshold early warning strategy, so that the collision false alarm rate is effectively reduced.

The invention discloses an exoskeleton driving method based on real moment constraint and continuous gait phase modeling, which comprises the following steps: acquiring multi-modal signal data in the gait process of a subject by using a multi-sensor synchronous acquisition system, including IMU signals, EMG signals, plantar pressure signals and optical motion capture data; preprocessing the multi-mode signal data to obtain a preprocessed signal; performing continuous gait phase prediction and gait pattern recognition according to the preprocessed signal; estimating a joint instantaneous moment by using a pre-trained moment estimation network based on the predicted gait phase and gait mode; and converting the obtained instantaneous moment of the joint into a driving signal of an exoskeleton actuator. According to the method, discrete event triggering is replaced by continuous gait phase modeling, so that the smoothness and robustness of control are ensured; meanwhile, a mechanical real constraint and individualized moment estimation mechanism is introduced, stable man-machine cooperative control is achieved, and the adaptability and use experience of the exoskeleton in a complex environment are remarkably improved.

The invention discloses a method and system for driving and controlling a lower limb exoskeleton through brain-computer interface intention confidence, and relates to the technical field of medical rehabilitation. The method comprises the following steps: firstly, identifying an original electroencephalogram signal to obtain intention confidence, and mapping an impedance parameter group containing joint stiffness and a damping coefficient in combination with gait phase information; secondly, extracting a motion position and an interaction force error, comparing an intention confidence coefficient with an actual execution state by using a sliding time window to obtain a nerve matching error, and fusing the three into a total composite error; meanwhile, the human-machine coupling compliance is evaluated based on the human body joint angle and the exoskeleton torque variation, and a safety gain coefficient is discriminated and generated; and finally, performing gain operation on the total composite error based on the impedance parameter group, and performing safety gain coefficient correction to obtain a target control torque to be output to a driving actuator. In this way, a nerve-force-motion three-closed-loop framework is constructed, and compliant self-adaptive on-demand assistance and high-safety man-machine collaborative rehabilitation with defense protection are achieved.

The invention provides an electrical stimulation and lower limb exoskeleton device fusion control method and system, and the method comprises the steps: driving an exoskeleton to move through bottom admittance control based on the actual human-computer interaction torque and expected torque of a knee joint and reference trajectory data of the knee joint in a gait cycle, and triggering ankle joint electrical stimulation through a fixed stimulation parameter group; after the gait cycle is finished, torque errors are calculated, and periodic iteration updating is conducted on the knee joint reference trajectory data through an iterative learning control algorithm till convergence is conducted to obtain target knee joint reference trajectory data; and then determining a target moment of a gait phase based on the target knee joint reference trajectory data, and obtaining a corresponding target stimulation parameter group from the set stimulation parameter time mapping library to replace the fixed stimulation parameter group for subsequent stimulation. According to the method, through single-cycle and cross-cycle trajectory iteration, the adaptation defect of a traditional fixed trajectory is avoided, the stimulation parameters are synchronously optimized, and the dorsiflexion assisting effect is indirectly improved.

This disclosure relates to the field of exoskeleton technology, and discloses an exoskeleton control method, storage medium, and exoskeleton. The exoskeleton control method includes: acquiring gait feature values; inputting the gait feature values ​​into a fuzzy neural network model to obtain a target gait output value; determining candidate gait phases and previous gait phases based on the target gait output value, wherein the gait output value of the previous gait phase is temporally adjacent to the target gait output value; determining the target gait phase based on the candidate gait phases and the previous gait phase; and controlling the exoskeleton based on the target gait phase. This embodiment identifies gait phases by comprehensively considering previous gait phases and candidate gait phases, which helps improve the accuracy and reliability of gait phase identification, thereby enabling the exoskeleton to more effectively drive the user's limbs, thus providing human-machine compliance.

The invention relates to the technical field of rehabilitation medical robots, and particularly provides a joint angle vector generation method and a gait control method.The generation method comprises the steps that a first gait phase and multi-sensor data at the current moment are obtained; the first gait phase is the gait phase of the lower limb exoskeleton robot at the previous moment, and the multi-sensor data comprises multi-modal motion data and visual data; predicting a second gait phase according to the first gait phase and the multi-modal motion data; the second gait phase is the gait phase of the lower limb exoskeleton robot at the current moment; adjusting a reference joint angle vector and a time scaling factor of a self-adaptive joint angle vector generator of which the weight is optimized in advance according to the visual data, and then generating an expected joint angle vector according to the second gait phase by utilizing the adjusted self-adaptive joint angle vector generator; according to the method, the joint angle vector which adapts to the current environment and is high in gait bionic performance can be generated.

The present invention relates to a device for training a personalized gait cycle model configured to generate gait phase information representative of a progression state of a patient in a gait cycle. The invention also relates to a device and a method for calculating a starting time and an ending time and an intensity of at least one sequence of electrical pulses applied to at least one of nerves and muscles of a patient using at least one pair of electrodes based on gait phase information representing a progression state of the patient in a gait cycle, the gait phase information is obtained using a personalized model of gait cycles obtained from a device for training.

The application discloses a portrait inspection and identification method based on gait cycle decomposition and multi-phase force analysis, and comprises the following steps: extracting the skeleton key points of a target object in a target object image set in a comparison video and the confidence score of each skeleton key point; extracting the features of the target object in the target object image set according to the skeleton key points of the target object, and obtaining the target object features, wherein the target object features comprise basic features, gait cycle features, gait stage features and gait parameter features; integrating the basic features, the gait cycle features and the gait stage features to obtain comprehensive features; respectively calculating the similarity of the target object features and the comprehensive features in the comparison video to obtain similarity calculation results and overall similarity calculation results; and obtaining the identity identification result of the target object according to the similarity calculation results and the overall similarity calculation results.

The application discloses a multi-modal fusion motion intention recognition method which can be embedded and deployed, is based on a multi-modal sensor perception system, integrates inertial measurement unit data and human-computer interaction force data, adopts a hierarchical sliding window strategy to adapt to recognition requirements of different time scales, constructs a light-weight CNN-LSTM double-branch deep learning model, extracts local space-time features of multi-modal data through a CNN module, models gait timing dependency by means of an LSTM module, realizes collaborative recognition of gait phases and motion modes, reduces model calculation complexity through a light-weight optimization strategy, completes embedded platform deployment adaptation, ensures balance between recognition accuracy and real-time performance, can be directly deployed on an STM32 or other resource-limited embedded controller, is suitable for portable devices such as lower limb rehabilitation exoskeletons, has high recognition accuracy and strong generalization ability, and can provide accurate and real-time intention decision support for human-robot collaborative control of rehabilitation robots.

This invention relates to the field of quadruped robot navigation technology and discloses an intelligent navigation system for quadruped robots that combines behavior tree and visual recognition. The system includes modules for path processing, turn prediction, behavior tree decision-making, obstruction judgment, visual recognition and language evaluation, and navigation decision-making. The path processing module integrates multi-sensor data to acquire pose and gait state; the turn prediction module calculates safe turning speed and trigger distance based on a gait phase sliding window; the behavior tree decision-making module dynamically adjusts node priorities according to trigger intensity, and the switching timing is controlled by a gait phase arbiter; the obstruction judgment module generates path obstruction flags; the visual recognition module adopts a two-level architecture, calling a visual language model to evaluate the landing area for static traversable obstacles; and the navigation decision-making module integrates landing semantic confidence and outputs commands to cross, detour, wait, or return. This invention enables quadruped robots to achieve proactive predictive navigation and intelligent obstacle avoidance decision-making in complex industrial environments.

This application discloses a method and system for controlling a lower limb exoskeleton using a brain-computer interface based on intent confidence, relating to the field of medical rehabilitation technology. First, the system identifies the original electroencephalogram (EEG) signal to obtain intent confidence, and maps an impedance parameter set including joint stiffness and damping coefficients by combining gait phase information. Second, it extracts motion position and interaction force errors, compares the intent confidence with the actual execution state using a sliding time window to obtain neural matching errors, and merges these three into a total composite error. Simultaneously, it assesses human-machine coupling compliance based on changes in human joint angles and exoskeleton torque, and generates a safety gain coefficient. Finally, it performs gain calculations on the total composite error based on the impedance parameter set, and corrects it with the safety gain coefficient to obtain the target control torque for output to the actuator. This constructs a neural-force-motor three-loop architecture, achieving compliant and adaptive on-demand assistance and highly safe human-machine collaborative rehabilitation with defensive protection.

This invention discloses a gait prediction method and system based on a support-aware multi-scale temporal convolutional network. The method includes the following steps: S1, collecting plantar pressure data through pressure insoles and ground reaction force data through a force plate to construct a gait dataset; S2, performing data preprocessing and feature enhancement on the dataset from step S1; S3, establishing a multi-scale support-aware temporal convolutional network (MSA-TCN) model; S4, training the network model from step S3 using supervised learning, inputting the dataset from step S2 into the trained network model for prediction, and outputting the predicted ground reaction force. This invention overcomes the shortcomings of existing technologies in multi-scale temporal feature extraction, gait phase-aware modeling, and synchronous high-precision prediction of multi-dimensional mechanical parameters.

The invention discloses a quantitative weight-reducing lower limb exoskeleton rehabilitation robot which comprises a moving frame and further comprises mounting plates symmetrically fixed to the moving frame. The lower limb exoskeleton assembly is mounted on the moving frame through the mounting plate. An exoskeleton weight reduction module integrates a plantar pressure sensor and an IMU (inertial measurement unit), a support phase / swing phase time sequence node is accurately judged through an improved gait phase recognition algorithm, an output force value of an electric push rod is dynamically adjusted through closed-loop control on the basis of a personalized weight reduction threshold value, and the weight reduction precision is improved. Quantitative output and real-time calibration of weight reduction are achieved, the technical bottleneck that existing equipment is poor in weight reduction precision is solved, the stress load of a fracture part is precisely reduced, secondary injury is avoided, and gait naturalness and coordination are optimized; meanwhile, the corresponding joint driving modules can be installed and activated as required according to the fracture part of the patient, the monitoring modules are flexibly arranged, and the equipment suitability and the training effectiveness are improved.

This invention belongs to the technical field of rehabilitation robots, specifically relating to a continuous phase control method for a lower limb rehabilitation exoskeleton that adapts to user gait. It employs a multi-task learning model, simultaneously outputting user identity probability and movement pattern probability to generate a probability map representing the similarity between the current gait and a certain movement pattern of a sample user in a sample database. A gait phase variable parameter prediction model is used to predict gait phase variable parameters, employing angular velocity data from the initial stage of the current gait cycle and the matched movement pattern to predict the gait phase variable parameters for the current cycle. Finally, a pre-constructed relational constraint model corresponding to the matched sample user and their movement pattern is used to obtain the desired joint angles from the real-time gait phase, and the exoskeleton is tracked and controlled via a PD controller. This invention can adaptively adjust the exoskeleton's motion assistance mode according to the user's gait characteristics, and is suitable for rehabilitation training of patients with lower limb motor dysfunction.

The application relates to the technical field of behavior feature recognition, and discloses an anti-crosstalk biological rehabilitation electrical signal safety control method and system in a high-density scene. The method comprises the following steps: calculating a bilateral average distance according to an environmental distance; calculating a surface electromyogram signal root mean square value according to an electromyogram original signal; analyzing a pelvis tilt angle standard deviation according to the pelvis tilt angle; calculating a hip joint adduction auxiliary torque according to the surface electromyogram signal root mean square value and a gait phase; calculating a hip joint coronal plane impedance value according to the gait phase and the pelvis tilt angle; calculating an intention state according to the surface electromyogram signal root mean square value, the pelvis tilt angle and a channel state; calculating a hip joint driving torque according to the hip joint adduction auxiliary torque and the hip joint coronal plane impedance value; and calculating a system mode according to the bilateral average distance, the surface electromyogram signal root mean square value and the pelvis tilt angle standard deviation. The application improves the lower limb rehabilitation training efficiency by adaptively adjusting the gait.

The invention relates to a multi-joint layered cooperative intelligent control method for a power-assisted exoskeleton, and belongs to the field of robot exoskeletons. Comprising the following steps: acquiring a real-time hip joint angle, updating an adaptive oscillator model based on the real-time hip joint angle, and obtaining a real-time gait phase; segmenting the real-time gait phase to obtain gait phases of a plurality of first sampling points; acquiring the height, the weight, the lower limb joint center and the stride frequency of the user to construct a power-assisted curve of the relation between the lower limb joint power-assisted torque and the gait phase; and acquiring the actual torque of the lower limb joint corresponding to each first sampling point, inputting the gait phase corresponding to each first sampling point into the power-assisted curve for updating to obtain an optimized power-assisted curve, and realizing power assistance of the lower limb joint based on the optimized power-assisted curve. According to the method, the hip joint angle and the moment thereof and the knee joint angle and the moment thereof are associated through the same gait phase, cooperative assistance of the hip joint and the knee joint is achieved, the effectiveness and smoothness of exoskeleton assistance are improved, and the user experience is improved.

The invention relates to a cerebral apoplexy gait phase recognition method and system under class imbalance, and belongs to the technical field of cerebral apoplexy gait analysis. The method comprises the following steps: acquiring original gait signal data of a stroke patient; inputting the original gait signal data into a feature enhancement unit, and performing global average pooling, time sequence feature enhancement branch and channel feature enhancement branch processing to obtain a first fusion feature; inputting the first fusion feature into a long-short sequence feature extraction unit to obtain a second fusion feature; inputting the second fusion feature into a feature calibration unit to obtain a calibration feature; the calibration features pass through a linear layer classifier to obtain a stroke gait phase recognition result; in the training process, a self-adjusting cost-sensitive loss function is adopted to carry out parameter optimization. The accuracy of stroke gait phase recognition can be improved.

The invention relates to a gait detection method and system, electronic equipment and a readable medium. The method comprises the following steps: acquiring multi-source sensor signals acquired by a waist inertial measurement unit and a hip joint motor, and constructing a multi-dimensional instantaneous feature vector based on the multi-source sensor signals; inputting the multi-dimensional instantaneous feature vector into a lightweight support vector machine model for classification prediction to obtain a preliminary classification result of the current gait; and inputting the preliminary classification result into a preset finite state machine for sequential logic verification, and outputting a final effective gait when the state conversion satisfies the state conversion path constraint and the last state duration satisfies the minimum threshold condition, thereby solving the technical problem of misclassification caused by lack of sequential verification during gait phase recognition in the related technology, and improving the recognition accuracy of the gait phase. The technical effect of improving the gait classification precision is achieved.

The embodiment of the invention discloses a gait phase recognition method and system based on a convolutional neural network, and the method comprises the steps: obtaining a plurality of sample motion signals of a sample motion object, carrying out the preprocessing of the sample motion signals, and obtaining a sample motion vector; inputting the sample motion vector into a gait phase recognition model for feature extraction to obtain a gait phase feature, and performing gait phase prediction based on the gait phase feature to obtain a gait phase prediction tag; the gait phase loss is determined according to the difference between the gait phase prediction label and the gait phase real label, a gait phase recognition model is trained based on the gait phase loss, and in the training process, the gait phase recognition model is optimized, so that the difference between the gait phase prediction label and the gait phase real label is reduced; the real-time motion signal of the target motion object is obtained, gait phase recognition is carried out based on the real-time motion signal, the gait phase of the target motion object is obtained, and the accuracy of gait phase recognition can be improved.

The invention provides an intelligent wearable health monitoring system, and relates to the technical field of sports biomechanics, and the method comprises the steps: collecting original sensing data streams of a triaxial accelerometer and a gyroscope in real time, and converting the original sensing data streams to a human skeleton coordinate system through a rotation matrix; calculating the Euler angle of the limb joint based on the converted data, and extracting gait phase parameters in combination with the curvature features of the motion trail; calculating a real-time stride frequency according to the gait phase parameter; when a plurality of continuous stride frequency sampling values exceed a preset medium-high intensity motion threshold value, generating a medium-high intensity motion state identifier; synchronously executing according to the medium-high intensity motion state identification, constructing a blood glucose dynamic evaluation window based on a continuous sampling sequence of a blood glucose sensor, calculating a time sequence gradient of blood glucose reading in the window, and generating a dynamic compensation coefficient according to the time sequence gradient. According to the invention, dynamic and accurate health monitoring and risk early warning in a motion scene are realized.

The present application relates to the field of physiological motion detection in medical devices, and particularly relates to a knee joint injury postoperative rehabilitation action recognition method, device and storage medium. The method comprises: acquiring multi-source sensor data and preprocessing to generate a preprocessing sequence; based on the preprocessing sequence, posture and gait phase analysis is performed to generate a posture sequence and a phase sequence; then segment boundary detection and action candidate generation are performed, and time sequence features are extracted, and through hierarchical action atlas matching and sequence discrimination model reasoning, an action category and a completion degree score are generated; finally, quality evaluation and safety threshold linkage comparison are performed, and strategy mapping is performed to generate a parameter update package. The present application can realize accurate recognition and quantitative evaluation of rehabilitation actions, and form a closed loop feedback, effectively improving the safety and individualization level of rehabilitation training.