Remote monitoring and guiding system for rehabilitation training of orthopedics department

Abstract

The invention discloses a remote monitoring and guiding system for rehabilitation training in the orthopedics department, particularly relates to the technical field of medical rehabilitation intelligent monitoring, and is used for solving the problems that in an existing remote rehabilitation system, timing sequence dislocation exists in action recognition and regulation and control instruction execution, and joint movement safety early warning is delayed. According to the method, localized time sequence alignment processing is performed on joint movement data and electromyographic signals based on edge computing nodes, a joint linkage action time sequence difference is generated, and an abnormal action risk level is judged in real time by combining an electromyographic signal pre-activation feature and a conflict detection mechanism of a standard action intention; historical abnormal records are matched through the cloud platform to generate a hierarchical regulation and control instruction set, and progressive safety intervention is triggered by the patient end equipment according to the dynamic deviation of the joint activity; through a collaborative mechanism of edge side action stage accurate analysis, electromyographic signal feed-forward verification and multi-stage instruction dynamic binding, millisecond-level response of joint linkage abnormity in staged rehabilitation training is realized, and the risk of joint over-limit activity is effectively avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring for medical rehabilitation, and more specifically, to a remote monitoring and guidance system for orthopedic rehabilitation training. Background Art

[0002] In the field of orthopedic postoperative rehabilitation, rehabilitation monitoring can be carried out based on wearable devices and remote communication technologies. By collecting the joint range of motion, gait, and physiological data of patients through multi-modal sensors (such as inertial sensors and electromyogram patches), and combining with a cloud platform for offline analysis and generating rehabilitation suggestions. For example, devices such as smart bracelets and pressure insoles can achieve real-time monitoring of motion parameters, and some systems have integrated video communication functions to support remote doctor-patient interaction.

[0003] However, in the end-cloud collaborative architecture, there are defects in the timing misalignment of action phase recognition and feedback instructions. That is, due to network transmission delays, multi-node data processing time-consuming, and overly long instruction transmission chains, the control instructions generated by the system (such as joint movement limit warnings and training intensity adjustments) often lag behind the actual action phases of patients, which may lead to the failure of dynamic adjustment of the rehabilitation training plan, resulting in risks of excessive joint movement or the solidification of compensatory actions, especially posing safety hazards to rehabilitation scenarios that require precise phased control such as knee joint replacement and spinal surgery. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a remote monitoring and guidance system for orthopedic rehabilitation training to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solutions: A remote monitoring and guidance system for orthopedic rehabilitation training, including the following modules: A data acquisition module, used to collect joint movement data of the target joint and electromyogram signal data of the associated muscle group and send them to the edge computing node; A timing processing module, provided in the edge computing node, used to perform timing alignment processing on the joint movement data to generate the action timing difference between different joints; An instruction generation module, used to generate a joint linkage correction instruction when the action timing difference exceeds the preset physiological linkage range; A conflict detection module, used to extract the pre-activation features before the occurrence of the target action in the electromyogram signal data, and judge whether there is a conflict based on the time-domain correlation between the pre-activation features and the standard action intention. If there is a conflict, mark the joint linkage correction instruction as a high-risk instruction; An instruction management module, provided in the cloud platform, used to match historical abnormal action records according to the high-risk instructions and joint movement data to generate a set of control instructions; An execution control module, which is installed in the patient-side device, is used to receive a set of regulation instructions, and execute high-risk instructions and interrupt training when the action time difference exceeds the preset physiological linkage range again.

[0006] In a preferred embodiment, collecting joint motion data of the target joint and electromyogram signal data of the associated muscle group and sending them to the edge computing node includes: Collecting joint motion data of the target joint through an inertial sensor, where the joint motion data includes the joint angle change rate and angular acceleration of the target joint; Collecting electromyogram signal data of the associated muscle group through an electromyogram patch, where the electromyogram signal data includes muscle activation intensity and duration; Adding timestamps to the joint angle change rate, angular acceleration, muscle activation intensity, and duration and sending them to the edge computing node, where the timestamp is used to identify the data collection time; Among them, when collecting electromyogram signal data, the electromyogram patch synchronously monitors the pre-activation stage of the target action, and the pre-activation stage is the period when the muscle electrical signal reaches the preset activation threshold before the target action starts.

[0007] In a preferred embodiment, the edge computing node performs time series alignment processing on the joint motion data to generate the action time difference between different joints, including: Performing clock deviation compensation on the joint motion data of the target joint based on the timestamp; Performing cross-modal alignment between the compensated joint motion data and the action trigger event of the electromyogram signal data; Dividing the time window of the joint motion data according to the cross-modal alignment result; Performing linkage phase difference analysis on the divided time windows of different joints; Generating the action time difference between different joints based on the linkage phase difference analysis result.

[0008] In a preferred embodiment, the division boundary of the time window is determined by the coincidence interval between the action trigger event of the electromyogram signal data and the peak value of the joint angle change rate. The linkage phase difference analysis is realized by calculating the time shift amount of the angle change trajectories of the main joint and the slave joint in the same action stage, and the action time difference is the interval time between the peak value of the main joint angle change rate and the peak value of the slave joint angle change rate.

[0009] In a preferred embodiment, when the action time difference exceeds the preset physiological linkage range, generating a joint linkage correction instruction includes: Performing multi-dimensional threshold comparison between the action time difference between different joints and the preset physiological linkage range, and the multi-dimensional threshold comparison includes positive and negative polarity detection of the action time difference and time length matching; When the positive and negative polarities of the action timing difference indicate that the main joint action lags behind and the time length exceeds the upper limit of the preset physiological linkage range, a joint linkage acceleration correction instruction is generated; When the positive and negative polarities of the action timing difference indicate that the main joint action is ahead and the time length exceeds the lower limit of the preset physiological linkage range, a joint linkage deceleration correction instruction is generated; Based on the detection result of the conflict value between the pre-activation feature and the standard action intention in the electromyogram signal data, dynamic priority marking is performed on the joint linkage acceleration correction instruction or the joint linkage deceleration correction instruction; The joint linkage correction instruction marked as high-risk is associated with a preset mandatory intervention level, and the non-high-risk instruction is associated with an adaptive adjustment level.

[0010] In a preferred embodiment, the pre-activation feature before the occurrence of the target action in the electromyogram signal data is extracted, and based on the time-domain correlation between the pre-activation feature and the standard action intention, it is judged whether there is a conflict. If there is a conflict, the joint linkage correction instruction is marked as a high-risk instruction, including: Based on the starting time point of the pre-activation stage of the electromyogram signal data and the action stage segmentation result of the joint motion data, the pre-activation feature before the occurrence of the target action is extracted; The pre-activation feature is divided into an action trigger sub-feature and an intention maintenance sub-feature; Based on the reference waveform of the standard action intention, the time-domain correlation weights of the action trigger sub-feature and the intention maintenance sub-feature are calculated; The conflict value between the pre-activation feature and the standard action intention is calculated through the weighted time-domain correlation weight. When the conflict value exceeds the preset conflict threshold, it is determined that there is an intention conflict; The joint linkage correction instruction corresponding to the intention conflict is marked as a high-risk instruction, and the marking of the high-risk instruction triggers the real-time cross-validation mechanism of the electromyogram signal data and the joint motion data.

[0011] In a preferred embodiment, the action trigger sub-feature is the root mean square value of the electromyogram signal during the period from the starting time point of the pre-activation stage to the peak value of the joint angle change rate, and the intention maintenance sub-feature is the duration fluctuation coefficient of the electromyogram signal during the period; The time-domain correlation weight is dynamically adjusted according to the action type. In walking training, the weight of the action trigger sub-feature is higher than that of the intention maintenance sub-feature, and in strength training, the weight of the intention maintenance sub-feature is higher than that of the action trigger sub-feature.

[0012] In a preferred embodiment, the cloud platform matches the historical abnormal action records according to the high-risk instructions and the joint motion data, and generates a set of regulation instructions, including: The joint motion data associated with the high-risk instructions is classified according to the action type and the rehabilitation stage, and similarity matching is performed with the abnormal data with the same classification label in the historical abnormal action record library; Screen out the historical abnormal action records in the historical abnormal action records whose action timing deviation degree and EMG signal conflict value from the current joint motion data are both higher than the preset similarity threshold based on the similarity matching results; Generate a candidate regulation instruction set according to the screened historical abnormal action records, and the candidate regulation instruction set includes a forced interruption instruction, a resistance gradient adjustment instruction, and an action guidance correction instruction; Perform priority sorting on the candidate regulation instruction set based on the real-time change trend of the current joint motion data, and the real-time change trend is the acceleration or deceleration state of the joint angle change rate; Bind the candidate regulation instruction set after priority sorting to the forced intervention level of the high-risk instruction to generate a dynamically updated regulation instruction set.

[0013] In a preferred embodiment, the patient-side device receives the regulation instruction set and executes the high-risk instruction and interrupts the training when the action timing difference exceeds the preset physiological linkage range again, including: The patient-side device receives the dynamically updated regulation instruction set and real-time monitors the action timing difference of the joint motion data; When the action timing difference exceeds the preset physiological linkage range again, execute the high-risk instruction according to the priority sorting in the regulation instruction set; Before executing the high-risk instruction, verify in real time through the EMG signal data whether the conflict value between the pre-activation feature and the standard action intention continuously exceeds the preset conflict threshold; If the conflict value continuously exceeds the preset conflict threshold, trigger the forced interruption instruction of the high-risk instruction and stop the power output of the training device; After the forced interruption instruction is executed, record the current joint motion data and the EMG signal conflict value, and transmit them back to the cloud platform to update the historical abnormal action record library.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The edge computing node is used to perform local time series alignment processing on the joint motion data and the EMG signal, and combined with the EMG pre-activation feature and the action intention conflict detection mechanism, the millisecond-level response ability to abnormal actions is realized; compared with the traditional cloud offline analysis mode, by completing the core action stage analysis and high-risk instruction pre-generation at the network edge, the delay interference of the data transmission link is effectively eliminated, so that the joint linkage correction instruction can reach the execution end within the physiological safety time window, avoiding the risk of joint over-limit activities caused by feedback lag, which not only ensures the dynamic adaptability of the rehabilitation training, but also strengthens the real-time correction ability for compensatory actions.

[0015] 2. By constructing an intelligent decision-making system for multi-dimensional data fusion, deeply correlating and analyzing electromyogram pre-activation features, joint movement trajectory deviations, and historical anomaly records to form a clinically interpretable dynamic regulation strategy; by introducing the action intention conflict value as the trigger threshold for high-risk instructions and combining it with a real-time updated physiological linkage parameter threshold library, enabling the rehabilitation intensity adjustment to be precisely matched with the patient's individualized rehabilitation stage; the two-way data flow closed-loop design between the cloud platform and the patient-side device not only realizes the continuous iterative optimization of training anomaly events, but also implements progressive intervention at the initial stage of abnormal joint range of motion through a multi-level safety redundancy mechanism for mechanical braking and power output, providing a solution with both safety and adaptability for postoperative staged rehabilitation. Brief Description of the Drawings

[0016] Figure 1 It is a schematic structural diagram of a remote monitoring and guidance system for orthopedic rehabilitation training according to the present invention. Detailed Embodiment

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Embodiment: Figure 1 A schematic structural diagram of a remote monitoring and guidance system for orthopedic rehabilitation training according to the present invention is given. A remote monitoring and guidance system for orthopedic rehabilitation training includes the following modules: A data acquisition module, configured to acquire joint movement data of a target joint and electromyogram signal data of associated muscle groups and send them to an edge computing node; A timing processing module, provided in the edge computing node, configured to perform timing alignment processing on the joint movement data to generate an action timing difference between different joints; An instruction generation module, configured to generate a joint linkage correction instruction when the action timing difference exceeds a preset physiological linkage range; A conflict detection module, configured to extract pre-activation features before the occurrence of a target action in the electromyogram signal data, and determine whether there is a conflict based on the time-domain correlation between the pre-activation features and the standard action intention. If there is a conflict, mark the joint linkage correction instruction as a high-risk instruction; An instruction management module, provided in the cloud platform, configured to match historical abnormal action records according to the high-risk instruction and the joint movement data to generate a set of regulation instructions; An execution control module, which is provided in the patient-side device, is configured to receive a set of regulation instructions and execute a high-risk instruction and interrupt the training when the action time difference exceeds the preset physiological linkage range again.

[0019] Collect the joint motion data of the target joint and the electromyogram signal data of the associated muscle group and send them to the edge computing node. The specific implementation is as follows: Collect the joint motion data of the target joint through an inertial sensor. The joint motion data includes the joint angle change rate and angular acceleration of the target joint. The inertial sensor is a multi-axis inertial measurement unit, which is fixed at the anatomical positioning point of the target joint, such as the lateral side of the patella of the knee joint or the greater trochanter of the femur of the hip joint. The joint angle change rate and angular acceleration are calculated by measuring the acceleration and angular velocity data. The joint angle change rate is the change amount of the joint angle per unit time, which is obtained through the integration operation of the angular velocity data. The integration operation is realized by accumulating the time series of the angular velocity data. The angular acceleration is the change rate of the angular velocity, which is obtained by the difference calculation of the angular velocity data at adjacent time points. The sampling frequency of the inertial sensor is set in the high-frequency range commonly used by medical wearable devices, such as a sampling frequency of 100 Hz, which meets the human motion capture requirements. The high-frequency range is adjusted according to the activity speed of the target joint. A higher sampling frequency, such as 200 Hz, corresponds to fast movements, and a lower sampling frequency, such as 50 Hz, corresponds to slow movements. The original data collected by the inertial sensor is filtered by a low-pass filter to remove high-frequency noise. The cut-off frequency of the low-pass filter is set according to the physiological activity frequency band of human joint movement. For example, high-frequency noise signals above 20 Hz are removed to retain effective motion characteristics.

[0020] Collect the electromyogram signal data of the associated muscle group through an electromyogram patch. The electromyogram signal data includes the muscle activation intensity and duration. The electromyogram patch is attached to the belly area of the muscle group related to the target action, such as the midpoint of the belly of the quadriceps femoris. After collecting the original electromyogram signal, it is amplified and filtered to obtain an effective electromyogram signal. The filtering process includes band-pass filtering, and its passband range is set according to the typical frequency band of human electromyogram signals. For example, the frequency band from 10 Hz to 500 Hz is retained to eliminate power frequency interference and baseline drift. The muscle activation intensity is a signal intensity index of the electromyogram signal within a preset time range, which is calculated by a time-domain analysis method, such as calculating the root mean square value or average absolute value of the electromyogram signal. The calculation method of the root mean square value is to square the amplitude of the electromyogram signal within a preset time window, such as 100 ms, take the average, and then take the square root. The duration is determined by the time period when the amplitude of the electromyogram signal exceeds the baseline level of the resting state. The baseline level of the resting state is obtained through statistical analysis of the electromyogram signals continuously collected in the patient's static state. For example, the electromyogram signals are continuously collected for 5 seconds when the patient maintains a relaxed posture, and the average value is taken as the baseline reference value. The baseline reference value is recalibrated every 10 minutes to adapt to the changes in the patient's state.

[0021] The joint angle change rate and angular acceleration, along with the muscle activation intensity and duration, are sent to the edge computing node after adding timestamps, which are used to identify the data acquisition time; the timestamps are generated through the clock synchronization mechanism between the edge computing node and the sensor device. The clock synchronization mechanism uses the Network Time Protocol, which is a standard protocol for device clock synchronization in the prior art. For example, the clock deviation between the sensor device and the edge computing node is controlled within 1 millisecond accuracy by broadcasting time signals through a wireless network; the joint motion data and electromyogram signal data are encapsulated into data packets in timestamp order and transmitted to the edge computing node through a wireless communication protocol. The wireless communication protocol includes one of Bluetooth or wireless local area network, and its transmission interval is dynamically adjusted according to the data acquisition frequency. For example, when the data acquisition frequency is 100 Hz, the transmission interval is set to 20 milliseconds to avoid data accumulation.

[0022] Among them, the electromyogram patch synchronously monitors the pre-activation stage of the target action when collecting electromyogram signal data. The pre-activation stage is the period when the muscle electrical signal reaches the preset activation threshold before the target action starts; the preset activation threshold is set according to the statistical data of the patient's historical data when performing standard actions. For example, when the patient performs the standard knee flexion action, the amplitude change data of the electromyogram signal of the quadriceps femoris is collected, and the root mean square value distribution range of the electromyogram signal in the pre-activation stage is statistically 2 to 4 times the standard deviation of the resting state. The lower limit value of the distribution range, that is, 2 times the standard deviation, is used as the preset activation threshold; the period of the pre-activation stage is defined as the interval time from when the electromyogram signal first exceeds the preset activation threshold to the actual occurrence of the target action. The interval time is set with reference to the pre-activation characteristics of different action types in clinical rehabilitation training. For example, the pre-activation stage time window for walking training is 50 milliseconds to 100 milliseconds, and for stair climbing training is 100 milliseconds to 200 milliseconds. The interval time is adjusted in combination with the patient's individual differences, including age, muscle reaction speed, and postoperative recovery stage.

[0023] The edge computing node performs timing alignment processing on the joint motion data to generate the action timing difference between different joints. The specific implementation is as follows: Based on the timestamp, clock deviation compensation is performed on the joint motion data of the target joint. The timestamp is added by the inertial sensor and the electromyogram patch during data acquisition. The clock deviation compensation is achieved by calculating the difference between the local clock of the edge computing node and the internal clock of the sensor device; the edge computing node obtains the internal clock timestamp of the sensor device, compares it with the local clock timestamp, calculates the clock difference between the two, and the clock difference is averaged by the sliding window method to eliminate instantaneous fluctuations and obtain a stable clock deviation compensation value; the timestamp of the compensated joint motion data is calibrated by accumulating the clock deviation compensation value to ensure the consistency of the time reference for multi-sensor data.

[0024] Perform cross-modal alignment between the compensated joint motion data and the action trigger events of the electromyogram signal data. The action trigger event is the starting time point when the pre-activation phase in the electromyogram signal data reaches the preset activation threshold. The starting time point of the pre-activation phase is determined by detecting the moment when the amplitude of the electromyogram signal first exceeds the preset activation threshold. After calibrating the timestamps of the electromyogram signal data and the joint motion data, using the action trigger event as the reference point, align the starting point of the corresponding action phase in the joint motion data to this reference point. The cross-modal alignment is achieved through time axis translation. For example, align the time point of the action trigger event of the electromyogram signal with the time point when the joint angle change rate starts to increase in the joint motion data to form a unified multi-modal action phase starting marker.

[0025] Segment the time window of the joint motion data according to the cross-modal alignment result. The segmentation boundary of the time window is determined by the coincidence interval between the action trigger event of the electromyogram signal data and the peak value of the joint angle change rate. The peak value of the joint angle change rate is the local maximum point in the joint angle change rate curve, which is identified by the sliding window extreme value detection method. The starting point of the segmentation boundary of the time window is the time point of the action trigger event, and the ending point is the time point of the peak value of the joint angle change rate. When the time interval between the action trigger event and the peak value of the joint angle change rate is within the preset reasonable range, it is determined as a valid coincidence interval and the time window is segmented. The preset reasonable range is set according to the physiological characteristics of typical actions in clinical rehabilitation training. For example, in walking training, the interval time between the action trigger event and the peak value of the joint angle change rate is usually 100 milliseconds to 300 milliseconds.

[0026] Perform a coupled phase difference analysis on the segmented time windows of different joints. The coupled phase difference analysis is achieved by calculating the time shift amount of the angle change trajectories of the main joint and the secondary joint within the same action phase. The division of the main joint and the secondary joint is based on the dominant relationship of biomechanical actions. For example, in walking training, the hip joint is the main joint and the knee joint is the secondary joint. The time shift amount of the angle change trajectory is the difference between the time point of the peak value of the joint angle change rate of the main joint and the time point of the peak value of the joint angle change rate of the secondary joint within the same action phase. The time shift amount is calculated by the dynamic time warping algorithm, which is a matching algorithm used to align time series waveforms in the prior art. By stretching or compressing the time axis, the morphological difference between the two joint angle change trajectories is minimized, thereby extracting the time shift amount.

[0027] Generate the action timing difference between different joints based on the analysis result of the linkage phase difference. The action timing difference is the interval time between the peak value of the main joint angle change rate and the peak value of the slave joint angle change rate. The interval time is calculated by directly subtracting the peak time point of the main joint angle change rate from the peak time point of the slave joint angle change rate. When the peak of the main joint is earlier than that of the slave joint, the interval time is positive, and vice versa. The calculation result of the action timing difference is used to determine whether the multi-joint linkage conforms to the preset physiological linkage parameter range. For example, in the knee flexion movement, the physiological linkage parameter range between the hip joint and the knee joint is set as the peak of the hip joint being 50 milliseconds to 150 milliseconds earlier than the peak of the knee joint.

[0028] When the action timing difference exceeds the preset physiological linkage range, generate a joint linkage correction instruction. The specific implementation is as follows: Conduct multi-dimensional threshold comparison between the action timing difference between different joints and the preset physiological linkage range. The multi-dimensional threshold comparison includes the positive and negative polarity detection of the action timing difference and the time length matching. The positive and negative polarity detection is used to judge the sequence relationship of the main joint action relative to the slave joint. The positive polarity indicates that the main joint action is ahead, and the negative polarity indicates that the main joint action is lagging. The time length matching is achieved by comparing the absolute value of the action timing difference with the upper and lower limit values of the preset physiological linkage range. The upper and lower limit values of the preset physiological linkage range are set according to the statistical results of the normal timing difference of multi-joint coordinated movement in the clinical rehabilitation medicine guidelines. For example, in the walking training, the upper limit value of the physiological linkage range between the hip joint and the knee joint is 150 milliseconds, and the lower limit value is 50 milliseconds.

[0029] When the positive and negative polarity of the action timing difference indicates that the main joint action is lagging and the time length exceeds the upper limit of the preset physiological linkage range, generate a joint linkage acceleration correction instruction. The content of the joint linkage acceleration correction instruction includes sending a tactile feedback signal to the patient-end device. The vibration frequency of the tactile feedback signal is positively correlated with the degree of the action timing difference exceeding the range. For example, when the action timing difference exceeds the upper limit value by 50 milliseconds, the vibration frequency is twice per second, and when it exceeds 100 milliseconds, it is adjusted to four times per second. The tactile feedback signal is used to prompt the patient to speed up the main joint action speed to shorten the action timing difference.

[0030] When the positive and negative polarity of the action timing difference indicates that the main joint action is ahead and the time length exceeds the lower limit of the preset physiological linkage range, generate a joint linkage deceleration correction instruction. The content of the joint linkage deceleration correction instruction includes sending a visual warning signal to the patient-end device. The flashing frequency of the visual warning signal is negatively correlated with the amplitude of the action timing difference being lower than the lower limit value. For example, when the action timing difference is lower than the lower limit value by 30 milliseconds, the flashing frequency is once per second, and when it is lower than 60 milliseconds, it is adjusted to once every two seconds. The visual warning signal is used to prompt the patient to slow down the main joint action speed to extend the action timing difference.

[0031] Based on the detection result of the conflict value between the pre-activation feature and the standard action intention in the EMG signal data, dynamic priority marking is performed on the joint linkage acceleration correction instruction or the joint linkage deceleration correction instruction; the conflict value detection result is obtained by calculating the time-domain correlation deviation between the pre-activation feature and the standard action intention, and the time-domain correlation deviation is the absolute value of the difference between the root mean square value of the EMG signal during the pre-activation phase and the reference root mean square value corresponding to the standard action intention; when the absolute value of the difference exceeds the preset conflict threshold, it is determined that the conflict value exceeds the safe range, and the joint linkage correction instruction is marked as a high-risk instruction; the preset conflict threshold is set according to the EMG signal feature differences between safe actions and abnormal actions in the patient's historical training data. For example, the upper limit of the root mean square value fluctuation range of the EMG signal of safe actions is used as the conflict threshold.

[0032] The joint linkage correction instruction marked as high-risk with dynamic priority is associated with a preset forced intervention level, and the non-high-risk instruction is associated with an adaptive adjustment level; the execution strategy corresponding to the preset forced intervention level is to immediately interrupt the patient's current training action and initiate an emergency braking protocol, such as cutting off the power output of the rehabilitation training device; the execution strategy corresponding to the adaptive adjustment level is to gradually adjust the training intensity and continuously monitor the action time difference. For example, the training resistance value is reduced by 10% every five seconds until the action time difference returns to the physiological linkage range.

[0033] Extract the pre-activation feature before the target action occurs in the EMG signal data, and judge whether there is a conflict based on the time-domain correlation between the pre-activation feature and the standard action intention. If there is a conflict, mark the joint linkage correction instruction as a high-risk instruction. The specific implementation is as follows: Based on the starting time point of the pre-activation phase of the EMG signal data and the action phase segmentation result of the joint motion data, extract the pre-activation feature before the target action occurs; the starting time point of the pre-activation phase is the time point when the EMG signal amplitude first exceeds the preset activation threshold, and the action phase segmentation result is obtained by the time window segmentation step of the joint motion data. The starting point defined in the time window segmentation step is the action trigger event time point, and the ending point is the peak time point of the joint angle change rate; the extraction range of the pre-activation feature is the time period between the starting time point of the pre-activation phase and the peak time point of the joint angle change rate. The root mean square value of the EMG signal within this time period is calculated by a sliding window. The width of the sliding window is set according to the action type. For example, the width of the sliding window is 50 milliseconds during walking training and 100 milliseconds during strength training. The moving step of the sliding window is one-fourth of the window width.

[0034] The pre-activation feature is divided into action triggering sub-feature and intention maintenance sub-feature. The action triggering sub-feature is the maximum value of the root mean square value of the electromyographic signal in the period from the starting time point of the pre-activation stage to the peak value of the joint angle change rate. The intention maintenance sub-feature is the ratio of the standard deviation of the root mean square value of the electromyographic signal in the same period to the duration; the duration is calculated by the length of the period when the electromyographic signal amplitude exceeds the resting baseline level, and the resting baseline level is determined by the average value of the electromyographic signal when the patient is in a static state; the action triggering sub-feature reflects the intensity of the action initiation, and the intention maintenance sub-feature reflects the stability of muscle activation during the execution of the action.

[0035] The time-domain correlation weights of the action triggering sub-feature and the intention maintaining sub-feature are calculated based on the reference waveform of the standard action intention, and the time-domain correlation weights are dynamically adjusted according to the action type; the reference waveform of the standard action intention is generated by the statistics of the electromyographic signal data when healthy subjects perform the standard action, for example, the reference waveform of walking training corresponds to the root mean square value curve of the electromyographic signal in the hip flexion phase; the adjustment rule of the time-domain correlation weight is as follows: the weight of the action triggering sub-feature in walking training is set to 0.7, the weight of the intention maintaining sub-feature is set to 0.3, the weight of the intention maintaining sub-feature in strength training is set to 0.6, and the weight of the action triggering sub-feature is set to 0.4. The weight ratio is determined by optimizing the clinical rehabilitation effect evaluation data, that is, it is determined by regression analysis of the historical data of the patient group, with the action completion degree (such as the rate of achieving the standard range of motion) as the dependent variable, and the action triggering sub-feature and the intention maintaining sub-feature as the independent variables, and the optimal weight distribution ratio is obtained by fitting.

[0036] The conflict value between the pre-activation feature and the standard action intention is calculated by the weighted time-domain correlation weight. The conflict value is the absolute value of the difference between the action trigger sub-feature and the maximum root mean square value of the reference waveform multiplied by the action trigger sub-feature weight, plus the absolute value of the difference between the intention maintenance sub-feature and the standard deviation duration ratio of the reference waveform multiplied by the intention maintenance sub-feature weight; the standard deviation duration ratio is the ratio of the standard deviation of the electromyographic signal to the duration in the reference waveform; when the conflict value exceeds the preset conflict threshold, it is determined to be an intention conflict. The preset conflict threshold is set according to the conflict value distribution of safe actions and abnormal actions in the patient's historical training data. For example, 1.2 times the maximum value of the safe action conflict value is taken as the threshold. For example, if the maximum value of the safe action conflict value is 80, the threshold is set to 96. The multiple is determined by counting the minimum deviation value of the abnormal action.

[0037] Mark the joint linkage correction instruction corresponding to the intention conflict as a high-risk instruction, and the marking of the high-risk instruction triggers the real-time cross-validation mechanism of the electromyogram signal data and the joint motion data; the real-time cross-validation mechanism re-collects the electromyogram signal and the joint motion data within the current period through the edge computing node, performs pre-activated feature extraction and conflict value calculation. If the results of the two conflict value calculations both exceed the preset conflict threshold, the validity of the high-risk instruction is confirmed, and the forced interruption instruction of the execution control module is activated.

[0038] The cloud platform matches the historical abnormal action records according to the high-risk instruction and the joint motion data, and generates a set of regulation instructions. The specific implementation is as follows: Classify the joint motion data associated with the high-risk instruction according to the action type and the rehabilitation stage, and perform similarity matching with the abnormal data with the same classification label in the historical abnormal action record library; the action types include but are not limited to walking training, up and down stairs training, and strength training. The rehabilitation stage is divided into early, middle, and late according to the postoperative time. The classification label is automatically generated through the action type label and the rehabilitation stage label predefined in the patient's rehabilitation plan document; the historical abnormal action record library stores the joint motion data marked as high-risk instructions and the corresponding electromyogram signal conflict values in previous trainings. The similarity matching is realized by calculating the cosine similarity of the deviation degree of the action time sequence difference between the current joint motion data and the deviation degree of the action time sequence difference in the historical record. When calculating, the time windows of the two need to be aligned. The calculation result range of the cosine similarity is from 0 to 1. When the result is higher than the preset similarity threshold, it is determined that the matching is successful; the preset similarity threshold is set according to the statistical analysis of historical data. For example, take 1.5 times the highest similarity value of the safe action matching result as the similarity threshold. If the highest similarity value of the safe action is 0.6, the similarity threshold is set to 0.9.

[0039] Based on the similarity matching results, screen out the historical abnormal action records in which the deviation degree of the action time sequence difference and the electromyogram signal conflict value of the current joint motion data are both higher than the preset similarity threshold; the deviation degree of the action time sequence difference is the absolute value of the difference between the current action time sequence difference and the median value of the preset physiological linkage range. For example, if the median value of the hip-knee linkage physiological linkage range is 100ms and the current action time sequence difference is 150ms, the deviation degree is 50ms; the electromyogram signal conflict value is the ratio of the current conflict value to the preset conflict threshold. For example, if the current conflict value is 120 and the preset conflict threshold is 100, the ratio is 1.2; the setting of the preset similarity threshold needs to meet the screening conditions of both the deviation degree and the conflict value. For example, the deviation degree threshold is set to 20% of the upper limit value of the physiological linkage range, and the conflict value threshold is set to 1.2 times the preset conflict threshold.

[0040] Generate a set of candidate regulation instructions based on the selected historical abnormal action records. The set of candidate regulation instructions includes forced interruption instructions, resistance gradient adjustment instructions, and action guidance correction instructions. The forced interruption instructions correspond to cases in the historical records where emergency braking is triggered due to joint over-limit, such as the emergency stop instruction triggered when the knee joint movement angle of the patient exceeds the safe range. The resistance gradient adjustment instructions correspond to the historical strategy of gradually reducing the training resistance, such as reducing the resistance value by 10% every 5 seconds. The action guidance correction instructions correspond to the historical scheme of adjusting the action trajectory through tactile cues, such as guiding the patient to adjust the gait through the vibration frequency. The generation rule of the candidate regulation instructions is to associate at least one regulation instruction with each historical abnormal action record. For example, one historical record is associated with one forced interruption instruction and two resistance gradient adjustment instructions. The specific number of associations is adjusted according to the execution success rate of the historical instructions.

[0041] Rank the set of candidate regulation instructions based on the real-time change trend of the current joint motion data. The real-time change trend is the acceleration or deceleration state of the joint angle change rate. In the acceleration state, the resistance gradient adjustment instructions and the action guidance correction instructions are preferentially ranked. For example, when the acceleration value of the joint angle change rate is positive and exceeds 5° / s², the priority score of the resistance gradient adjustment instructions increases by 30%. In the deceleration state, the forced interruption instructions are preferentially ranked. For example, when the acceleration value is negative and lower than -5° / s², the priority score of the forced interruption instructions increases by 50%. The priority score is achieved through weighted calculation. For example, the basic weight of the resistance gradient adjustment instructions is 0.6, and the final weight after dynamic adjustment according to the real-time trend is 0.78.

[0042] Bind the ranked set of candidate regulation instructions to the forced intervention levels of the high-risk instructions to generate a dynamically updated set of regulation instructions. The forced intervention levels are the attribute labels of the high-risk instructions, including first-level forced interruption and second-level adaptive adjustment. For example, the first-level forced interruption binds the forced interruption instructions and tops the priority, and the second-level adaptive adjustment binds the resistance gradient adjustment and action guidance correction instructions and sorts them according to the weights. The dynamic update mechanism refreshes the set of regulation instructions according to the change cycle of the real-time joint motion data. The refresh cycle is set according to the training stage. For example, it is refreshed every 5 seconds in the early postoperative stage and every 10 seconds in the middle stage. The refresh cycle is determined by the experience suggestions of clinical rehabilitation experts.

[0043] The patient-side device receives the set of regulation instructions and executes the high-risk instructions and interrupts the training when the action time difference exceeds the preset physiological linkage range again. The specific implementation is as follows: The patient-side device receives the dynamically updated regulation instruction set and monitors the action time sequence difference of joint motion data in real time; the regulation instruction set is transmitted from the edge computing node to the patient-side device through a wireless communication protocol. The patient-side device is built with an instruction parsing engine to parse the instruction content in real time and monitor the joint motion data; the monitoring of the action time sequence difference is achieved by calculating the difference between the peak time point of the current joint angle change rate and the median value of the preset physiological linkage range. The median value of the preset physiological linkage range is set according to the statistical data of joint coordinated motion of healthy people in the clinical rehabilitation medicine guidelines. For example, the median value of the physiological linkage range of the coordinated flexion of the hip joint and the knee joint is 100 milliseconds, and it is monitored in real time whether the interval time of the angle change rates of the two exceeds this range.

[0044] When the action time sequence difference exceeds the preset physiological linkage range again, high-risk instructions are executed according to the priority ranking in the regulation instruction set; the priority ranking is generated and dynamically updated by the cloud platform. After the patient-side device reads the ranking result, the instructions are executed in order from high to low according to the priority; for example, the regulation instruction set includes a forced interruption instruction (priority 1), a resistance gradient adjustment instruction (priority 2), and an action guidance correction instruction (priority 3). The device first executes the forced interruption instruction. If this instruction is not activated, the subsequent instructions are executed, and the execution logic is synchronized with the instruction update cycle of the cloud platform.

[0045] Before executing the high-risk instruction, it is verified in real time through the electromyogram signal data whether the conflict value between the pre-activation feature and the standard action intention continuously exceeds the preset conflict threshold; the verification process includes extracting the root mean square value and the duration fluctuation coefficient of the pre-activation stage of the current electromyogram signal, and calculating the time-domain correlation conflict value with the reference waveform of the standard action intention; the calculation method of the conflict value is the difference of the action trigger sub-feature multiplied by the weight plus the difference of the intention maintenance sub-feature multiplied by the weight. For example, the difference of the action trigger sub-feature is 10 (the maximum root mean square value of the reference waveform is 50, and the current value is 60), the weight is 0.7, the difference of the intention maintenance sub-feature is 5 (the standard deviation continuous ratio of the reference waveform is 2, and the current value is 7), the weight is 0.3, then the conflict value is 10×0.7 + 5×0.3 = 8.5. The calculation process is completed by the embedded processor of the patient-side device.

[0046] If the conflict value continuously exceeds the preset conflict threshold, the forced interruption instruction of the high-risk instruction is triggered and the power output of the training device is stopped; the execution of the forced interruption instruction includes cutting off the motor power supply and the air pump power output of the rehabilitation training device, and at the same time starting the mechanical braking device to lock the joint activity range; for example, in a knee joint rehabilitation device, after the forced interruption instruction is triggered, the air pump pressure drops to zero within 0.5 seconds, and the mechanical lock catches the joint rotating shaft within 1 second. The time parameters are set according to the maximum response speed of the device mechanical structure to ensure braking within the physiological safety time.

[0047] After the forced interruption instruction is executed, the current joint motion data and the myoelectric signal conflict value are recorded and transmitted back to the cloud platform to update the historical abnormal action record library; the recorded data includes the joint angle change rate, angular acceleration, root mean square value of the myoelectric signal, and conflict value, and the data is transmitted back to the cloud platform through an encrypted transmission protocol, and the encrypted transmission protocol is the common transport layer security protocol in the prior art; the cloud platform adds the data to the historical abnormal action record library of the corresponding patient and marks it as a new training abnormal event for subsequent similarity matching optimization of the generated regulation instructions. The update period of the historical library is synchronized with the patient training plan, for example, batch data update is performed at 0:00 every day.

[0048] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0049] It should be noted that the present invention can be deployed on the device itself to achieve an embedded application, or can also run on a PC or other terminal with a user interface, so as to meet various hardware environments and usage requirements.

[0050] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on the computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server or data center that includes one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0051] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the above-described systems, devices, and modules can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0052] In several embodiments provided in the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division. In actual implementation, there may be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces. The indirect couplings or communication connections of the devices or modules can be in electrical, mechanical, or other forms.

[0053] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules. They can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0054] In addition, in each embodiment of the present application, the functional modules can be integrated in a processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0055] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0056] As described above, the above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0057] Finally, the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A remote monitoring and guidance system for orthopedic rehabilitation training, characterized in that, It includes the following modules: The data acquisition module is used to acquire the joint motion data of the target joint and the electromyogram signal data of the associated muscle group and send them to the edge computing node; The timing processing module is set in the edge computing node and is used to perform timing alignment processing on the joint motion data to generate the action timing difference between different joints; The instruction generation module is used to generate a joint linkage correction instruction when the action timing difference exceeds the preset physiological linkage range; The conflict detection module is used to extract the pre-activation features before the target action in the electromyogram signal data, and judge whether there is a conflict based on the time-domain correlation between the pre-activation features and the standard action intention. If there is a conflict, the joint linkage correction instruction is marked as a high-risk instruction; The instruction management module is set in the cloud platform and is used to match the historical abnormal action records according to the high-risk instructions and the joint motion data to generate a set of control instructions; The execution control module is set in the patient-side device and is used to receive the set of control instructions, and execute the high-risk instruction and interrupt the training when the action timing difference exceeds the preset physiological linkage range again.

2. The remote monitoring and guidance system for orthopedic rehabilitation training according to claim 1, characterized in that, Acquiring the joint motion data of the target joint and the electromyogram signal data of the associated muscle group and sending them to the edge computing node includes: Collecting the joint motion data of the target joint through an inertial sensor, and the joint motion data includes the joint angle change rate and angular acceleration of the target joint; Collecting the electromyogram signal data of the associated muscle group through an electromyogram patch, and the electromyogram signal data includes the muscle activation intensity and duration; Adding timestamps to the joint angle change rate, angular acceleration, muscle activation intensity, and duration and sending them to the edge computing node, where the timestamp is used to identify the data acquisition time; Among them, when collecting the electromyogram signal data, the electromyogram patch synchronously monitors the pre-activation stage of the target action, and the pre-activation stage is the time period when the muscle electrical signal reaches the preset activation threshold before the target action starts.

3. The remote monitoring and guidance system for orthopedic rehabilitation training according to claim 1, characterized in that, The edge computing node performs timing alignment processing on the joint motion data to generate the action timing difference between different joints, including: Performing clock deviation compensation on the joint motion data of the target joint based on the timestamp; Performing cross-modal alignment on the compensated joint motion data and the action trigger event of the electromyogram signal data; Dividing the time window of the joint motion data according to the cross-modal alignment result; Performing linkage phase difference analysis on the divided time windows of different joints; Generating the action timing difference between different joints based on the linkage phase difference analysis result.

4. The remote monitoring and guidance system for orthopedic rehabilitation training according to claim 3, characterized in that, The segmentation boundary of the time window is determined by the coincidence interval between the action trigger event of the electromyogram signal data and the peak value of the joint angle change rate. The linkage phase difference analysis is realized by calculating the time shift amount of the angle change trajectories of the main joint and the slave joint in the same action stage. The action timing difference is the interval time between the peak value of the main joint angle change rate and the peak value of the slave joint angle change rate.

5. An orthopedic rehabilitation training remote monitoring and guidance system according to claim 1, characterized in that, When the action timing difference exceeds the preset physiological linkage range, generating a joint linkage correction instruction includes: Performing multi-dimensional threshold comparison between the action timing difference between different joints and the preset physiological linkage range. The multi-dimensional threshold comparison includes the positive and negative polarity detection and time length matching of the action timing difference; When the positive and negative polarities of the action timing difference indicate that the main joint action lags behind and the time length exceeds the upper limit of the preset physiological linkage range, a joint linkage acceleration correction instruction is generated; When the positive and negative polarities of the action timing difference indicate that the main joint action is ahead of schedule and the time length exceeds the lower limit of the preset physiological linkage range, a joint linkage deceleration correction instruction is generated; Based on the detection result of the conflict value between the pre-activation feature and the standard action intention in the electromyogram signal data, dynamic priority marking is performed on the joint linkage acceleration correction instruction or the joint linkage deceleration correction instruction; The joint linkage correction instruction with a dynamic priority marked as high risk is associated with a preset mandatory intervention level, and the non-high-risk instruction is associated with an adaptive adjustment level.

6. The remote monitoring and guidance system for orthopedic rehabilitation training according to claim 1, characterized in that, Extract the pre-activation feature before the target action occurs in the electromyogram signal data, and judge whether there is a conflict based on the time-domain correlation between the pre-activation feature and the standard action intention. If there is a conflict, mark the joint linkage correction instruction as a high-risk instruction, including: Based on the starting time point of the pre-activation stage of the electromyogram signal data and the action stage segmentation result of the joint motion data, extract the pre-activation feature before the target action occurs; Divide the pre-activation feature into an action trigger sub-feature and an intention maintenance sub-feature; Calculate the time-domain correlation weights of the action trigger sub-feature and the intention maintenance sub-feature based on the reference waveform of the standard action intention; Calculate the conflict value between the pre-activation feature and the standard action intention through the weighted time-domain correlation weights. When the conflict value exceeds the preset conflict threshold, it is determined as an intention conflict; Mark the joint linkage correction instruction corresponding to the intention conflict as a high-risk instruction, and the marking of the high-risk instruction triggers the real-time cross-validation mechanism of the electromyogram signal data and the joint motion data.

7. An orthopedic rehabilitation training remote monitoring and guidance system according to claim 6, characterized in that, The action trigger sub-feature is the root mean square value of the electromyogram signal during the period from the starting time point of the pre-activation stage to the peak value of the joint angle change rate, and the intention maintenance sub-feature is the duration fluctuation coefficient of the electromyogram signal during the period; The time-domain correlation weights are dynamically adjusted according to the action type. In walking training, the weight of the action trigger sub-feature is higher than that of the intention maintenance sub-feature, and in strength training, the weight of the intention maintenance sub-feature is higher than that of the action trigger sub-feature.

8. The remote monitoring and guidance system for orthopedic rehabilitation training according to claim 1, characterized in that, The cloud platform matches the historical abnormal action records according to the high-risk instructions and the joint motion data, and generates a set of regulation instructions, including: Classify the joint motion data associated with the high-risk instructions according to the action type and the rehabilitation stage, and perform similarity matching with the abnormal data with the same classification label in the historical abnormal action record library; Based on the similarity matching result, screen out the historical abnormal action records in which the deviation degree of the action timing difference and the electromyogram signal conflict value of the current joint motion data are both higher than the preset similarity threshold; Generate a candidate regulation instruction set according to the screened historical abnormal action records. The candidate regulation instruction set includes a forced interruption instruction, a resistance gradient adjustment instruction, and an action guidance correction instruction; Perform priority sorting on the candidate regulation instruction set based on the real-time change trend of the current joint motion data. The real-time change trend is the acceleration or deceleration state of the joint angle change rate; Bind the candidate regulation instruction set after priority sorting to the mandatory intervention level of the high-risk instruction, and generate a dynamically updated regulation instruction set.

9. The remote monitoring and guidance system for orthopedic rehabilitation training according to claim 1, wherein The patient-side device receives a set of regulation instructions and executes high-risk instructions and interrupts training when the action time sequence difference exceeds the preset physiological linkage range again, including: The patient-side device receives a dynamically updated set of regulation instructions and real-time monitors the action time sequence difference of joint movement data; When the action time sequence difference exceeds the preset physiological linkage range again, execute high-risk instructions according to the priority sorting in the set of regulation instructions; Before executing the high-risk instruction, verify in real time through the electromyogram signal data whether the conflict value between the pre-activation feature and the standard action intention continuously exceeds the preset conflict threshold; If the conflict value continuously exceeds the preset conflict threshold, trigger a forced interruption instruction for the high-risk instruction and stop the power output of the training device; After the forced interruption instruction is executed, record the current joint movement data and the electromyogram signal conflict value, and transmit them back to the cloud platform to update the historical abnormal action record library.

Images