Smart rehabilitation device

Abstract

The embodiments of the present application disclose a smart rehabilitation device, which aims to solve the problems of the prior art, such as the inability to adjust position and height, low evaluation accuracy, and low rehabilitation management efficiency. The device comprises a mobile carriage, an embedded host computer, a mobile unit, and a wearable unit, each attached to the mobile carriage. The mobile unit and the wearable unit are both connected to the embedded host computer, which is equipped with a rehabilitation training module, a rehabilitation evaluation module, a data collection module, a data processing module, a movement control module, and a storage module. The mobile unit includes rollers, an electric push rod, and a drive unit installed on the mobile carriage, and the drive unit is connected to the movement control module. The wearable unit includes rehabilitation training gloves and a sensory device, and both the rehabilitation training gloves and the sensory device perform rehabilitation training via the rehabilitation training module and transmit rehabilitation training data to the data collection module.

Description

[Technical Field]

[0001] The embodiments of this application relate to the medical technology field, and more specifically to a smart rehabilitation device. [Background technology]

[0002] As people's living standards improve and the population ages rapidly, cardiovascular diseases have become a threat to health, with stroke being a major cause of death and permanent disability. Clinical research and practice have shown that with rapid and effective rehabilitation training, approximately 90% of patients can recover a certain level of daily movement and independence in living. Clinically, the period from 0 to 6 months is called the golden recovery period, but rehabilitation is a long-term process and should continue for more than a year in most cases. Furthermore, given the very large number of stroke patients and the severe shortage of rehabilitation medical resources, there is a very high demand for smart management devices.

[0003] Current rehabilitation treatment generally involves a rehabilitation therapist assisting the patient by holding their hands or using mechanical assistive devices to treat the patient one-on-one. This method has several drawbacks: (1) it is inefficient and expensive; (2) it is subjective in its rehabilitation assessments, making it difficult to ensure the accuracy of the evaluation results of rehabilitation training; (3) the selection of training movements is not closely related to daily life, creating a certain level of difficulty for patients to return to normal social life; and (4) the position and height of the equipment are fixed, making it difficult to use for patients who have difficulty moving. [Overview of the project] [Problems that the invention aims to solve]

[0004] The embodiment of the present invention provides a smart rehabilitation device that can improve the efficiency of patient rehabilitation treatment and enhance the intelligence of the rehabilitation management device. [Means for solving the problem]

[0005] The technical solutions relating to the embodiments of this application are as follows. A smart rehabilitation device, the device comprising a mobile trolley, The mobile trolley includes a built-in host computer, a mobile unit, and a wearable unit, and both the mobile unit and the wearable unit are communicated to the built-in host computer. The embedded host computer includes a rehabilitation training module, a rehabilitation evaluation module, a data acquisition module, a data processing module, a mobility control module, and a storage module. The rehabilitation training module is designed to provide a scenario-based training game based on rehabilitation movements, and the scenario-based training game is designed to enable training of the upper limbs, lower limbs, and hands of rehabilitation patients. The rehabilitation evaluation module is for evaluating the motor function of the upper limbs, lower limbs, and hands of the rehabilitation training patient. The data collection module is for collecting information from the rehabilitation training patient, and the information includes eigenvectors corresponding to multiple dimensions of the patient, with each eigenvector representing patient information in each dimension, and the dimensions include the patient's medical history file, lifestyle, environmental factors, and rehabilitation training status. The data processing module is used to predict the degree of influence of patient information in each dimension on patient rehabilitation based on the eigenvectors of each dimension, obtain a first prediction vector corresponding to the eigenvector of each dimension, the first prediction vector representing the degree of influence of patient information in each dimension on patient rehabilitation, and to obtain a target vector by feature-combining each of the first prediction vectors; to predict the rehabilitation effect corresponding to patient information in each dimension based on the target vector, obtain a second prediction vector representing the rehabilitation effect corresponding to patient information in each dimension; and to generate prescription information for the patient's rehabilitation based on the second prediction vector, the prescription information including medication information and drug dosage. The movement control module acquires the position of the rehabilitation training patient, measures the distance from the rehabilitation training patient, and controls the mobile trolley to perform position movement and height adjustment. The movement unit includes rollers, electric push rods, and a driving device installed on the mobile trolley. The driving device is connected to the movement control module. The movement control module drives the rollers through the driving device to perform movement and position adjustment, and controls the electric push rods to perform height adjustment. The wearable unit includes a rehabilitation training glove and a somatosensory device. Both the rehabilitation training glove and the somatosensory device perform rehabilitation training by the rehabilitation training module and transmit rehabilitation training data to the data collection module.

[0006] Furthermore, the data collection module includes a receiver and a camera. The receiver is wirelessly connected to the rehabilitation training glove and the somatosensory device respectively. The receiver is for receiving the motion data of the rehabilitation training glove and the somatosensory device. The camera captures the movements of the rehabilitation training patient to realize the collection of motion data. The data processing module is further for processing the motion data by fusing multi-source information.

[0007] Furthermore, for the man-machine interaction type support column, a card reading area and a button area are installed on the man-machine interaction type support column. The card reading area is for individually identifying the rehabilitation training patient. A plurality of buttons are installed in the button area, and the buttons are for assisting in the progress of the scenario training game. For the cloud server, a motion prescription is stored in the cloud server. The embedded host computer is for downloading the motion prescription by the cloud server or uploading the rehabilitation training data. Regarding the power supply module, the power supply module includes a magnetic core, a transformer, a filter, and a data line, and is for supplying power and processing power signals.

[0008] Furthermore, the rehabilitation training glove includes a glove body and a first housing attached to the glove body. A first battery, a first circuit board, and a bending sensor are installed in the first housing, and a first indicator lamp is installed on the first housing.

[0009] Furthermore, the somatosensory device includes a band and a second housing attached to the band. A second circuit board and a second battery are installed in the second housing, and a second indicator lamp is installed on the second housing.

[0010] Furthermore, the rehabilitation evaluation module includes an evaluation assistance unit and an evaluation unit. The evaluation unit includes a smart scale evaluation sub-unit and a compensatory movement quantification evaluation sub-unit. The evaluation assistance unit is for completing the identification task of the rehabilitation training movement and obtaining the identification result. The evaluation unit is for obtaining the corresponding evaluation result based on the identification result. Furthermore, the embedded host computer realizes the interconnection of terminals via a network and is used for remote medical diagnosis or online rehabilitation training guidance.

[0011] Furthermore, a radar detector and a positioning module are connected to the movement control module. The radar detector is for detecting the position of the rehabilitation training patient, and the positioning module is for determining the current position of the mobile trolley.

[0012] Furthermore, the mobile trolley includes a pedestal and a connection block. The roller is installed at the bottom of the pedestal, the electric push rod is installed in the connection block, and the embedded host computer is connected to the top of the electric push rod.

[0013] Furthermore, an infrared sensor and an infrared signal transmitter are connected to the motion control module, and the device adjusts its position and height using the infrared sensor, the infrared signal transmitter, and the motion control module during use. [Effects of the Invention]

[0014] Compared to the prior art, the beneficial effects of the embodiments of this application are as follows:

[0015] Firstly, an embodiment of the present invention provides a smart rehabilitation device comprising a mobile cart and an embedded host computer, a mobile unit, and a wearable unit, each attached to the mobile cart, wherein the embedded host computer controls the mobile unit to realize smart position / height identification and height adjustment functions. Secondly, the embodiment of the present application provides a smart rehabilitation device, which performs rehabilitation training tasks in the form of a virtual game using a rehabilitation training module on an embedded host computer, acquires visualized quantifiable rehabilitation evaluation results using a rehabilitation evaluation module, is easy to operate, highly efficient, and highly engaging. Thirdly, the embodiment of the present application provides a smart rehabilitation device that can perform rehabilitation training on the upper limbs, lower limbs, and fingers in multiple directions using rehabilitation training gloves and a sensory device, and by identifying rehabilitation training movements and establishing scale evaluation and compensatory evaluation, it forms an accurate evaluation of rehabilitation training movements, and compared to conventional evaluation systems and devices, the evaluation results are more accurate, the evaluation model can be further learned, and it has a wide application market. Fourthly, the embodiment of the present invention provides a smart rehabilitation device that establishes rehabilitation files using a human-machine interactive support column, allows for real-time monitoring of the rehabilitation progress of all patients, and improves the user experience. Furthermore, it reduces the burden on operators through detailed management files and standardizes the entire rehabilitation training process. Fifth, the embodiment of the present invention provides a smart rehabilitation device that loads various rehabilitation training games using a rehabilitation training module and selects interesting games that are close to everyday life to assist the patient's rehabilitation training, and the rehabilitation training module can adequately train the patient's upper limbs, lower limbs and fingers, thereby accelerating the patient's rehabilitation progress. Sixth, the embodiment of the present application provides a smart rehabilitation device that, by acquiring dimensional information of the patient, can predict the effectiveness of the patient's rehabilitation and generate prescription information for the patient's rehabilitation based on the dimensional information of the patient, thereby improving the efficiency of the patient's rehabilitation. [Brief explanation of the drawing]

[0016] [Figure 1] Figure 1 is a first structural diagram of a smart rehabilitation device according to an embodiment of the present invention. [Figure 2] Figure 2 is a second structural diagram of a smart rehabilitation device according to an embodiment of the present invention. [Figure 3] Figure 3 is a third structural diagram of a smart rehabilitation device according to an embodiment of the present invention. [Figure 4] Figure 4 is a schematic diagram of a smart rehabilitation device according to an embodiment of the present invention. [Figure 5] Figure 5 is a structural diagram of a rehabilitation training glove in a smart rehabilitation device according to an embodiment of the present invention. [Figure 6] Figure 6 is a structural diagram of the sensory device in a smart rehabilitation device according to an embodiment of the present invention. [Figure 7] Figure 7 is a fourth structural diagram of a smart rehabilitation device according to an embodiment of the present invention. [Modes for carrying out the invention]

[0017] To further clarify the purpose, technical proposal, and advantages of the embodiments of this application, the technical proposal of the embodiments of this application will be clearly and completely described below with reference to the drawings of the embodiments of this application. Of course, the embodiments described below are only a part of the embodiments of this application, not all of them. In general, the components of the embodiments of this application described and shown in these drawings may be arranged and designed in a variety of different configurations.

[0018] Accordingly, the detailed description of embodiments of the present application provided below with reference to the drawings is for illustrative purposes only and does not limit the scope of the claims of the present application. Any other embodiments that a person skilled in the art can obtain based on the embodiments of the present application without requiring inventive work are all within the scope of the protection of the present application.

[0019] It should be understood that, in the description of embodiments of the present invention, terms such as "first," "second," etc., are for descriptive purposes only and should not be understood as indicating or implying relative importance, or implicitly indicating the number of technical features being referred to. Accordingly, features limited by "first," "second," etc., may explicitly or implicitly include one or more of the aforementioned features.

[0020] In describing embodiments of the present invention, unless otherwise specifically defined and limited, the terms “installation,” “mounting,” “connection,” and “connection” should be understood in a broad sense, for example, as fixed connections, removable connections, or integral connections; mechanical connections, electrical connections, or intercommunicative connections; direct connections, indirect connections via intermediate elements, internal communication between two elements, or interaction relationships between two elements. Those skilled in the art will be able to understand the specific meaning of the above terms in embodiments of the present invention depending on the specific circumstances.

[0021] Figure 1 is a block diagram of the overall structure of a smart rehabilitation device according to an embodiment of the present application. Referring to Figure 1, the embodiment of the present application provides a smart rehabilitation device. The system comprises a mobile cart 1, an embedded host computer 11 and a mobile unit 12, both attached to the mobile cart 1, and a wearable unit 2, both of which are connected to the embedded host computer 11.

[0022] The embedded host computer 11 includes a processor and memory. The processor is used to run programs for the data acquisition module 111, data processing module 112, mobility control module 113, rehabilitation training module 114, and rehabilitation evaluation module 115.

[0023] The processor may be a single general-purpose processor (CPU, central processing unit), a microprocessor, an application-specific integrated circuit (ASIC, application-specific integrated circuit), or one or more integrated circuits for executing the program of this application.

[0024] The memory may be the storage module 116 of the present invention, and the memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, a random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, an electrically erasable programmable read-only memory (EEPROM), a magnetic disk storage medium or other magnetic storage device, or any other medium that is accessible by a computer and capable of carrying or storing a desired program code having an instruction or data structure format.

[0025] As shown in Figure 2, the wearable unit 2 includes a rehabilitation training glove 201 and a sensory device 202. Both the rehabilitation training glove 201 and the sensory device 202 perform rehabilitation training using the rehabilitation training module 114 and transmit the rehabilitation training data to the data collection module 111. The wearable unit 2 achieves wireless communication with the embedded host computer 11 using ZigBee, Bluetooth®, or Wi-Fi wireless communication protocols, thereby enabling it to perform rehabilitation training based on the training content of the rehabilitation training module 114 and transmit the rehabilitation training data to the data collection module 111.

[0026] The data collection module 111 is for collecting information from the rehabilitation training patient, and the information includes eigenvectors corresponding to multiple dimensions of the patient, with each eigenvector representing patient information in each dimension, and the dimensions include the patient's medical history file, lifestyle, environmental factors, and rehabilitation training status.

[0027] A patient's medical history file includes the patient's chief complaint, present illness, past medical history, personal life history, marriage and childbirth history, family history, diagnostic records, treatment suggestions, surgical records, and pathology reports. Lifestyle includes eating habits, exercise habits, smoking history, alcohol consumption, and work breaks. Environmental factors include temperature, humidity, air dust, wind speed, ambient noise in decibels, and air quality. Rehabilitation training includes motor function, balance function, and cognitive function. Naturally, in addition to the elements of these dimensions influencing the patient's rehabilitation, other factors such as the patient's gender, address, age, education level, family circumstances, marriage and childbirth history, economic situation, occupation, job type, and work environment also affect the patient's rehabilitation. Taking lifestyle habits as an example, we establish an eigenvector for the lifestyle habit dimension (0,0,1,1,0,0…) which includes eating and drinking habits (normal:0, saltier:1, bland:2, excessive oil intake:3, etc.), whether or not one exercises (moderate exercise:0, no exercise:1, small amount of exercise:2, large amount of exercise:3, excessive exercise:4, etc.), whether or not one smokes, whether or not one drinks alcohol, and whether or not work breaks are regular.

[0028] The data processing module 112 includes a first prediction unit, a processing unit, a second prediction unit, and a generation unit. The first prediction unit predicts the degree of influence of patient information in each dimension on patient rehabilitation based on the eigenvectors of each dimension, and obtains a first prediction vector corresponding to the eigenvectors of each dimension, the first prediction vector representing the degree of influence of patient information in each dimension on patient rehabilitation. The processing unit features each of the first prediction vectors to obtain a target vector. The second prediction unit predicts the rehabilitation effect corresponding to patient information in each dimension based on the target vector, and obtains a second prediction vector, the second prediction vector representing the rehabilitation effect corresponding to patient information in each dimension. The generation unit generates prescription information for the patient's rehabilitation based on the second prediction vector, the prescription information including medication information and drug dosages.

[0029] Prescription information includes information such as the name of the medical institution, the patient's name, gender, age, outpatient or inpatient medical history number, department or ward and bed number, clinical diagnosis, and date of creation, as well as information such as the name of the drug, dosage form, specifications, quantity, and usage / dosage.

[0030] As an option, the first prediction unit specifically inputs eigenvectors corresponding to each dimension into a first prediction model corresponding to each dimension to obtain a first prediction vector corresponding to each eigenvector, the first prediction vector indicates the degree of influence of patient information in each dimension on patient rehabilitation, each of the first prediction models is an LSTM-based neural network model, and each of the LSTM-based neural network models is a model trained using multiple patient information samples and corresponding tag information.

[0031] Specifically, the second prediction unit inputs the target vector into a predetermined second prediction model to obtain a second prediction vector corresponding to each of the first prediction vectors, the second prediction vector represents the predicted result of the rehabilitation effect corresponding to the patient information of each dimension, the second prediction model is an attention mechanism-based neural network model, and the attention mechanism-based neural network model is a model trained using samples of patient information from each of the multiple dimensions that have an impact on the patient's rehabilitation and corresponding tag information.

[0032] The processing unit is further used to obtain an activation function, which includes weight coefficients for each dimension and a predetermined bias constant, and to input the second prediction vector into the activation function and use the activation function to determine at least one target sub-vector from the second prediction vector. The generation unit is further for generating prescription information for the patient's rehabilitation based on the at least one target sub-vector.

[0033] The processing unit is further used to acquire a rehabilitation prescription pattern database, to ensure that the rehabilitation prescription pattern database contains multiple prescription information, to generate each prescription information in the rehabilitation prescription pattern database as a corresponding prescription vector, to acquire a matching function and a weight vector, to ensure that the matching function is for matching the corresponding prescription vector based on the target subvector and the weight vector, and to construct a prescription generation model based on the prescription vector, the matching function, and the weight vector, where the weight vector is a predetermined parameter of the model.

[0034] The processing unit is further used to obtain predicted prescription information generated by multiple patients using a prescription generation model, and to obtain actual prescription information that was actually executed corresponding to each predicted prescription information, and to ensure that the actual prescription information includes the patient's actual medication information, to obtain the loss function of the prescription generation model, and to train the prescription generation model using each predicted prescription information and the corresponding actual prescription information until the threshold of the loss function of the prescription generation model satisfies a predetermined condition, thereby obtaining a target prescription generation model.

[0035] Specifically, the generation unit inputs the at least one target sub-vector into the target prescription generation model to obtain prescription information for the patient's rehabilitation.

[0036] Specifically, the second prediction unit may be arbitrarily selected to output predictions of the influence on rehabilitation effects, such as lifestyle habits, environmental factors, and rehabilitation training, using a multiple output control policy, and the multiple output control policy is: JPEG0007851487000001.jpg5170 However, σ is the activation function, which is initialized as the sigmod function or another function and selected according to the actual needs, x is the second prediction vector, W and V are the weight coefficients, and b and c are the bias constants.

[0037] Furthermore, the loss function of the prescription generation model is: JPEG0007851487000002.jpg28170

[0038] Furthermore, the second prediction unit may further improve the training efficiency of the model by utilizing a reward and punishment system and an error feedback method during the model training process. The reward and punishment system is JPEG0007851487000003.jpg11170 where i and j represent the dimension numbers, N(i) is the predicted prescription information vector, N(j) is the actual prescription information vector, and l is the reward / punishment constant.

[0039] The error feedback method is: JPEG0007851487000004.jpg11170 However, observed t This is actual prescription information, predicted t is the predictive prescription information, RMSE represents the system's root mean square error, N is the eigenvector dimension, and t is the prescription vector number.

[0040] Furthermore, to improve the visualization of patient rehabilitation execution management, the processing unit may also monitor the patient's execution status through compliance calculations. JPEG0007851487000005.jpg6170 However, Task shows the predicted prescription information vector, Execute shows the actual prescription information vector, and Compliance i represents the desired execution power of the i-th predictive prescription information, and n is the number of the information vector.

[0041] JPEG0007851487000006.jpg12170 However, w i This represents the weight vector of the i-th predicted prescription information. Finally, the output is the patient's execution pattern and the type of composite factor composite calculation.

[0042] In embodiments of the present invention, the first prediction unit, the processing unit, the second prediction unit, and the generation unit may each be one or more processors, controllers, or chips having a communication interface and capable of implementing a communication protocol, and may include memory and associated interfaces, a system transmission bus, etc., if necessary, and the processors, controllers, or chips realize the corresponding functions by executing code related to the program. Alternatively, an alternative solution is that the first prediction unit, the processing unit, the second prediction unit, and the generation unit share a single integrated chip, or share devices such as processors, controllers, and memory. The shared processors, controllers, or chips realize the corresponding functions by executing code related to the program.

[0043] The movement control module 113 acquires the position of the rehabilitation training patient and measures the distance to the rehabilitation training patient, thereby controlling the mobile cart 1 to move towards the rehabilitation training patient and adjusting its height to suit rehabilitation training patients of different heights, thereby obtaining a better user experience and providing great convenience for patients who have difficulty moving.

[0044] Rehabilitation training module 114 is designed to provide scenario-based training games that are based on rehabilitation movements, and these games are intended to enable rehabilitation training of the upper limbs, lower limbs, and hands of patients undergoing rehabilitation.

[0045] The rehabilitation assessment module 115 is used to evaluate the motor function of the upper limbs, lower limbs, and hands of rehabilitation training patients. After the rehabilitation training module 114 completes the rehabilitation training, the rehabilitation assessment module 115 performs an evaluation in the form of a score based on the training results.

[0046] As an option, as shown in Figure 3, the rehabilitation assessment module 115 consists of an assessment support unit and an assessment unit. The assessment support unit is for completing the task of identifying rehabilitation assessment movements. The assessment support unit is a cascade model of two parts, each containing a first classification subunit and a second classification subunit. The assessment unit consists of a smart scale assessment subunit and a compensatory movement quantification assessment subunit.

[0047] The first classification subunit uses a neural network structure and is for identifying which of the upper limb arm movement, palm movement, lower limb movement, and upper limb movement the rehabilitation evaluation movement the patient is performing belongs to. The neural network has an input that is a movement-specific vector (a vector obtained by threshold judgment of the original signal of the wearable unit), and an output that is one of the above four types of movement categories. The neural network includes one input layer, and the number of neurons in this layer is the length of the movement-specific vector, which is preset as N. The network includes two hidden layers. The number of neurons in the first hidden layer is 2N, and the number of neurons in the second hidden layer is N. The network includes one output layer, and the number of neurons is 4. Finally, the network completes the final output through a softmax layer and one-hot encoding. The network calculation method is JPEG0007851487000007.jpg35170 provided that a j is the output of the hidden layer, σ is the activation function, N is the number of neurons, i is the loop parameter, W ij is the weight, X i is the input, b j is the bias parameter, z k is the k-th output of the output layer, W jk is the weight, b k is the bias parameter, output is the final output, ont-hot is the encoding function, and softmax(z) is the activation function.

[0048] The second classification subunit uses a gated recurrent neural network, and its role is to identify specific rehabilitation evaluation movements. The input is the original signal of the wearable device for rehabilitation, and the output is the specific identification result. Its calculation method is JPEG0007851487000008.jpg3'2170 In the formula, W r 、W z and W h are weight parameters, b r 、b h and bz This is the bias parameter, which is automatically optimized by the backpropagation algorithm, h t The state is timestamp t, h t-1 This is the state at timestamp t-1, h t ~ This is the state of the timestamp after the update, g z This is the control vector for the reset gate, and g r σ is the control vector of the update gate, σ is the activation function, commonly used as the sigmoid function, and tanh is the tanh activation function.

[0049] The smart scale evaluation subunit is for completing the quantitative scoring of rehabilitation movements. This subunit takes signals from a wearable unit as input and outputs a smart quantitative score of the movement. This part is completed using a self-attention mechanism-based LSTM network. First, attention coding is performed on the multi-channel signals (x1-xn) from the wearable unit using the self-attention mechanism, and the calculation of the attention mechanism module is as follows: JPEG0007851487000009.jpg43170 However, W q This is the query matrix, and W k This is the key matrix, which is automatically learned and obtained during the network training backpropagation process, and x i and x j x are input eigenvectors at different time points, AS is the attention score, AW is the attention weight, and x j ′ is the reconstructed eigenvector at a single point in time, i and k are loop parameters, exp is the base-e exponential function, and softmax is the activation function.

[0050] After self-aware coding, the original signal (x1-x n ) is the reconstructed signal (x1'-x n ′) and then (z1-z nIt is named as ). At this time, the reconstruction signal is input to the LSTM network to complete the classification task. Assuming that the score for a single action in the original clinical scale is k-point (0, 1...k), the LSTM completes the k-classification task. The calculation of the LSTM is as follows: JPEG0007851487000010.jpg40170 However, the input gate is cell state C t (C t-1 It acts on the cell state (which is the cell state at the previous timestamp) and determines which new information to store in the cell state. The input gate consists of two parts. For the input part, it constructs the input information and determines which information to add to the unit state as new internal memory. For the candidate state part, one candidate cell state C t ~ Construct the following. After constructing these two parts, the unit state C t Update the output gate to the final output h t (h t-1 This determines the hidden state of the previous timestamp. The output is the new unit state C. t Based on this, the process is completed in two steps. The first step is to determine which part of the cell state needs to be output. The second step is to control the final output. W f , W i and W o The weight parameter is b f , b i , b c and b o x is the bias parameter, which is automatically optimized by the backpropagation algorithm. t The input eigenvector at time t is i t is the input gate, and o t σ is the output gate, σ is the activation function, which is generally used as the sigmoid function, and tanh is the tanh activation function.

[0051] The output of the LSTM passes through the softmax layer, which in turn results in the final probability output (p1-p k) can be obtained, and further classification output, i.e., the final smart score, can be completed through one-hot coding. The smart scale evaluation subunit not only provides the smart score of the scale, but also provides the scale score s after subdivision. Its calculation method (where i is the loop parameter) is: JPEG0007851487000011.jpg13170 The Compensatory Movement Quantification Evaluation subunit includes compensatory movement detection and compensatory movement quantification. In the compensatory movement detection model, an internal classifier determines the probability of identifying a compensatory movement as a type with different motor parameter characteristics (a total of 7 types), and the highest probability value and the name of the corresponding compensatory movement are taken as the detection result. If compensation is detected, the Compensatory Movement Quantification Evaluation model uses the quantitative threshold interval of the compensatory movement severity as the evaluation tag for the quantitative category, determines the quantitative threshold interval corresponding to the motor parameter characteristics based on the K-nearest neighbor algorithm, and outputs the quantitative category corresponding to the quantitative threshold interval as the compensatory movement quantification evaluation result.

[0052] The compensatory movement detection internal classifier is a support vector machine model. First, feature extraction is performed on limb movement data based on the completeness, smoothness, and variability of the target patient's limb movement test movements. Movement parameter features related to the target patient's trunk, shoulder joint, and elbow joint are selected, and movement eigenvectors are obtained. To detect the seven types of compensatory movements present in the target patient, seven binary classification models are constructed and used as representatives of the compensatory movements in each category. Finally, the output results of the seven SVM binary classifiers are integrated to achieve detection and classification of multiple types of compensatory movements. That is, each binary classification model takes an eigenvector as input and outputs probabilities p0 and p1 that classify the movement as 0 or 1. If p1 is greater than p0, the largest p1 output from the seven models is the final compensatory movement type. If p1 is never greater than p0, no compensation exists.

[0053] The compensatory movement quantification evaluation model uses the quantification threshold interval of compensatory movement severity as an evaluation tag for the quantification category, determines the quantification threshold interval corresponding to the motor parameter characteristics based on the K-nearest neighbor algorithm, and outputs the quantification category corresponding to the quantification threshold interval as the compensatory movement quantification evaluation result.

[0054] Specifically, the quantitative threshold interval for compensatory movement severity is determined by dividing the same compensatory movement into K levels according to the difference in its severity, and the division range for each severity level is the quantitative threshold interval for compensatory movement. The quantitative threshold intervals for mild, moderate, and severe compensation are determined by performing clustering analysis on multiple motor parameter features of stroke patients with such compensatory movements using the K-clustering algorithm. Clustering analysis is performed on a large sample data using the K-mean clustering algorithm to obtain tags for the compensatory movement severity division intervals, and K-level compensatory movement severity evaluation thresholds are obtained by parameter setting, with K being 3 or greater. When K is equal to 3, the quantitative threshold intervals corresponding to mild, moderate, and severe compensation are the first type of threshold interval, the second type of threshold interval, and the third type of threshold interval, respectively. Subsequently, the data is divided into training and test sets using a supervised machine learning algorithm, and the motor parameter features and multi-level evaluation tags are used as input to a compensatory movement quantification evaluation model, and a K-nearest neighbor machine learning algorithm is used to realize quantitative evaluation of compensatory movement severity based on multi-level threshold tags.

[0055] In embodiments of the present invention, the first prediction unit, the processing unit, the second prediction unit, the generation unit, the evaluation support unit, the evaluation unit, the smart scale evaluation subunit, the compensatory motion quantification evaluation subunit, the first classification subunit, and the second classification subunit each have a communication interface and may be one or more processors, controllers, or chips capable of implementing a communication protocol, and may further include memory and related interfaces, a system transmission bus, etc., if necessary, and the code relating to the processor, controller, or chip execution program realizes the corresponding function. Alternatively, an alternative solution is that the first prediction unit, the processing unit, the second prediction unit, the generation unit, the evaluation support unit, the evaluation unit, the smart scale evaluation subunit, the compensatory motion quantification evaluation subunit, the first classification subunit, and the second classification subunit share a single integrated chip, or share devices such as a processor, controller, and memory. The shared processor, controller, or chip realizes the corresponding function by executing the code relating to the program.

[0056] As shown in Figure 4, the mobile unit 12 includes rollers 123, an electric push rod 122, and a drive unit 121 installed on the mobile trolley 1, with the drive unit 121 connected to the mobile control module 113. The mobile control module 113 includes a controller that controls the drive unit 121 to drive the rollers 123 to move and controls the electric push rod 122 to extend and retract, thereby enabling height adjustment relative to the embedded host computer 11 and adapting its height to the current rehabilitation training patient.

[0057] In this embodiment, the embedded host computer 11 is an SK-27A type all-in-one personal computer, and includes a 27-inch capacitive touch panel, an Intel Core i5-3210 CPU, Longsys 4G memory, and a Longsys 128G solid-state drive, as well as a power adapter, connecting cables, and a stylus. The all-in-one personal computer is embedded and mounted on top of the mobile trolley 1 body, and the mobile trolley 1 further includes a base 101 and a connecting block 102, as shown in Figure 2. A roller 123 is installed at the bottom of the base 101, and an electric push rod 122 is installed inside the connecting block 102. The embedded host computer 11 is connected to the output terminal of the electric push rod 122, thereby enabling the height of the embedded host computer 11 to be controlled by the electric push rod 122. The electric push rod is a TJC-C4 type electric push rod.

[0058] It should be noted that in this embodiment, the rehabilitation training module 114 and the rehabilitation evaluation module 115 constitute the rehabilitation training and evaluation system. After logging into the rehabilitation training and evaluation system, the user may select rehabilitation training and evaluation content as needed. If rehabilitation training content is selected, the user must first put on the rehabilitation training gloves 201 and the sensory device 202, and then select the rehabilitation training function based on the display window. Since many rehabilitation training games are loaded into the rehabilitation training function, the rehabilitation trainee may select a difficulty level corresponding to their own situation. For patients using the device for the first time, an explanatory video is also loaded into the system. The rehabilitation trainee may learn how to use the device based on the video, which helps the user to use the device better and further improves the user experience. The rehabilitation training games are selected to be interesting games that are close to everyday life, and the games can sufficiently train the patient's upper limbs, lower limbs and fingers, accelerating the patient's rehabilitation progress.

[0059] As shown in Figure 5, the rehabilitation training glove 201 includes a glove body and a first housing attached to the glove body. A first battery, a first circuit board, and a bending sensor are installed in the first housing, and a first indicator lamp is installed in the first housing. In this embodiment, the glove body is made of nylon material, the opening of the glove is elastic to improve comfort of use, and the glove comes in multiple different types to be applicable to many people. The first battery is a selected 602035 type lithium battery, which provides power support to the remaining components on the glove. The bending sensor is an FS-L-0055-253-ST bending sensor, which measures relevant data when a rehabilitation patient is performing training. The first circuit board transmits the bending sensor data to the data acquisition module 111 via an integrated wireless transmission module. Simultaneously, the first circuit board also controls the flashing of the first indicator lamp. When the channel device corresponding to the rehabilitation training glove 201 is functioning normally, the first circuit board controls the indicator lamp to green (or flashing). If there is no data in the corresponding channel device, the indicator lamp controls the indicator lamp to gray. If data reception by the corresponding channel device is abnormal, the indicator lamp controls the indicator lamp to red.

[0060] As shown in Figure 6, the sensory device 202 includes a band and a second housing attached to the band, the second circuit board and second battery installed in the second housing, and the second indicator lamp installed in the second housing. In this embodiment, there may be multiple sensory devices 202, mainly used for collecting upper and lower limb movement data. Different dimensions of bands may be selected and used as needed, for fixing the second housing and the second circuit board, second battery, etc. inside it in corresponding positions. The second battery may be of the same type as the first battery and is for providing power support. A corresponding sensor, such as a posture sensor, may be installed in the second housing and connected to the second circuit board as needed, the second circuit board is for controlling the flashing of the second indicator lamp and for transmitting sensor data to the data acquisition module 111 via an integrated wireless transmission module.

[0061] As can be understood, in this embodiment, in addition to the rehabilitation training gloves 201 and the sensory device 202, other wearable devices or rehabilitation training devices such as walking machines and rehabilitation training helmets may also be used.

[0062] Preferably, a receiver 1111 and a camera 1112 are connected to the data acquisition module 111. Both receivers 1111 are wirelessly connected to the rehabilitation training gloves 201 and the sensory device 202, and are used to receive movement data from devices such as the rehabilitation training gloves 201 and the sensory device 202. The camera 1112 collects movement data by capturing the movements of the rehabilitation training patient. In this embodiment, the camera 1112 may be a camera 1112 attached to an integrated PC, or it may be a separately connected camera 1112. The type of camera 1112 is not particularly required, as long as it is compatible with the integrated PC and the rehabilitation training and evaluation system.

[0063] Preferably, in this embodiment, the all-in-one PC may be connected to a cloud server via a network, the exercise prescription may be stored on the cloud server, and the exercise prescription may be downloaded or training data may be uploaded from the cloud server. In actual application, the rehabilitation trainee may also be connected to a terminal device such as a hospital via the cloud server, enabling remote online guidance and diagnosis by a doctor.

[0064] As shown in Figure 4, the device further includes a human-machine interactive support column 13, which is connected to an integrated PC by wire or wireless means. The human-machine interactive support column 13 is equipped with a card reading area and a button area. The card reading area is for individual identification of rehabilitation training patients, and the button area is equipped with multiple buttons. These buttons assist in the progress of the scenario training game. Before performing rehabilitation training using the device, rehabilitation training patients must first swipe their ID card or identification card into the card reading area. The integrated PC performs personal authentication using the ID card or identification card. Then, the rehabilitation training patient can enter the rehabilitation training and evaluation system and perform the rehabilitation training game by pressing the buttons in the button area. As can be understood, a rehabilitation training patient management system can be established based on the human-machine interactive support column 13, and files can be established for each rehabilitation training and evaluation session of the patient. This allows for tracking of rehabilitation progress at any time, improving the user experience, and also reduces the burden on operators through detailed management files, making the entire rehabilitation training process more standardized.

[0065] Preferably, the device further comprises a power module 14, which includes a magnetic core, transformer, filter, and data lines, and is for supplying power and processing power signals. The magnetic core is of type V18004, V18005, or V18007 and is used to achieve electromagnetic shielding. The filter is of type RSEN-2006L EMC filter and is used to process power signals. The transformer is of type H0128-823-0250 isolation transformer. The magnetic core, transformer, and filter are all installed in empty cavities within the base 101, thereby increasing the weight of the base 101 and making it easier to maintain stability during the movement of the embedded host computer 11 connected to it. Each component is connected to the power supply and to the embedded host computer 11 as needed, and a detailed explanation is omitted here.

[0066] As shown in Figure 7, a radar detector is connected to the mobile control module 113, which is used to locate the rehabilitation patient, and the mobile control module 113 controls the drive unit 121 and the electric push rod 122 to operate. In this embodiment, the radar detector may be a SENKYLASER SK-C10 2D laser scanning radar, and the mobile control module 113 uses a Gmapping algorithm to achieve autonomous positioning and map construction. As can be understood, the device further includes a GPS positioning module to determine the current position of the mobile cart 1, then the radar detector scans the surrounding environment to determine the position of the rehabilitation patient, and finally the mobile control module 113 controls the rollers 123 to move toward the rehabilitation patient by the drive unit 121. In this embodiment, there are four rollers 123, which are installed circumferentially at the bottom of the mobile carriage 1. The drive unit 121 uses motors, and there are two motors in total. One motor is connected to two rollers 123, and the controller controls the forward and reverse rotation of this motor to drive and rotate the rollers 123, thereby enabling the mobile carriage 1 to move forward and backward. The other motor is connected to two other rollers 123, and the controller controls the forward and reverse rotation of this other motor to drive and rotate the other two rollers 123, thereby enabling the mobile carriage 1 to turn left and right. The method of controlling the forward, backward, left and right movement of the device using motors is conventional technology and is similar to the method of controlling the forward, backward, left and right movement of a toy car using a steering wheel. The controller may be a remote control, and the remote control and motor are connected via WIFI. The remote control has forward, backward, left and right buttons, and the remote control enables control of the forward, backward, left and right movement of the device.

[0067] What needs to be explained is that the drive unit 121 uses a servo motor as the drive unit, and its model may be from the HC-KFS, HC-MFS, HC-SFS, HC-RFS, and HC-UFS series. The output terminal of the drive unit 121 is connected to the roller 123, and the drive unit 121 is controlled by the movement control module 113, thereby realizing the movement of the trolley 1. To better realize the forward, backward, left, and right movement of the trolley 1, the number of rollers 123 is set to four and the number of drive units 121 is set to two. The four rollers 123 are installed at the bottom of the base 101, two rollers 123 are connected to one drive unit 121, and the forward and reverse rotation of the drive unit 121 is controlled to realize the forward and backward movement of the trolley 1, and two other rollers 123 are connected to the other drive unit 121, and the forward and reverse rotation of this drive unit 121 is controlled to realize the left and right turns of the trolley 1.

[0068] Since different users have different heights, to enable automatic height adjustment of the device, one infrared sensor connected to the movement control module 113 is installed on top of the mobile trolley 1, and the patient simply wears the corresponding infrared signal transmitter when using the device. When use begins, the movement control module 113 controls the electric push rod 122 to operate within a certain range, moving the embedded host computer 11 within a certain height range. When the infrared sensor and the infrared signal transmitter are at the same height, the infrared sensor receives an infrared signal transmitted from the infrared signal transmitter, and at this time, the movement control module 113 controls the electric push rod 122 to stop operation, i.e., to obtain the height most suitable for the current patient.

[0069] In other embodiments, the device can achieve manual height adjustment. The lifting and lowering program for the electric push rod 122 is integrated into the movement control module 113, and by expanding the lifting and lowering control screen for the electric push rod 122 in the movement control module 113 and clicking the lift or lower option, the built-in host computer 11 can be raised or lowered.

[0070] In other embodiments, the movement control module 113 includes a remote control, which is connected to an embedded host computer 11 via Wi-Fi, and the corresponding raise or lower buttons on the remote control enable control of the motorized push rod 122 of the movement control module 113, thereby enabling the raised or lowered movement of the embedded host computer 11.

[0071] In this embodiment, the data acquisition module 111 collects the original limb movement signals of the rehabilitation training patient using the receiver 1111 and camera 1112, and transmits them to the data processing module 112 for processing. The data processing module 112 first preprocesses the limb movement signals to remove interference and influence of system noise, and then needs to extract features from the original limb movement signals. In this embodiment, it attempts to obtain one-dimensional eigenvectors of the rehabilitation training movements by extracting features from the perspective of the completeness and smoothness of the patient's movement completion. Finally, it needs to quantify and evaluate the multi-dimensional features using a multi-source fusion algorithm, and based on the results, provide driving (control) and human-machine interaction feedback to the tasks and characters in the virtual reality scene.

[0072] The rehabilitation assessment module 115 establishes an action recognition model for a wearable device based on a sequence network, and after the patient completes the rehabilitation training movement, the quality of the training movement needs to be evaluated. After obtaining one-dimensional eigenvectors of the rehabilitation training movement processed by the data acquisition module 111, a training movement assessment model is established using an attention mechanism-based LSTM scale assessment network and a machine learning-based compensation detection network. In the movement assessment, the process is predicted, i.e., the process is evaluated (giving an action score and a compensation score), and this data is placed in the training set to update the training model and give the model a self-learning function. Visualized quantified assessment parameters are designed on the model using nonlinear dynamics tools such as Poincare difference dispersion plots, maximized Lyapunov exponents, and distributed entropy, thereby enabling the quantified assessment model to provide visualized quantified scores for the training movement.

[0073] The above description merely represents the optimal specific embodiment of the present application; however, the scope of protection of the present application is not limited thereto, and any changes or substitutions within the technical scope disclosed herein should also be included within the scope of protection. Accordingly, the scope of protection of the present application should be in accordance with the claims described above. [Explanation of Symbols]

[0074] 1 Mobile cart 101 Pedestal 102 Connection Blocks 11 Embedded host computer 111 Data Acquisition Module 1111 Receiver 1112 Camera 112 Data Processing Modules 113. Movement control module 114 Rehabilitation Training Modules 115 Rehabilitation Assessment Module 116 Memory Modules 12 Mobile Units 121 Drive unit 122 Electric Push Rod 123 Laura 2 Wearable Units 201 Rehabilitation Training Gloves 202 Sensory Device 13. Human-machine interactive support column 14 Power Modules

Claims

1. A device for use in rehabilitation training and evaluation, The system comprises a mobile cart, an embedded host computer attached to the mobile cart, a mobile unit, and a wearable unit, wherein both the mobile unit and the wearable unit are connected to the embedded host computer. The aforementioned embedded host computer is equipped with a rehabilitation training module, a rehabilitation evaluation module, a data acquisition module, a data processing module, a mobility control module, and a memory module. The rehabilitation training module is designed to provide a scenario-based training game based on rehabilitation movements, and the scenario-based training game is designed to enable training of the upper limbs, lower limbs, and hands of rehabilitation patients. The rehabilitation evaluation module is for evaluating the motor function of the upper limbs, lower limbs, and hands of the rehabilitation training patient. The aforementioned data collection module is for collecting training data from the rehabilitation training patient, The aforementioned data processing module is for processing the training data, The movement control module acquires the position of the rehabilitation training patient and measures the distance to the rehabilitation training patient, thereby controlling the mobile cart to move its position and adjust its height. The moving unit includes rollers, an electric push rod, and a drive unit installed on the moving trolley, the drive unit being connected to the movement control module, the movement control module controlling the electric push rod to move and adjust the position of the rollers by the drive unit, and to adjust the height. The wearable unit includes a rehabilitation training glove and a sensory device, and both the rehabilitation training glove and the sensory device perform rehabilitation training using the rehabilitation training module, and transmit rehabilitation training data to the data collection module. This device is used for rehabilitation training and evaluation.

2. A smart rehabilitation device, wherein the device includes a mobile trolley, A built-in host computer, a mobile unit, and a wearable unit are installed on the mobile trolley, and both the mobile unit and the wearable unit are connected to the built-in host computer via communication. The embedded host computer includes a rehabilitation training module, a rehabilitation evaluation module, a data acquisition module, a data processing module, a mobility control module, and a storage module. The rehabilitation training module is designed to provide a scenario-based training game based on rehabilitation movements, and the scenario-based training game is designed to enable training of the upper limbs, lower limbs, and hands of rehabilitation patients. The rehabilitation evaluation module is for evaluating the motor function of the upper limbs, lower limbs, and hands of the rehabilitation training patient. The data collection module is for collecting information from the rehabilitation training patient, and the information includes eigenvectors corresponding to multiple dimensions of the patient, with each eigenvector representing patient information in each dimension, and the dimensions include the patient's medical history file, lifestyle, environmental factors, and rehabilitation training status. The data processing module is used to predict the degree of influence of patient information in each dimension on patient rehabilitation based on the eigenvectors of each dimension, obtain a first prediction vector corresponding to the eigenvector of each dimension, the first prediction vector representing the degree of influence of patient information in each dimension on patient rehabilitation, and to obtain a target vector by feature-combining each of the first prediction vectors; to predict the rehabilitation effect corresponding to patient information in each dimension based on the target vector, obtain a second prediction vector representing the rehabilitation effect corresponding to patient information in each dimension; and to generate prescription information for the patient's rehabilitation based on the second prediction vector, the prescription information including medication information and drug dosage. The movement control module acquires the position of the rehabilitation training patient and measures the distance to the rehabilitation training patient, thereby controlling the mobile cart to move its position and adjust its height. The moving unit includes rollers, an electric push rod, and a drive unit installed on the moving trolley, the drive unit being connected to the movement control module, the movement control module controlling the electric push rod to move and adjust the position of the rollers by the drive unit, and to adjust the height. The smart rehabilitation device is characterized in that the wearable unit includes a rehabilitation training glove and a sensory device, and both the rehabilitation training glove and the sensory device perform rehabilitation training using the rehabilitation training module and transmit rehabilitation training data to the data collection module.

3. The data acquisition module includes a receiver and a camera, and the receiver is wirelessly connected to the rehabilitation training glove and the sensory device, respectively. The receiver is for receiving the exercise data of the rehabilitation training gloves and the sensory device, and the camera collects the exercise data by capturing the movements of the rehabilitation training patient. The apparatus according to claim 2, wherein the data processing module is further for processing the motion data by fusing multi-source information.

4. The aforementioned device further, A man-machine interactive support column is provided, wherein a card reading area and a button area are installed, the card reading area being for individual identification of the rehabilitation training patient, and the button area is provided with multiple buttons, the buttons being for assisting the progress of the scenario training game. A power supply module comprising a magnetic core, transformer, filter, and data lines, for supplying power and processing power signals, The aforementioned embedded host computer is further connected to a cloud server. The exercise prescription is stored in the aforementioned cloud server. The apparatus according to claim 2, further characterized in that the embedded host computer is for downloading the exercise prescription from the cloud server or uploading the rehabilitation training data.

5. The rehabilitation training glove includes a glove body and a first housing attached to the glove body. The apparatus according to claim 2, characterized in that a first battery, a first circuit board, and a bending sensor are installed in the first housing, and a first indicator lamp is installed in the first housing.

6. The tactile device includes a band and a second housing attached to the band. The apparatus according to claim 2, characterized in that a second circuit board and a second battery are installed in the second housing, and a second indicator lamp is installed in the second housing.

7. The rehabilitation evaluation module includes an evaluation support unit and an evaluation unit, the evaluation unit includes a smart scale evaluation subunit and a compensatory movement quantification evaluation subunit. The aforementioned evaluation support unit is for completing the task of identifying rehabilitation evaluation movements and obtaining the identification results. The apparatus according to claim 2, characterized in that the evaluation unit is for obtaining a corresponding evaluation result based on the identification result.

8. The apparatus according to claim 2, characterized in that the embedded host computer enables interconnection of terminals via a network and is used for remote medical diagnosis or online rehabilitation training instruction.

9. The aforementioned movement control module is connected to a radar detector and a positioning module. The aforementioned radar detector is for detecting the location of the rehabilitation training patient, The apparatus according to claim 2, characterized in that the positioning module is for determining the current position of the mobile cart.

10. The aforementioned mobile trolley includes a base and a connecting block, The apparatus according to any one of claims 2 to 9, characterized in that the roller is installed at the bottom of the base, the electric push rod is installed in the connecting block, and the built-in host computer is connected to the top of the electric push rod.

11. The apparatus according to claim 2, wherein an infrared sensor and an infrared signal transmitter are connected to the motion control module, and the apparatus is characterized in that, when in use, the position and height are adjusted by the infrared sensor, the infrared signal transmitter and the motion control module.

Images