Evaluation Method of the Patient's Upper Limb Recovery Evaluation System Based on Inertial Information

Abstract

The present invention proposes an evaluation method for a patient's upper limb recovery evaluation system based on inertial information. It includes multiple flexible wearable nodes, a data aggregation node, and a data processing terminal. The flexible wearable node is composed of a flexible circuit board, an inertial sensor module, a processor module, a wireless transceiver module, a power supply module, and a housing installed thereon. The data aggregation node is composed of a wireless transceiver module and a serial port transmission module. The multiple flexible wearable nodes send sensor data to the data aggregation node, and the data aggregation node sends the sensor data to the data processing terminal. The data processing terminal processes the sensor data. The method identifies upper limb movements based on a support vector machine and detects symptoms such as twitching and jitter, so as to realize the identification and evaluation of the movement state and continuity of the patient's upper limb. The present invention can detect the rehabilitation movement of the patient's upper limb, guide the adjustment of the rehabilitation training strategy, and improve the rehabilitation effect.

Description

Technical Field

[0001] The present invention relates to the cross - field of instrument science, sensor technology, computer science, human health monitoring technology, and wearable instrument devices, and specifically to an evaluation method for a patient's upper limb recovery evaluation system based on inertial information. Background Art

[0002] The patient upper limb movement recovery evaluation system and method are of great significance. At present, the rehabilitation training instruments in medical institutions are expensive, bulky, and require professional technology for assistance. The rehabilitation evaluation of patients needs to be carried out by professionals, resulting in a slow rehabilitation process and limited effects for some patients. Moreover, some patients' upper limbs will have slight twitching and shaking symptoms, which are difficult to observe with the naked eye. In social life, stroke and sports injury events occur from time to time. After a stroke, patients generally show symptoms such as upper limb weakness and hemiplegia, while sports injuries can cause damage to the upper limb muscles and bones of patients. The damage to the upper limb requires patients to undergo certain upper limb rehabilitation training after receiving normal treatment to avoid osteoporosis, muscle atrophy, and gradual decline in physical fitness caused by long - term bed rest. More importantly, it will miss the good opportunity for functional rehabilitation, making the recovery of limb function unable to reach the best state. And the rehabilitation movements made by patients may be very different from those in the normal state. During the recovery process of the patient's upper limb, there may also be involuntary shaking and twitching of the upper limb that are difficult to observe, resulting in the patient being unable to obtain good rehabilitation effects. In addition, during the upper limb movement in rehabilitation training, it is impossible to perform continuous movements like normal people. The jitter amplitude is calculated through acceleration and combined with the jitter amplitude threshold to evaluate the continuity of the patient's upper limb movement. Therefore, supervising the patient's upper limb movement and evaluating its recovery effect are convenient for patients to take more effective measures in subsequent rehabilitation training.

[0003] With the development of materials science, instrument science, computer science, sensor technology, wireless communication technology, etc., wearable medical instrument devices have developed rapidly in recent years. Therefore, applying these emerging technologies to research and develop a portable instrument device that can provide support for patient upper limb movement recognition is one of the current tasks for those skilled in the art. Summary of the Invention

[0004] Aiming at the problems of a large number of stroke and sports injury patients and unsatisfactory rehabilitation effects, the present invention proposes an evaluation method for a patient's upper limb recovery evaluation system based on inertial information, so as to be able to monitor the accuracy, type, and recovery level of human upper limb movement in real - time, and achieve the purpose of improving the patient's rehabilitation effect.

[0005] To achieve the above - mentioned purpose, the technical solution adopted by the present invention is:

[0006] The present invention provides a patient upper limb recovery evaluation system based on inertial information. The evaluation system supporting equipment includes at least three wearable nodes respectively worn on the forearms and upper arms of the patient. The wearable node includes a protective layer and the overall structure is a wristband. The hardware of the evaluation system is arranged between two protective layers, and is characterized in that: the evaluation system includes an inertial sensor module, a processor module, a wireless transceiver module and a power module. The inertial sensor module, the processor module, the wireless transceiver module and the power module are integrated in the hardware of the evaluation system and are connected through a flexible circuit board;

[0007] The power module supplies power to the entire system for operation;

[0008] The inertial sensor module includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer to collect the movement information of the patient's arm;

[0009] The processor module completes data acquisition work by connecting to the inertial sensor module. The processor module receives data from the inertial sensor and sends it to the wireless transceiver module through a serial port;

[0010] The wireless transceiver module is the link between the flexible wearable node and the data aggregation node, and sends data information to the data aggregation node. The data aggregation node consists of a wireless communication module and a serial port transmission module. The wireless communication module of the data aggregation node will receive the data of the flexible wearable node; the serial port transmission module will be connected to the wireless communication module and the data processing terminal respectively; communicate with the wireless communication module and transmit the data of the wireless communication module to the data processing terminal;

[0011] The data processing terminal receives the data packet {addr, a 1 , a 2 , a 3 , …, a n} from the data aggregation node, and unpacks, parses and preprocesses the data to obtain available data including Euler angles and three-axis acceleration {roll, yaw, pitch, a x , a y , a z}, and the specific steps are as follows:

[0012] (1) Collect user inertial information including: roll angle roll, pitch angle pitch, heading angle yaw, x-axis output a x , y-axis acceleration output a y , z-axis acceleration output a z ;

[0013] (2) Preprocess the three-axis acceleration output to filter out the gravity component and eliminate the tiny data jitter;

[0014] (3) Collect the attitude angle data {roll, pitch, yaw} and the three-axis acceleration data {a x , a y , a z} to train the human upper limb motion recognition model and the twitch and jitter recognition models;

[0015] (4) According to the human upper limb motion recognition, judge the upper limb movements of the patient through the attitude angle data {roll, yaw, pitch};

[0016] (5) According to the twitch and jitter recognition model, judge whether there are jitters and twitches during the user's rehabilitation training process through the acceleration data {a x , a y , a z};

[0017] (6) Evaluate the upper limb movement continuity level.

[0018] As a further improvement of the evaluation method of the present invention, in the step (1), collect the user's inertial information, and the specific steps are:

[0019] (1.1) Use the inertial module to collect the patient's inertial information. The inertial module transmits the inertial information to the processor module, the processor module sends the inertial information to the wireless transceiver module, the wireless transceiver module sends the inertial information to the data aggregation node one by one in byte form, and the data aggregation node transmits the data to the data processing terminal. The data processing terminal converts the byte string into {roll, yaw, pitch, a x , a y , a z}.

[0020] As a further improvement of the evaluation method of the present invention, in the step (2), preprocess the three-axis acceleration output to filter out the gravity component and eliminate the small data jitter;

[0021] (2.1) Receive the acceleration data {a x , a y , a z} transmitted from the wearable node, and use the first-order low-pass filter Y(n) = αX(n) + (1 - α)Y(n - 1) to obtain the components of the gravitational acceleration on the three axes {g x , g y , g z}, and subtract {g x , g y , g z} from the original data {a x , g y , g z}The three-axis linear acceleration data {a'} with the gravity component filtered out is obtained. x , a' y , a' z};

[0022] (2.2) Perform mean filtering on the three-axis linear acceleration data. Set its filtering window to L, move the filtering window to perform mean filtering on the data, eliminate minor data jitters, and smooth the data to obtain {a (2) x , a (2) y , a (2) z}.

[0023] As a further improvement of the evaluation method of the present invention, in step (3), the attitude angle data {roll, pitch, yaw} and the three-axis acceleration data {a x , a y , a z} are collected to train the human upper limb motion recognition model and the twitch and jitter recognition models. The specific steps are as follows:

[0024] (3.1) A normal person wears a wearable node to collect the attitude angles {roll, pitch, yaw} that rotate 360° under specific actions, avoiding coordinate transformation, and constructing a motion information data set. Map the non-linear {roll i , pitch i , yaw i} to a high-dimensional space, where i = 1, 2,..., n, and use the Gaussian kernel function K(x i , x j ) = e^[-||x 1 - x 2 || 2 / (2σ 2 )] to find the inner product in the high-dimensional space, thereby training the human upper limb motion recognition model;

[0025] (3.2) Normal people and patients wear wearable nodes respectively to collect the acceleration data {a (2) xi , a (2) yi , a (2) zi} during their uniform and stable upper limb movements, where i = 1, 2,..., n, and establish an acceleration data set A. Construct a sliding window with a length of L and slide it in the data set A. For the data within the window, use the FFT transform to obtain the discrete spectrum of the sequence and extract its maximum component frequency {f xmax , f ymax , f zmax}, when the sliding window slides to the end of the dataset, the maximum component frequency {f is obtained (j) xmax , f (j) ymax , f (j) zmax}, j = 1, 2, …, n - L, establish the maximum component frequency dataset F, and map the non-linear

[0026] {f (j) xmax , f (j) ymax , f (j) zmax} to a high-dimensional space, and use the Gaussian kernel function K(x i , x j ) = e^[-||x 1 - x 2 || 2 / (2σ 2 )] to calculate the inner product in the high-dimensional space, thereby training the twitch and jitter recognition model.

[0027] As a further improvement of the evaluation method of the present invention, in step (4), according to the support vector machine classifier, the upper limb movement of the patient is judged through the attitude angle data {roll, pitch, yaw}, and the specific steps are as follows:

[0028] (4.1) Real-time collect the patient's attitude angle information {roll, pitch, yaw}, transmit it to the data processing terminal, and the data processing terminal calls the support vector machine classifier model for action recognition.

[0029] As a further improvement of the evaluation method of the present invention, in step (5), according to the support vector machine classifier, whether there is jitter and twitch during the user's rehabilitation training is judged through the acceleration data {a x , a y , a z} The specific steps are as follows:

[0030] (5.1) Real-time obtain the patient's three-axis acceleration information {a (2) x , a (2) y , a (2) z}, perform FFT transformation on the data within the sliding window with a length of L to obtain {f xmax , f ymax , f zmax}, call the twitch and jitter recognition model to determine whether there is twitch and jitter, and record the position of the first element of the sliding window when the model determines that there is twitch and jitter as the serial number of the sliding window, denoted as Lk , where k = 1, 2, j, t.

[0031] As a further improvement of the evaluation method of the present invention, in the step (6), according to the evaluation of the upper limb movement continuity level, the specific steps are as follows:

[0032] (6.1) Preset the jitter amplitude threshold {α 0 , α 1 , α 2 , α 3}, where α 0 < α 1 < α 2 < α 3 ;

[0033] (6.2) When the patient has convulsions and jitters, calculate the convulsion and jitter amplitude of the patient's upper limb. Based on the sampling frequency f k obtain the sampling time interval Δt, and use integration to obtain the jitter amplitude on the three axes and synthesize it as d s , and in the discrete space, the integration of acceleration is discretely approximated as: k

[0034]

[0035] Solve for the maximum jitter amplitude d max of the patient during the entire rehabilitation training process, where d k = max(d ) and k = 1, 2,..., t;

[0036] (6.3) Determine the four levels of excellent, good, medium, and poor according to the comparison between the jitter amplitude threshold {α 0 , α 1 , α 2 , α 3} and d max :

[0037] Excellent: d max < α 0

[0038] Good: α 0 <= d max < α 1

[0039] Medium: α 1 <= d max < α 2

[0040] Poor: d max >= α 2

[0041] Finally, information such as the accuracy of the patient's upper limb movement, the presence of tremors, convulsions, and the level of continuity of the upper limb movement is displayed on the data processing terminal.

[0042] Adopting the technical solution of the present invention will have the following beneficial effects:

[0043] (1) A patient upper limb recovery evaluation system based on inertial information disclosed by the present invention has the advantages of small volume, light weight, simple structure, convenient use, etc., and can be applied to stroke and sports injury patients for upper limb movement recognition and rehabilitation effect evaluation.

[0044] (2) The system of the present invention integrates multi-dimensional data of the human upper limb, performs real-time human motion posture recognition, tremor and convulsion recognition, and can judge the continuity of the patient's upper limb movement, which can help users correct actions in time during rehabilitation training, reflect the state of the patient's upper limb, facilitate targeted rehabilitation training, and improve the rehabilitation training effect.

[0045] (3) The patient upper limb motion recognition method of the present invention can avoid coordinate transformation when using the attitude angle, reducing the calculation. Brief Description of the Drawings

[0046] Figure 1 is a schematic diagram of the composition of a patient upper limb recovery evaluation system based on inertial information of the present invention.

[0047] Figure 2 is a flowchart of model construction of a patient upper limb recovery evaluation method based on inertial information of the present invention.

[0048] Figure 3 is an example diagram of an embodiment of a flexible wearable node of the present invention.

[0049] Figure 4 is an example diagram of an embodiment of a flexible wearable node worn on the left and right upper arms of a patient's upper limb of the present invention.

[0050] Figure 5 is an example diagram of an embodiment of a flexible wearable node worn by a patient with both arms raised high of the present invention.

[0051] Figure 6 is an example diagram of an embodiment of a flexible wearable node worn on the right hand of a patient of the present invention.

[0052] Figure 7 is an example diagram of a sliding window in a patient upper limb recovery evaluation method of the present invention.

[0053] Figure 8 is an example diagram of an embodiment of a flexible wearable node worn by a patient with the right hand raised high and the left finger touching the nose of the present invention.

[0054] Figure 9This is an exemplary diagram of an embodiment of the flexible wearable node worn by a patient with both hands extended horizontally to the sides in the present invention. Detailed implementation manners

[0055] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners:

[0056] The present invention discloses a patient upper limb recovery evaluation system based on inertial information. The patient upper limb recovery evaluation system based on inertial information is as Figure 1 shown, and includes a plurality of wearable nodes respectively worn on the forearms and upper arms of both hands of the patient. The wearable node includes a protective layer, which is composed of two layers of elastic materials, stitched together between the two layers, and the main circuit is wrapped therein, playing a dual protection role of protecting the circuit and protecting the human body from directly contacting the circuit, and a switch and a charging opening are reserved.

[0057] Here, the protective layer can be made of cloth, rubber, resin, etc., and its overall structure can be regarded as a wristband.

[0058] The hardware modules are all placed between the two layers of the protective layer, and are integrated with an inertial sensor module, a processor module, a wireless transceiver module and a power module. These four modules are all highly integrated, with the characteristics of small size, light weight and convenient installation. At the same time, they are fixed to the upper and lower layers of the protective layer and cannot move or deform. These four modules complete the main functions of the hardware device.

[0059] The flexible circuit board is responsible for connecting the four hardware function modules. The whole uses flexible plates, and a certain length is reserved by folding. Therefore, when wearing the device, a certain force can be applied from the inside to the outside, so as to expand the elastic material of the device, and thus it can be worn on the arm. During this process, due to the elasticity of the upper and lower protective layers, and the bendability of the flexible circuit board and the reserved length is sufficient for deformation, the main function modules will not be damaged.

[0060] After the device is worn, it is subjected to elastic action and tightens inward against the arm. It not only plays a stabilizing role and is not easy to fall off, but also can improve the measurement accuracy. To ensure that it can be used by different people, a variety of materials can be used to set up devices with a variety of elastic specifications, so as to ensure that the overall size can fit the upper and lower arms of the human body, which is both stable and practical and will not exert strong pressure on the arm.

[0061] The power module plays a role in supplying power for the whole system to work. To ensure convenient use and simple operation, we adopt a design of lithium battery plus integrated charge and discharge. A switch is designed to control the working mode of the device to be charging or discharging. At the same time, a type C interface is designed. When the device is adjusted to the charging mode through the switch, it can be charged through this interface. The overall usage method is very simple and suitable for various people to use.

[0062] The sensor module is mainly an inertial sensor. The inertial sensor includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, and is used to collect the motion information of the patient's arm.

[0063] The processor module completes data acquisition by connecting to the inertial sensor module. The processor module receives the data from the inertial sensor and sends it to the wireless transceiver module through a serial port.

[0064] The wireless transceiver module is the link between the flexible wearable node and the data aggregation node, and it can send data information to the data aggregation node. The data aggregation node consists of a wireless communication module and a serial port transmission module. The wireless communication module of the data aggregation node receives the data of the flexible wearable node; the serial port transmission module will be connected to the wireless communication module and the data processing terminal respectively; it communicates with the wireless communication module and transmits the data of the wireless communication module to the data processing terminal.

[0065] The data processing terminal receives the data packet {addr, a 1 , a 2 , a 3 , …, a n} from the data aggregation node, and unpacks, parses, and preprocesses the data to obtain the available data including Euler angles and three-axis acceleration {roll, yaw, pitch, a x , a y , a z}.

[0066] The upper limb rehabilitation assessment method for patients uses the support vector machine algorithm to classify and identify the upper limb movement of patients, whether there are jitters, twitches, and the level of continuity of their upper limb movement. Its steps are as follows:

[0067] (1) Collect user inertial information including: roll angle roll, pitch angle pitch, heading angle yaw, x-axis output a x , y-axis acceleration output a y , z-axis acceleration output a z .

[0068] (3) Preprocess the three-axis acceleration output, filter out the gravity component, and eliminate the tiny data jitter.

[0069] (3) Collect the attitude angle data {roll, pitch, yaw} and the three-axis acceleration data {a x , a y , a z} to train the human upper limb movement recognition model and the twitch and jitter recognition models.

[0070] (4) Determine the upper limb movements of the patient based on the body's upper limb movement recognition through the attitude angle data {roll, yaw, pitch}.

[0071] (5) Determine whether there are tremors and convulsions during the user's rehabilitation training according to the tremor and convulsion recognition model through the acceleration data {a x , a y , a z}.

[0072] (6) Evaluate the continuity level of upper limb movements.

[0073] Among them, in step (1), collect the user's inertial information. The specific steps are as follows:

[0074] (1.1) Use the inertial module to collect the patient's inertial information. The inertial module transmits the inertial information to the processor module. The processor module sends the inertial information to the wireless transceiver module. The wireless transceiver module sends the inertial information to the data aggregation node one by one in byte form. The data aggregation node transmits the data to the data processing terminal. The data processing terminal converts the byte string into {roll, yaw, pitch, a x , a y , a z}.

[0075] In step (2), preprocess the output of the three-axis acceleration, filter out the gravity component, and eliminate the tiny data jitter.

[0076] (2.1) Receive the acceleration data {a x , a y , a z} transmitted from the wearable node. Use the first-order low-pass filter Y(n) = αX(n) + (1 - α)Y(n - 1) to obtain the components of the gravitational acceleration on the three axes {g x , g y , g z}. Subtract {g x , g y , g z} from the original data {a x , a y , a z} to obtain the three-axis linear acceleration data {a’ x , a’ y , a’ z} with the gravity component filtered out.

[0077] (2.2) Perform mean filtering on the three-axis linear acceleration data. Set its filtering window to L. Move the filtering window to perform mean filtering on the data to eliminate the tiny data jitter and smooth the data, obtaining {a (2) x , a(2) y , a (2) z}。

[0078] In step (3), collect the attitude angle data {roll, pitch, yaw} and the three-axis acceleration data {a x , a y , a z} to train the human upper limb movement recognition model and the twitch and jitter recognition model. The specific steps are as follows:

[0079] (3.1) A normal person wears a wearable node to collect the attitude angles {roll, pitch, yaw} that rotate 360° under specific actions, avoiding coordinate transformation, and constructing a motion information data set. Map the non-linear {roll i , pitch i , yaw i} (where i = 1, 2,..., n) to a high-dimensional space, and use the Gaussian kernel function K(x i , x j ) = e^[-||x 1 - x 2 || 2 / (2σ 2 )] to calculate the inner product in the high-dimensional space, thereby training the human upper limb movement recognition model.

[0080] (3.2) Both normal people and patients wear wearable nodes to collect the acceleration data {a (2) xi , a (2) yi , a (2) zi} (where i = 1, 2,..., n) during uniform and stable upper limb movements. Establish an acceleration data set A. Construct a sliding window of length L and slide it in the data set A. Use the FFT transform for the data within the window to obtain the discrete spectrum of the sequence and extract its maximum component frequency {f xmax , f ymax , f zmax}. Then, when the sliding window slides to the end of the data set, obtain the maximum component frequency {f (j) xmax , f (j) ymax , f (j) zmax} (where j = 1, 2,..., n - L), and establish a maximum component frequency data set F. Map the non-linear {f (j) xmax , f (j) ymax , f (j)zmax} is mapped to a high-dimensional space, and the Gaussian kernel function K(x i , x j ) = e^[-||x 1 - x 2 || 2 / (2σ 2 )] is used to calculate the inner product in the high-dimensional space, thereby training a twitch and tremor recognition model.

[0081] In step (4), according to the support vector machine classifier, the upper limb movements of the patient are judged based on the attitude angle data {roll, pitch, yaw}. The specific steps are as follows:

[0082] (4.1) Real-time collect the patient's attitude angle information {roll, pitch, yaw}, transmit it to the data processing terminal, and the data processing terminal calls the support vector machine classifier model for action recognition.

[0083] In step (5), according to the support vector machine classifier, it is judged whether there are tremors and twitches during the user's rehabilitation training based on the acceleration data {a x , a y , a z}. The specific steps are as follows:

[0084] (5.1) Real-time obtain the patient's three-axis acceleration information {a (2) x , a (2) y , a (2) z}, perform FFT transformation on the data within a sliding window of length L to obtain {f xmax , f ymax , f zmax}, call the twitch and tremor recognition model to determine whether there are twitches and tremors, and record the position of the first element of the sliding window when the model determines that there are twitches and tremors as the serial number of the sliding window, denoted as L k , k = 1, 2,..., t.

[0085] In step (6), according to the evaluation of the upper limb movement continuity level, the specific steps are as follows:

[0086] (6.1) Preset the jitter amplitude threshold {α 0 , α 1 , α 2 , α 3}, where α 0 < α 1 < α 2 < α 3 .

[0087] (6.2) According to Lk When the patient has convulsions and tremors, calculate the amplitude of the convulsions and tremors of the patient's upper limbs. Based on the sampling frequency f s The sampling time interval Δt is obtained, and the tremor amplitude on the three axes is obtained by integration and synthesized as d k . In the discrete space, the integration of acceleration is discretely approximated as follows:

[0088]

[0089] Solve for the maximum tremor amplitude d of the patient during the entire rehabilitation training process max = max(d k ), k = 1, 2, …, t.

[0090] (6.3) According to the tremor amplitude thresholds {α 0 , α 1 , α 2 , α 3} compared with d max , determine four grades: excellent, good, medium, and poor:

[0091] Excellent: d max < α 0

[0092] Good: α 0 <= d max < α 1

[0093] Medium: α 1 <= d max < α 2

[0094] Poor: d max >= α 2

[0095] Finally, display information such as the accuracy of the patient's upper limb movement, whether there are tremors and convulsions, and the grades of the continuity of the patient's upper limb movement on the data processing terminal.

[0096] Taking the example that the patient wears flexible wearable nodes on the left upper arm, left lower arm, right upper arm, and right lower arm, referring to Appendix Figure 5 , within the specified time T, make the following three movements uniformly, sequentially, and continuously: (1) Raise both hands horizontally to the sides, (2) Raise both hands straight up, (3) Raise the left hand high and touch the nose with the right finger.

[0097] Refer to Appendix Figure 3, the flexible wearable node 1 includes a flexible circuit board 11, an upper protective layer 12-1 of the flexible housing 12, and a lower protective layer 12-2 of the flexible housing 12. The flexible circuit board 11 is provided with an inertial sensor module 111, a processor module 112, a wireless transceiver module 113, and a power module. A switch 114-1 and a type C charging interface 114-2 are arranged on the power module.

[0098] The data aggregation node 2 includes a wireless transceiver module 21 and a serial port transmission module 22, and is connected to the data processing terminal 3 through a USB port.

[0099] After receiving the inertial data of the inertial sensor module 111, the data processing terminal 3 preprocesses the data, divides it into attitude angle data and three-axis acceleration data, trains the upper limb movement recognition model and the jitter and twitch recognition model of the patient, recognizes whether the patient makes actions (1), (2), (3), and whether there are involuntary upper limb jitters and twitches after the whole process, and evaluates the continuity of the patient's upper limb movement.

[0100] The method for evaluating the upper limb recovery of a patient based on inertial information according to the present invention, the process reference appendix Figure 2 , the steps are as follows:

[0101] S1: After receiving the inertial data of the inertial sensor module 111, the data processing terminal 3 preprocesses the data, divides it into attitude angle data {roll, pitch, yaw} and three-axis acceleration data {a x , a y , a z}

[0102] S1.1: Refer to appendix Figure 4 , by performing low-pass filtering on the three-axis acceleration data {a x , a y , a z} transmitted by the flexible wearable node 4 on the left forearm and the flexible wearable node 2 on the right forearm when performing actions (1)-(3) to obtain the gravity component, and then subtracting the gravity component to obtain the three-axis linear acceleration {a' x , a' y , a' z}, and then performing smoothing filtering to eliminate small jitters to obtain {a (2) x , a (2) y , a (2) z}, and the direction of this acceleration refers to appendix Figure 6 .

[0103] S1.2: Obtain the attitude angle data of the flexible wearable nodes 1, 2, 3, and 4 on the left and right upper arms of the human body.

[0104] S2: Train the recognition model. Respectively train the human upper limb motion recognition model and the jitter and twitch recognition model.

[0105] S2.1: Refer to Appendix Figure 7 Use a sliding window 5 with a length of L to collect {a (2) x , a (2) y , a (2) z}. Perform FFT transformation on the acceleration data 6 in the sliding serial port, use the frequency domain characteristics to solve the maximum component frequency within the sliding window, and construct a data set to train the jitter and twitch models.

[0106] S2.2: Use the attitude angle data to train the human upper limb motion recognition model. Collect the actions of normal people making actions (1)-(3), refer to Appendix Figure 5 , Appendix Figure 8 and Appendix Figure 9 . And rotate 360 degrees in place, collect the corresponding attitude angles to construct a data set, and perform training on the patient's upper limb motion recognition model.

[0107] S3: Call the recognition model.

[0108] S3.1: Obtain the maximum component frequency within the sliding window, and call the jitter and twitch recognition model for recognition.

[0109] S3.2: Call the patient's upper limb motion recognition model to recognize the patient's real-time actions.

[0110] S4: Record the position of the sliding window when jitter occurs. Calculate the jitter amplitude d according to the acceleration data within the sliding window k .

[0111] S5: Determine the maximum jitter amplitude d through the jitter amplitude d calculated in the above process k , determine the maximum jitter amplitude d max = max(d k ). Combine the preset jitter threshold and the maximum jitter amplitude d max to evaluate the continuity of the patient's upper limb movement.

[0112] S6: Display the accuracy of the upper limb movement, whether there is jitter, twitch, and the continuity of the upper limb movement on the terminal.

[0113] The above is only a preferred embodiment of the present invention, and it is not a limitation of the present invention in any other form. Any modification or equivalent change made according to the technical essence of the present invention still belongs to the scope of protection required by the present invention.

Claims

1. Evaluation method of a patient's upper limb recovery assessment system based on inertial information. The supporting equipment of the assessment system includes at least three wearable nodes respectively worn on the forearms and upper arms of the patient. The wearable nodes include a protective layer and the overall structure is a wristband. The hardware of the assessment system is arranged between the two protective layers. Characterized in that: The assessment system includes an inertial sensor module, a processor module, a wireless transceiver module and a power module. The inertial sensor module, the processor module, the wireless transceiver module and the power module are integrated in the hardware of the assessment system and are connected through a flexible circuit board; The power module supplies power to the entire system to work; The inertial sensor module includes a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer to collect the movement information of the patient's arm. The processor module completes data collection work by connecting to the inertial sensor module. The processor module receives the data from the inertial sensor and sends it to the wireless transceiver module through a serial port; The wireless transceiver module is the link between the flexible wearable node and the data aggregation node, and sends the data information to the data aggregation node. The data aggregation node consists of a wireless communication module and a serial port transmission module. The wireless communication module of the data aggregation node will receive the data of the flexible wearable node; The serial port transmission module will be connected to the wireless communication module and the data processing terminal respectively; maintain communication with the wireless communication module, and transmit the data of the wireless communication module to the data processing terminal; the data processing terminal receives the data packet {addr, a 1 , a 2 , a 3 , …, a n} from the data aggregation node, and unpacks, parses, and preprocesses the data to obtain the available data including Euler angles and three-axis accelerations {roll, yaw, pitch, a x , a y , a z}, and the specific steps are as follows: (1)Collecting user inertial information includes: roll angle, pitch angle, yaw angle, x-axis acceleration output a x , y-axis acceleration output a y , z-axis acceleration output a z ; (2) Preprocess the output of the three-axis accelerometer to filter out the gravity component and eliminate the tiny data jitter; (3) Collect attitude angle data {roll, pitch, yaw} and three-axis acceleration data {a x , a y , a z} to train the human upper limb motion recognition model and the twitch and jitter recognition model. The specific steps are as follows: (3.1) A normal person wears a wearable node to collect the attitude angles {roll, pitch, yaw} that rotate 360° under specific actions, avoiding coordinate system conversion, constructing a motion information data set, mapping the non-linear {roll i ,pitch i ,yaw i}, where i = 1, 2, …, n to a high-dimensional space, and using the Gaussian kernel function K(x i ,x j ) = e^[-||x i -x j || 2 / (2σ 2 )] to find the inner product in the high-dimensional space, thereby training a human upper limb motion recognition model; (3.2) Wearable nodes are worn by normal people and patients respectively, and the acceleration data {a (2) xi , a (2) yi , a (2) zi} during their uniform and steady upper limb movements is collected. For i = 1, 2, …, n, an acceleration data set A is established, a sliding window with a length of L is constructed, and it slides in the data set A. For the data within the window, FFT transformation is used to obtain the discrete spectrum of the sequence and extract its maximum component frequency {f xmax , f ymax , f zmax}. Then, when the sliding window slides to the end of the data set, the maximum component frequencies {f (j) xmax , f (j) ymax , f (j) zmax} are obtained. For j = 1, 2, …, n - L, a maximum component frequency data set F is established. The non-linear {f (j) xmax , f (j) ymax , f (j) zmax} is mapped to a high-dimensional space, and the Gaussian kernel function K(x i , x j ) = e^[-||x i - x j || 2 / (2σ 2 )] is used to calculate the inner product in the high-dimensional space, thereby training a twitch and tremor recognition model; (4) Judge the upper limb movement of the patient according to the attitude angle data {roll, yaw, pitch} through human upper limb movement recognition; (5) Determine whether there are tremors and convulsions in the user's rehabilitation training process based on the twitch and tremor recognition model through the acceleration data {a x , a y , a z}. The specific steps are as follows: (5.1) Obtain the patient's three-axis acceleration information {a (2) x , a (2) y , a (2) z}, perform FFT transformation on the data within a sliding window of length L to obtain {f xmax , f ymax , f zmax}, call the twitch and jitter recognition model to determine whether there are twitches and jitters, and record the position of the first element of the sliding window when the model determines that there are twitches and jitters as the serial number of the sliding window, denoted as L k , k = 1, 2, …, t; (6) Evaluate the upper limb movement continuity level, and the specific steps are as follows: (6.1) Preset jitter amplitude threshold {α 0 , α 1 , α 2 , α 3}, where α 0 < α 1 < α 2 < α 3 ; (6.2) According to L k When the patient has twitching and jittering, calculate the amplitude of the patient's upper limb twitching and jittering. Based on the sampling frequency f s Obtain its sampling time interval as Δt, use integration to obtain the jitter amplitude on the three axes and synthesize it, denoted as d k , in the discrete space, the integration of acceleration is discretely approximated as follows: Solve for the maximum jitter amplitude d of the patient during the entire rehabilitation training process max = max(d k ), k = 1, 2, …, t; (6.3) Determine four levels of excellent, good, medium, and poor according to the comparison between the jitter amplitude threshold {α 0 , α 1 , α 2 , α 3} and d max : Advantage: d max <α 0 Good: α 0 <= d max < α 1 In: α 1 ≤ d max < α 2 Difference: d max >= α 2 Finally, display the grade information of the accuracy of the patient's upper limb movement, whether there is jitter, twitching, and the continuity of the upper limb movement on the data processing terminal.

2. The evaluation method of the patient's upper limb recovery assessment system based on inertial information according to claim 1, Characterized in that: In the step (1), collecting the user's inertial information, the specific steps are: (1.1) Collect the inertial information of the patient using the inertial module. The inertial module transmits the inertial information to the processor module, and the processor module sends the inertial information to the wireless transceiver module. The wireless transceiver module sends the inertial information to the data aggregation node one byte at a time. The data aggregation node transmits the data to the data processing terminal, and the data processing terminal converts the byte string into {roll, yaw, pitch, a x , a y , a z}.

3. The evaluation method of the patient's upper limb recovery assessment system based on inertial information according to claim 1, Characterized in that: In the step (2), preprocess the output of the three-axis accelerometer to filter out the gravity component and eliminate the tiny data jitter; (2.1) Receive the acceleration data {a x , a y , a z} transmitted from the wearable node. Using the first-order low-pass filter Y(n) = αX(n) + (1 - α)Y(n - 1), obtain the components of the gravitational acceleration in the three axes {g x , g y , g z}. Subtract {g x , g y , g z} from the original data {a x , a y , a z} to obtain the three-axis linear acceleration data {a’ x , a’ y , a’ z} after filtering out the gravitational components; (2.2) Perform mean filtering on the three-axis linear acceleration data. Set its filtering window to L, move the filtering window to perform mean filtering on the data, eliminate minor data jitters, and smooth the data to obtain {a (2) x , a (2) y , a (2) z}.

4. The evaluation method of the patient's upper limb recovery assessment system based on inertial information according to claim 1, Characterized in that: In the step (4), judge the upper limb movement of the patient according to the support vector machine classifier through the attitude angle data {roll, pitch, yaw}, and the specific steps are: (4.1) Real-time collect the patient's attitude angle information {roll, pitch, yaw}, transmit it to the data processing terminal, and the data processing terminal calls the support vector machine classifier model for action recognition.

Images