Lumbar vertebra protection method and system and medium

Abstract

The invention relates to a lumbar vertebra protection method and system and a medium, and relates to the technical field of intelligent wearing. According to the scheme, personal basic information and past medical history of a patient are collected, the personal basic information comprises the age, the gender, the height, the weight and the BMI value of the patient, and the past medical history comprises past disease types and current disease courses; on the basis of the personal basic information and the past medical history, a lumbar vertebra supporting strength dynamic adjusting strategy is obtained, and the lumbar vertebra supporting strength dynamic adjusting strategy comprises a control strategy for supporting strength of the patient in all activity states, a control strategy for a supporting mode and recognized bad postures on the basis of the collected electromyographic signals of the patient; and controlling the waist support protection device based on the obtained dynamic adjustment strategy of the lumbar support strength, wherein the control comprises the control of the support strength, the control of the support mode and the early warning control of bad postures. Compared with the prior art, more targeted support strength and support mode control strategies for the patient can be obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent wearable devices, and particularly to a lumbar protection method, system and medium. Background Art

[0002] With the increasing work pressure and sedentary lifestyle in modern society, low back pain has become one of the major hidden health problems. Especially in labor-intensive industries, office workers and the elderly population, low back health problems are particularly prominent. Factors such as long-term improper postures, excessive muscle burden, and lack of exercise often exacerbate low back pain and recurrence of diseases. Therefore, daily protection and early warning prompts for the waist are more critical than starting treatment after getting sick.

[0003] Currently, existing lumbar protection devices, such as lumbar support belts and lumbar cushions, can help patients maintain correct postures or relieve lumbar pressure, but usually only provide relatively passive and general support, lacking intelligent real-time monitoring and active adjustment functions based on AI big data according to the personalized needs of different patients. These traditional devices cannot dynamically adjust the support strength according to the individual's muscle state, nor can they effectively provide timely warnings for the patient's poor postures or overworked muscles during different work and life activities. Summary of the Invention

[0004] The technical problem to be solved by this application is to provide a lumbar protection method, system and medium, which have the characteristics of being able to obtain more targeted support strength and support mode control strategies for patients.

[0005] In a first aspect, in one embodiment, a lumbar protection method is provided, which can be applied to a lumbar support protection device to achieve the protection of the lumbar spine. The method includes: Collect the patient's personal basic information and past medical history; the personal basic information includes the patient's age, gender, height, weight and BMI value; the past medical history includes the types of past diseases and the current course of the disease; Based on the personal basic information and past medical history, obtain a dynamic adjustment strategy for lumbar support strength; the dynamic adjustment strategy for lumbar support strength includes a control strategy for the support strength, a control strategy for the support mode, and the identified poor postures for the patient in each activity state based on the collected electromyogram signals of the patient; wherein, the support mode represents the speed of change of the support strength; Control the lumbar support protection device based on the obtained dynamic adjustment strategy for lumbar support strength, including control of the support strength, control of the support mode, and early warning control of poor postures.

[0006] In one embodiment, obtaining a dynamic adjustment strategy for lumbar support based on the personal basic information and past medical history includes: preprocessing the personal basic information and past medical history and inputting them into a dynamic adjustment model for lumbar support force to obtain the required dynamic adjustment strategy for lumbar support force; the dynamic adjustment model for lumbar support force is a neural network model trained in combination with a loss function, and the training method of the dynamic adjustment model for lumbar support force includes: Collect the personal basic information and past medical history of multiple patients; collect the electromyogram signals and pressure signals of each patient in different activity states, and label the electromyogram signals and pressure signals, including labeling the support force, support mode, and type of bad posture, to obtain training data with labels of support force, support mode, and type of bad posture; among them, both the electromyogram signals and pressure signals are time series data; Perform one-hot encoding and / or normalization data preprocessing on the personal basic information and past medical history of the patient, and use the preprocessed data as the first input vector; segment the electromyogram signals to obtain an electromyogram signal matrix as the second input vector; segment the pressure signals to obtain a pressure signal matrix as the third input vector; Extract features from the first input vector, the second input vector, and the third input vector respectively to obtain their corresponding first feature vector, second feature vector, and third feature vector; Concatenate the first feature vector, the second feature vector, and the third feature vector to obtain a fourth feature vector; Predict the support force, support mode, and type of bad posture for the fourth feature vector respectively. Among them, obtain the first loss function based on the prediction of the support force, obtain the second loss function based on the prediction of the support mode, and obtain the third loss function based on the prediction of the type of bad posture; Obtain the total loss function based on the first loss function, the second loss function, and the third loss function, and backpropagate and train and optimize the total loss function until the training termination condition is met to obtain the dynamic adjustment model for lumbar support force.

[0007] In one embodiment, the support force includes weak support force, medium support force, and strong support force from weak to strong; the support mode includes a weak support mode and a strong support mode, where the change speed of the support force in the weak support mode is less than that in the strong support mode.

[0008] In one embodiment, obtaining the first loss function based on the prediction of the support force includes: Calculate the mean square error loss, mean absolute error loss, and categorical cross-entropy loss based on the prediction of the support force; Calculate the first loss based on the mean square error loss, mean absolute error loss, and categorical cross-entropy loss; Both the second loss function and the third loss function adopt the cross - entropy loss function.

[0009] In one embodiment, the prediction calculation of the support strength is based on the mean square error loss, the mean absolute error loss, and the categorical cross - entropy loss, including: Among them, 、 and respectively represent the mean square error loss, the mean absolute error loss, and the categorical cross - entropy loss; and respectively represent the predicted support strength and the true support strength of the i - th support element, represents the j - th true support strength classification of the i - th support element, represents the j - th predicted support strength classification of the i - th support element, 1 ≤ i ≤ n, 1 ≤ j ≤ C, where n represents the number of support elements, and C represents the number of support strength classifications; Based on the mean square error loss, the mean absolute error loss, and the categorical cross - entropy loss, calculating the first loss, including: Among them, represents the first loss, 、 and represent hyperparameters used to balance different loss weights.

[0010] In one embodiment, the feature extraction of the first input vector, the second input vector, and the third input vector respectively to obtain their corresponding first feature vector, second feature vector, and third feature vector, including: After passing the first input vector through the first fully - connected processing, ReLU activation is performed to obtain the first feature vector; the second input vector is subjected to feature extraction using the first convolutional neural network to obtain the second feature vector; the third input vector is subjected to feature extraction using the second convolutional neural network to obtain the third feature vector.

[0011] In one embodiment, the early warning control of the bad posture includes: warning of the vibration position for the type of bad posture; the control of the lumbar support protection device based on the obtained dynamic adjustment strategy of the lumbar support strength includes: controlling the vibration element corresponding to the part affected by the bad posture to vibrate based on the obtained type of bad posture.

[0012] Second aspect, in one embodiment, a lumbar protection system is provided, which can be applied to a lumbar support protection device to achieve the protection of the lumbar spine. The system includes a signal acquisition module, a control module, and an execution module. Among them, The signal acquisition module includes a plurality of electromyogram signal sensors for collecting electromyogram signals; The control module is used to load and execute the computer program of the dynamic adjustment strategy of lumbar support strength in any one of the lumbar protection methods in claims 1 to 7, and control the execution module based on the collected electromyogram signals to implement the corresponding dynamic adjustment strategy of lumbar support strength; The execution module includes a plurality of support elements and a warning unit; the plurality of support elements are used to implement the corresponding support strength and support mode based on the dynamic adjustment strategy of lumbar support strength of the control module; the warning unit is used to implement the warning of bad postures.

[0013] In one embodiment, the warning unit includes a plurality of vibration elements, and the plurality of vibration elements are used to implement the vibration of the vibration elements corresponding to the parts affected by bad postures based on the dynamic adjustment strategy of lumbar support strength of the control module.

[0014] Third aspect, in one embodiment, a computer-readable storage medium is provided. A program is stored in the medium, and the program can be loaded and executed by a processor to perform the lumbar protection method in any one of the above embodiments.

[0015] The beneficial effects of the present invention are: Since the dynamic adjustment strategy of lumbar support strength required by the user is obtained based on the patient's personal basic information and past medical history to control the lumbar support protection device, a more targeted support strength and support mode control strategy for the patient can be obtained based on the patient's age, gender, height, weight, BMI information, and past medical history. Also, since the collected training data includes electromyogram signals and pressure information of the patient in different activity states, a lumbar support force dynamic adjustment model capable of obtaining the dynamic adjustment strategy of lumbar support strength is trained, so that a more targeted support strength and mode control strategy can be better obtained in different activity states. Description of the Drawings

[0016] Figure 1 is a structural block diagram of a lumbar protection system according to an embodiment of the present application; Figure 2 is a schematic flowchart of a lumbar protection method according to an embodiment of the present application; Figure 3 is the present application Figure 2 A schematic flowchart of a method according to an embodiment of step S20 in; Figure 4It is a schematic flowchart of a method for obtaining a first loss function based on support strength according to an embodiment of the present application.

[0017] In the figure, 01 is a signal acquisition module, 02 is a control module, 03 is an execution module, 0301 is a support element, and 0302 is an early warning unit. Specific embodiments

[0018] The present invention will be further described in detail below in conjunction with the accompanying drawings through specific embodiments. Similar elements in different embodiments are labeled with related similar element numbers. In the following embodiments, many detailed descriptions are provided to enable a better understanding of the present application. However, those skilled in the art can easily recognize that some of these features can be omitted in different situations, or can be replaced by other elements, materials, or methods. In some cases, some operations related to the present application are not shown or described in the specification to avoid overwhelming the core part of the present application with excessive descriptions. For those skilled in the art, it is not necessary to describe these related operations in detail, and they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0019] In addition, the features, operations, or characteristics described in the specification can be combined in any appropriate manner to form various embodiments. At the same time, the steps or actions in the method description can also be reordered or adjusted in a manner obvious to those skilled in the art. Therefore, the various sequences in the specification and drawings are only for clearly describing a certain embodiment and do not mean that they are necessary sequences, unless it is stated that a certain sequence must be followed.

[0020] The serial numbers assigned to the components herein, such as "first", "second", etc., are only used to distinguish the described objects and do not have any sequential or technical meaning. The terms "connection" and "coupling" used in the present application, unless otherwise specified, both include direct and indirect connection (coupling).

[0021] For the convenience of explaining the inventive concept of the present application, the lumbar protection technology is briefly described below.

[0022] In the current lumbar protection solutions applied to lumbar support protection devices, the myoelectric data at various positions can be collected to obtain the required support strength at each position, so as to adjust the support strength and thus achieve the protection of the lumbar spine.

[0023] However, the applicant found in the research that although the current lumbar support protection devices (such as smart belts) can adjust the support force on the muscles based on the collected electromyography data, the support force required by each person in the same activity state may also vary. For example, older patients may have weaker muscle strength and poorer recovery ability, and thus require more meticulous support and protection; for patients with different genders and BMI (Body Mass Index) values, the distribution and physiological characteristics of their lumbar muscles also vary, which need to be considered when analyzing muscle status and adjusting the support force. For patients with lumbar diseases, the belt needs to provide targeted support force and mode according to the severity of their condition and the recovery stage. For example, patients with lumbar disc herniation need stronger and stable support during the acute attack period to relieve the pressure on the intervertebral disc; during the recovery period, the support force needs to be gradually adjusted to help the muscles recover their function. For patients with chronic diseases such as diabetes, due to possible complications such as neuropathy, their ability to perceive and respond to pressure may decline, so more caution is needed when monitoring and adjusting the support force to avoid problems such as skin damage caused by excessive local pressure.

[0024] In view of this, in the embodiments of the present application, a lumbar protection method, system and medium are provided. Since the dynamic adjustment strategy of the lumbar support force required by the user is obtained based on the personal basic information and past medical history of the patient to control the lumbar support protection device, a more targeted support force and support mode control strategy for the patient can be obtained based on the patient's age, gender, height, weight, BMI information and past medical history. Also, since the collected training data includes electromyography signals and pressure information of the patient in different activity states, a lumbar support force dynamic adjustment model that can obtain the dynamic adjustment strategy of the lumbar support force is trained, so that a more targeted support force and mode control strategy can be better obtained in different activity states.

[0025] To better understand the embodiments of the present application, a lumbar protection system in the embodiments of the present application will be introduced first. Please refer to Figure 1 , the lumbar protection system includes a signal acquisition module 01, a control module 02 and an execution module 03, which are arranged on the lumbar support protection device, and the lumbar support protection device can be a support belt, a lumbar cushion, an ergonomic chair, etc.

[0026] The signal acquisition module 01 includes multiple electromyography signal sensors for collecting electromyography signals. Taking the belt as an example, multiple electromyography sensors are evenly distributed at key positions where the belt contacts the human lumbar muscles, such as near the erector spinae muscle and the rectus abdominis muscle. These electromyography sensors can collect the weak electrical signals generated by the muscles during activity and transmit these signals to the control module 02 to control the execution module 03.

[0027] The control module 02 is used to load and execute the computer program of the dynamic adjustment strategy of lumbar support strength in the lumbar protection method, and control the execution module 03 based on the collected electromyogram signals to implement the corresponding dynamic adjustment strategy of lumbar support strength.

[0028] The execution module 03 includes a support element 0301 and a warning unit 0302. There are multiple support elements 0301, which are used to implement the corresponding support strength and support mode based on the dynamic adjustment strategy of lumbar support strength of the control module 02. Among them, the support mode represents the speed of change of the support strength. The warning unit 0302 is used to realize the warning of bad postures.

[0029] In one embodiment, the support unit can adopt shape memory alloy wires. By applying current to the shape memory alloy wires through a drive circuit, the temperature of the alloy wires rises, and the shape changes, thereby increasing the support strength. And by controlling the speed of temperature rise, the speed of change of the support strength is controlled, and the support strength is accurately adjusted by controlling the magnitude and on-off time of the current, which will not be elaborated here.

[0030] In one embodiment, the personal basic information and medical history of the patient can be collected through a smart terminal (such as a mobile phone), so as to obtain the dynamic adjustment strategy of lumbar support strength required by the patient. The control module 02 is loaded with the dynamic adjustment strategy of lumbar support strength of the patient, and based on the electromyogram signals of each electromyogram sensor collected, the required support strength, support mode and bad posture warning control are output, so as to control the support elements distributed on the lumbar protection device and support the muscles in different activity states with the preset support strength and support mode.

[0031] The applicant found in the research that although there is also a bad posture warning in the current lumbar protection devices, its warning method is relatively single. As long as there is a bad posture, a warning will be issued, but it is very difficult for patients to immediately understand the part where the bad posture is located.

[0032] In view of this, in one embodiment, the warning unit 0302 includes multiple vibration elements, and the multiple vibration elements are used to realize the vibration of the vibration elements corresponding to the part affected by the bad posture based on the dynamic adjustment strategy of lumbar support strength of the control module 02, so that the patient can adjust his posture in time.

[0033] Combined with any one of the above lumbar protection systems, it can be applied to a lumbar support protection device to realize the protection of the human lumbar spine. Please refer to Figure 2 , a lumbar protection method provided in the embodiment of the present application includes: Step S10, collect the patient's basic personal information and past medical history. Among them, the basic personal information includes the patient's age, gender, height, weight, and BMI information; the past medical history includes the types of past diseases and the current stage of the disease. For example, if the past medical history includes lumbar disc herniation, it is necessary to indicate the current stage of the disease, such as the acute attack stage or the recovery stage.

[0034] In one embodiment, the patient's basic personal information and past medical history can be collected through an APP on a smart terminal such as a mobile phone. In one embodiment, for the convenience of the patient's operation, the option selection method can be used to collect the past medical history.

[0035] Step S20, based on the basic personal information and past medical history, obtain the dynamic adjustment strategy for lumbar support strength. Among them, the dynamic adjustment strategy for lumbar support strength includes the control strategy for the support strength, the control strategy for the support mode, and the identified bad postures of the patient in various activity states based on the collected myoelectric signals of the patient; among them, the support mode represents the speed of change of the support strength.

[0036] In one embodiment, step S20 may include: preprocess the basic personal information and past medical history and then input them into the dynamic adjustment model of lumbar support force to obtain the required dynamic adjustment strategy for lumbar support strength; the dynamic adjustment model of lumbar support force is a neural network model trained in combination with a loss function.

[0037] Those skilled in the art can understand that the preprocessing process is to obtain the input form requirements that meet the dynamic adjustment model of lumbar support force, which is the same as the preprocessing of data during the model training process.

[0038] In one embodiment, please refer to Figure 3 , the training method of the dynamic adjustment model of lumbar support force includes: Step S201, collect the basic personal information and past medical history of multiple patients; collect the myoelectric signals and pressure signals of each patient in different activity states, and label the myoelectric signals and pressure signals, including labeling the support strength, support mode, and bad posture type, to obtain training data with labels of support strength, support mode, and bad posture type. Among them, both the myoelectric signals and pressure signals are time series data.

[0039] In one embodiment, the support strength ranges from weak to strong, including weak support strength, medium support strength, and strong support strength.

[0040] In one embodiment, the support mode includes a weak support mode and a strong support mode. Among them, the speed of change of the support strength in the weak support mode is less than that in the strong support mode.

[0041] To ensure data diversity, electromyography signals and pressure signals of patients with different genders, different age groups, different heights, different weights, different BMI values, and different past medical histories are collected under various activity states to collect suitable support strengths and bad postures in various situations.

[0042] In one embodiment, the collected data is divided according to the ratio of 70% training set, 15% validation set, and 15% test set. The training set is used for updating model parameters, the validation set is used for adjusting hyperparameters to prevent overfitting, and the test set is used for evaluating the performance of the final model.

[0043] In step S202, one-hot encoding and / or normalization data preprocessing is performed on the patient's personal basic information and past medical history, and the preprocessed data is used as the first input vector; the electromyography signal is segmented to obtain an electromyography signal matrix as the second input vector; the pressure signal is segmented to obtain a pressure signal matrix as the third input vector.

[0044] In the preprocessing, normalization processing can be performed on age, height, weight, and BMI value respectively to make their values within the range of 0 to 1. Taking age as an example, assuming the age range is from 18 to 90, the normalized value of the actual age of 50 is (50 - 18) / (90 - 18) = 0.44. For gender and past medical history, one-hot encoding processing can be performed respectively. For example, 0 represents male and 1 represents female. The severity of lumbar diseases (such as lumbar disc herniation, lumbar muscle strain) (mild, moderate, severe, corresponding to 1, 2, 3 respectively), the recovery stage (acute stage, early rehabilitation stage, late rehabilitation stage, corresponding to 1, 2, 3 respectively), and other chronic diseases (diabetes, etc. are recorded as 1, otherwise 0).

[0045] In step S203, feature extraction is respectively performed on the first input vector, the second input vector, and the third input vector to obtain the corresponding first feature vector, second feature vector, and third feature vector.

[0046] In one embodiment, performing feature extraction on the first input vector includes: after the first input vector undergoes fully connected processing, ReLU activation is performed to obtain the first feature vector.

[0047] The fully connected layer can be used to extract features from the patient's personal basic information and medical history vector, and then ReLU activation is performed to introduce non-linear features.

[0048] In one embodiment, feature extraction of the second input vector includes: using a first convolutional neural network to perform feature extraction on the second input vector to obtain a second feature vector. In one embodiment, the convolutional neural network includes a CNN (convolutional neural network) structure including a plurality of convolutional layers and pooling layers. For example, 3 convolutional layers are used, and each convolutional layer is followed by a max pooling layer. The convolutional kernel sizes can be set to 3×3, 3×3, 3×3 respectively, and the number of channels are 16, 32, 64 respectively. Finally, the output of the convolutional layer is flattened into a one-dimensional vector.

[0049] In one embodiment, feature extraction of the third input vector includes: using a second convolutional neural network to perform feature extraction on the third input vector to obtain a third feature vector. The second convolutional neural network can adopt the same network structure as the first convolutional neural network.

[0050] Step S204, concatenate the first feature vector, the second feature vector and the third feature vector to obtain a fourth feature vector.

[0051] Concatenate the first feature vector, the second feature vector and the third feature vector to obtain a fused fourth feature vector.

[0052] Step S205, predict the support strength, support mode and bad posture type for the fourth feature vector respectively. Among them, a first loss function is obtained based on the prediction of the support strength, a second loss function is obtained based on the prediction of the support mode, and a third loss function is obtained based on the prediction of the bad posture type.

[0053] In one embodiment, both the second loss function and the third loss function adopt the cross-entropy loss function.

[0054] The applicant found in the research that since the prediction of the support strength includes different degrees of support strength of multiple sensors and continuous values need to be obtained, general loss functions are difficult to meet the prediction requirements. In view of this, an embodiment of the present application provides a method for obtaining a first loss function to solve this technical problem. Please refer to Figure 4 , the method for obtaining the first loss function based on the prediction of the support strength includes: Step S100, calculate the mean square error loss, mean absolute error loss and categorical cross-entropy loss based on the prediction of the support strength.

[0055] In one embodiment, step S100 can be expressed as: Among them, 、 and respectively corresponding to and representing mean squared error loss, mean absolute error loss, and categorical cross-entropy loss; and respectively represent the predicted support strength and the true support strength of the i-th support element, represents the j-th true support strength classification of the i-th support element, represents the j-th predicted support strength classification of the i-th support element, where 1 ≤ i ≤ n, 1 ≤ j ≤ C, n represents the number of support elements, and C represents the number of support strength classifications.

[0056] By calculating the average of the squares of the differences between the predicted support strength value and the true value through mean squared error, it can effectively measure the degree of difference between the predicted value and the true value; by calculating the average of the absolute values of the differences between the predicted support strength value and the true value through mean absolute error, it can more intuitively reflect the average error magnitude between the predicted value and the true value. Based on categorical cross-entropy, the problem can be transformed into a multi-classification problem.

[0057] Step S200, calculate the first loss based on mean squared error loss, mean absolute error loss, and categorical cross-entropy loss.

[0058] In one embodiment, step S200 can be expressed as: where represents the first loss, , and represent hyperparameters used to balance different loss weights, which can be adjusted according to specific problems and experimental results.

[0059] Step S206, obtain the total loss function based on the first loss function, the second loss function, and the third loss function, and backpropagate and train the total loss function for optimization until the training termination condition is met to obtain the lumbar support force dynamic adjustment model.

[0060] Step S30, control the lumbar support protection device based on the obtained dynamic adjustment strategy for lumbar support strength, including control of support strength, control of support mode, and early warning control of bad postures.

[0061] In one embodiment, the early warning control of bad postures includes: giving an early warning of the vibration position for the type of bad posture; controlling the lumbar support protection device based on the obtained dynamic adjustment strategy for lumbar support strength, including: controlling the vibration elements corresponding to the parts affected by the bad posture to vibrate based on the obtained type of bad posture. In this way, the patient can adjust their posture in a timely manner.

[0062] Based on the lumbar protection method in any of the above embodiments, since it is based on the patient's personal basic information and past medical history to obtain the dynamic adjustment strategy of the lumbar support strength required by the user to control the lumbar support protection device, it is possible to obtain a more targeted support strength and support mode control strategy for the patient based on the patient's age, gender, height, weight, BMI information and past medical history. Also, since the collected training data includes the electromyogram signals and pressure information of the patient in different activity states, a lumbar support force dynamic adjustment model that can obtain the dynamic adjustment strategy of the lumbar support strength is trained, so that a more targeted support strength and mode control strategy can be better obtained in different activity states.

[0063] In an embodiment of the present application, a computer-readable storage medium is provided. A program is stored on the storage medium, and the stored program includes a method that can be loaded and processed by a processor in any of the above embodiments.

[0064] Those skilled in the art can understand that all or part of the functions of the above methods can be implemented in a hardware manner or in a computer program manner. When all or part of the functions in the above embodiments are implemented in a computer program manner, the program can be stored in a computer-readable storage medium. The storage medium may include: read-only memory, random access memory, magnetic disk, optical disk, hard disk, etc. The above functions are implemented by a computer executing the program. For example, the program is stored in the memory of the device, and when the processor executes the program in the memory, the above all or part of the functions can be implemented. In addition, when all or part of the functions in the above embodiments are implemented in a computer program manner, the program can also be stored in a storage medium such as a server, another computer, magnetic disk, optical disk, flash drive or mobile hard disk, downloaded or copied and saved to the memory of the local device, or the system of the local device is updated. When the processor executes the program in the memory, the above all or part of the functions in the above embodiments can be implemented.

[0065] The above uses specific examples to elaborate on the present invention, which is only used to help understand the present invention and is not intended to limit the present invention. For those skilled in the technical field to which the present invention belongs, based on the idea of the present invention, several simple deductions, deformations or substitutions can also be made.

Claims

1. A lumbar protection method, characterized in that, It can be applied to a lumbar support protection device to achieve the protection of the lumbar spine. The method includes: Collecting the patient's personal basic information and past medical history; the personal basic information includes the patient's age, gender, height, weight, and BMI value; the past medical history includes the types of past diseases and the current stage of the disease. Based on the personal basic information and past medical history, obtaining a dynamic adjustment strategy for lumbar support strength; the dynamic adjustment strategy for lumbar support strength includes a control strategy for support strength, a control strategy for support mode, and recognized bad postures for the patient in various activity states based on the collected myoelectric signals of the patient; where the support mode represents the speed of change of the support strength. Controlling the lumbar support protection device based on the obtained dynamic adjustment strategy for lumbar support strength, including control of support strength, control of support mode, and early warning control of bad postures.

2. The lumbar protection method according to claim 1, wherein The obtaining of the dynamic adjustment strategy for lumbar support strength based on the personal basic information and past medical history includes: preprocessing the personal basic information and past medical history and inputting them into a dynamic adjustment model for lumbar support force to obtain the required dynamic adjustment strategy for lumbar support strength; the dynamic adjustment model for lumbar support force is a neural network model trained in combination with a loss function, and the training method of the dynamic adjustment model for lumbar support force includes: Collecting the personal basic information and past medical history of multiple patients; collecting the myoelectric signals and pressure signals of each patient in different activity states, and annotating the myoelectric signals and pressure signals, including annotating support strength, support mode, and bad posture types, to obtain training data with labels of support strength, support mode, and bad posture types; where both the myoelectric signals and pressure signals are time series data. Performing one-hot encoding and / or normalization data preprocessing on the patient's personal basic information and past medical history, and using the preprocessed data as the first input vector; segmenting the myoelectric signals to obtain a myoelectric signal matrix as the second input vector; segmenting the pressure signals to obtain a pressure signal matrix as the third input vector. Performing feature extraction on the first input vector, the second input vector, and the third input vector respectively to obtain their corresponding first feature vector, second feature vector, and third feature vector. Performing splicing processing on the first feature vector, the second feature vector, and the third feature vector to obtain a fourth feature vector. Predicting the support strength, support mode, and bad posture types for the fourth feature vector respectively, where a first loss function is obtained based on the prediction of support strength, a second loss function is obtained based on the prediction of support mode, and a third loss function is obtained based on the prediction of bad posture types. Obtaining a total loss function based on the first loss function, the second loss function, and the third loss function, and performing backpropagation training optimization on the total loss function until the training termination condition is met to obtain a dynamic adjustment model for lumbar support force.

3. The lumbar spine protection method according to claim 2, wherein The support strength ranges from weak to strong, including weak support strength, medium support strength, and strong support strength. The support mode includes a weak support mode and a strong support mode, where the speed of change of the support strength in the weak support mode is less than that in the strong support mode.

4. The lumbar spine protection method according to claim 2, characterized in that, The prediction based on the support strength to obtain the first loss function includes: Calculating the mean square error loss, mean absolute error loss, and categorical cross-entropy loss based on the prediction of the support strength; Calculating the first loss based on the mean square error loss, mean absolute error loss, and categorical cross-entropy loss; Both the second loss function and the third loss function adopt the cross-entropy loss function.

5. The lumbar protection method according to claim 4, characterized in that, The calculation of the mean square error loss, mean absolute error loss, and categorical cross-entropy loss based on the prediction of the support strength includes: Among them, , and correspond to mean squared error loss, mean absolute error loss, and categorical cross-entropy loss respectively; and represent the predicted support strength and the true support strength of the i-th support element respectively, represents the j-th true support strength classification of the i-th support element, represents the j-th predicted support strength classification of the i-th support element, 1 ≤ i ≤ n, 1 ≤ j ≤ C, where n represents the number of support elements and C represents the number of support strength classifications; The calculation of the first loss based on the mean square error loss, mean absolute error loss, and categorical cross-entropy loss includes: ， Among them, represents the first loss, , and represent hyperparameters for balancing different loss weights.

6. The lumbar protection method according to claim 2, characterized in that, The feature extraction of the first input vector, the second input vector, and the third input vector respectively to obtain the corresponding first feature vector, second feature vector, and third feature vector includes: After the first input vector undergoes the first fully connected process, ReLU activation is performed to obtain the first feature vector; the second input vector is subjected to feature extraction using the first convolutional neural network to obtain the second feature vector; the third input vector is subjected to feature extraction using the second convolutional neural network to obtain the third feature vector.

7. The lumbar spine protection method according to claim 2, characterized in that, The early warning control of the bad posture includes: warning of the vibration position for the types of bad postures; the control of the lumbar support protection device based on the obtained dynamic adjustment strategy of the lumbar support strength includes: based on the obtained types of bad postures, controlling the vibration elements corresponding to the parts affected by the bad postures to vibrate.

8. A lumbar protection system, characterized in that, It can be applied to the lumbar support protection device to achieve the protection of the lumbar spine. The system includes a signal acquisition module, a control module, and an execution module. Among them, the signal acquisition module includes multiple electromyogram signal sensors for collecting electromyogram signals; The control module is used to load and execute the computer program of the dynamic adjustment strategy of the lumbar support strength in any one of the lumbar spine protection methods as claimed in claims 1 to 7, and control the execution module based on the collected electromyogram signals to implement the corresponding dynamic adjustment strategy of the lumbar support strength; The execution module includes multiple support elements and an early warning unit; the multiple support elements are used to implement the corresponding support strength and support mode based on the dynamic adjustment strategy of the lumbar support strength of the control module; the early warning unit is used to implement the early warning of bad postures.

9. The lumbar protection system according to claim 8, wherein, The early warning unit includes multiple vibration elements, and the multiple vibration elements are used to implement the vibration of the vibration elements corresponding to the parts affected by the bad postures based on the dynamic adjustment strategy of the lumbar support strength of the control module.

10. A computer-readable storage medium, characterized in that The medium stores a program, and the program can be loaded and executed by a processor to perform the lumbar spine protection method as claimed in any one of claims 1 to 7.

Images