Balance and stability assessment method, device and medium based on flexible imu sensor

Abstract

The application provides a balance and stability evaluation method based on a flexible IMU sensor, a device and a medium, and the method comprises the following steps: recording acceleration data during the action of a subject performing a Romberg test procedure; processing the acceleration data; filtering and normalizing the acceleration data; extracting features of the processed acceleration data; and inputting the extracted features into a trained neural network model to predict the index value of a force table device. The application uses low-cost, portable and small IMU to predict the index obtained by the force table device, provides a direct connection between the two balance evaluation methods, and thus adapts to the "gold standard". During the open-eye and closed-eye periods, the acceleration data is measured by an accelerometer, and a deep learning method is used to learn the complex relationship from the complex balance parameters to construct a prediction model, which is compared with the parameters obtained by the force table device, and a more portable and high-precision tool is provided for the evaluation of balance ability.

Description

Technical Field

[0001] This invention relates to the field of rehabilitation application assessment technology, and in particular to a method, device and medium for assessing balance and stability based on a flexible IMU sensor. Background Technology

[0002] Balance, or postural control, is defined as the ability to keep the center of gravity within a support area marked by the outer contours of the feet. Maintaining balance requires coordination and synergy between the vestibular, visual, and somatosensory systems (Heebner, Akins et al. 2015). Balance ability is a good indicator of the health of these complex systems in the human body and is widely used in monitoring early childhood development, assessing physical function in the elderly, and monitoring falls. Many specific diseases (such as Parkinson's disease, stroke, and multiple sclerosis) are closely related to balance ability. For example, patients with multiple sclerosis suffer from balance deficits due to slowed somatosensory conduction and impaired central integration (Sun, Moon et al. 2018), and balance ability is considered an important indicator for assessing the progression and rehabilitation of these diseases. Therefore, conducting effective balance ability assessments is crucial.

[0003] Currently, the main methods for assessing balance ability include observation, rating scales, and quantitative posture diagrams. Observation and rating scales are easy to operate and understand, but their excessive subjective judgment leads to low reliability, and their reliance on extensive clinical experience limits their widespread adoption. The quantitative posture diagram method, often considered the "gold standard," measures the trajectory of the center of pressure (CoP) using parameters such as the support reaction force and torque of a force table, and performs sway velocity analysis (CoPv) or trajectory spatial dispersion analysis (area, root mean square, path length, etc.) to generate quantitative data results (Jaworski, Ambrozy et al. 2020). It is divided into static and dynamic posture diagrams. Among these, the static balance task based on the Romberg test has become the most common and practical method in clinical practice due to its simplicity and ease of execution; however, the high cost and bulkiness of the force table equipment limit its further development and large-scale universal application. Summary of the Invention

[0004] To achieve the above-mentioned objectives and other advantages of the present invention, a first objective of the present invention is to provide a method for evaluating the balance and stability based on a flexible IMU sensor, comprising the following steps:

[0005] Acceleration data was recorded during the subjects' execution of the Lobo test procedure to obtain the acceleration data to be predicted;

[0006] The acceleration data is processed;

[0007] The processed acceleration data is then filtered and normalized.

[0008] Extract features from the processed acceleration data;

[0009] The extracted features are input into the trained neural network model to predict the index values ​​of the corresponding force measuring platform equipment.

[0010] Furthermore, the training of the neural network model includes the following steps:

[0011] Record acceleration data and force table equipment parameters during the subject's execution of the Lobo test procedure;

[0012] The acceleration data is processed;

[0013] The parameters of the force measuring table equipment are used as the indicators to be predicted.

[0014] The processed acceleration data is combined with the index to be predicted to form a new sample;

[0015] Filter and normalize all samples;

[0016] The healthy samples and balanced risk samples are divided by whether the true value of the indicator to be predicted is within the healthy range, and the number of positive and negative samples is balanced.

[0017] Extract features from healthy samples and balanced risk samples;

[0018] The extracted features are then input into the neural network model for training.

[0019] Furthermore, the acceleration data is the linear acceleration of motion within a three-dimensional frame fixed by a triaxial accelerometer.

[0020] Furthermore, the linear acceleration includes both motion and gravitational components.

[0021] Furthermore, the actions performed by the subject in the Lobo test procedure include closing their eyes, bending over, and opening their eyes.

[0022] Furthermore, the processing of the acceleration data includes the following steps:

[0023] Acceleration data for both closed and open eyes are retained;

[0024] After removing data during the action marking period, start and end data, and error data, the data with a preset duration in the middle of each state is retained and divided into multiple segments.

[0025] Furthermore, the indicators to be predicted are the elliptical area and the swing path length output by the force measuring table device.

[0026] Furthermore, the step of combining the processed acceleration data with the index to be predicted to form a new sample includes the following steps:

[0027] Acceleration data for each closed-eye and open-eye phase, along with the indicators to be predicted, are combined into two new samples.

[0028] Furthermore, the step of dividing healthy samples and balanced risk samples based on whether the true value of the indicator to be predicted is within the healthy range, and balancing the number of positive and negative samples, includes the following steps:

[0029] The healthy samples and balanced risk samples are divided by the health reference experience range of the ellipse area index and the health reference experience range of the swing path length index, and the number of positive and negative samples is balanced.

[0030] The number of samples that fall within or outside the health reference range of the ellipse area index and the swing path length index are respectively balanced according to the ellipse area index and the swing path length index.

[0031] Furthermore, the healthy reference range for the ellipse area index is 1.0 to 2.96;

[0032] The healthy reference range for the swing path length index is 1.0 to 1.73.

[0033] Furthermore, the extracted features include the mean, standard deviation, maximum, minimum, average peak value, average peak spacing, JERK parameters, root mean square, centroid frequency, mean / variance of acceleration spectrum slices, and whether the patient is a stroke patient.

[0034] Furthermore, the step of inputting the extracted features into the neural network model for training includes the following steps:

[0035] Calculate the root mean square error and the COD correlation parameter, and find the optimal weights for the root mean square error and the COD correlation parameter through a grid search method;

[0036] The optimal weights are normalized and used as weights for calculating the overall loss; wherein, the COD correlation parameter is calculated by linear fitting prediction index and the true value output by the force measuring table equipment;

[0037] During training, the fitting slope is controlled within the first quadrant by controlling the weights, and the mean squared error loss weights are decayed by annealing strategy.

[0038] For each weight, the model is trained, and the optimal weight combination is found by using an early stopping strategy. The model with the optimal parameters is then used to output the index values ​​of the force measuring platform equipment.

[0039] The fit is evaluated by calculating the fitting parameters between the predicted and true values, the prediction classification accuracy, precision, recall, and F1 score.

[0040] A second object of the present invention is to provide an electronic device comprising: a memory having program code stored thereon; and a processor connected to the memory, wherein the above-described method is implemented when the program code is executed by the processor.

[0041] A third objective of the present invention is to provide a computer-readable storage medium having program instructions stored thereon, which, when executed, implement the method described above.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] This invention provides a method, device, and medium for balance and stability assessment based on a flexible IMU sensor. It uses data measured by a low-cost, portable small IMU (accelerometer) to predict the indicators obtained from a force-measuring platform device, providing a direct link between two balance assessment methods, thus conforming to the "gold standard." Acceleration data is measured using a belt-type accelerometer device fixed at L5 on the lower back of the body during eye-opening (EO) and eye-closing (EC) states. Using methods such as deep learning, complex relationships are learned from numerous balance parameters to construct a predictive model, which is then compared with the parameters obtained from the force-measuring platform device, providing a more portable and high-precision tool for assessing balance ability.

[0044] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it according to the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Specific embodiments of the present invention are given in detail below with reference to the accompanying drawings. Attached Figure Description

[0045] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0046] Figure 1 This is a flowchart of the balance and stability evaluation method based on a flexible IMU sensor in Example 1;

[0047] Figure 2 This is a flowchart of the data acquisition process.

[0048] Figure 3 This is a schematic diagram of the system framework;

[0049] Figure 4 Flowchart for balance and stability assessment;

[0050] Figure 5 This is a schematic diagram of a neural network structure;

[0051] Figure 6 This is a flowchart of the training process for a neural network model.

[0052] Figure 7 Flowchart for processing acceleration data;

[0053] Figure 8 A flowchart for dividing healthy samples and balanced risk samples based on whether the true value of the indicator to be predicted is within the healthy range, and balancing the number of positive and negative samples;

[0054] Figure 9 Flowchart for training a neural network model by inputting the extracted features;

[0055] Figure 10 This is a schematic diagram of the electronic device in Example 2;

[0056] Figure 11 This is a schematic diagram of the computer storage medium in Example 3. Detailed Implementation

[0057] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments. It should be noted that, without conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0058] Balance and postural stability assessment is crucial in medical rehabilitation and daily life. In existing technologies, the high cost and bulkiness of force tables limit their further development and widespread application. In contrast, wearable sensors offer advantages such as ease of management and relatively low cost, thus potentially leading to a wider range of applications (Zemkova, Durinova et al. 2021). Therefore, wearable sensors are considered to have significant application potential in assessing postural stability. The most widely used type is the inertial measurement unit (IMU), represented by the accelerometer.

[0059] The purpose of this invention is to use a wearable accelerometer and a force platform to simultaneously record data under the Romberg test procedure, and to predict the balance index measured by the force platform using the accelerometer data, thereby replacing the former. This invention proposes a balance and stability evaluation method based on a flexible IMU sensor.

[0060] Example 1

[0061] Methods for evaluating the balance and stability of flexible IMU sensors, such as Figure 1 , Figure 4 As shown, it includes the following steps:

[0062] S1. Record the acceleration data of the subject during the execution of the Lobo test procedure to obtain the acceleration data to be predicted;

[0063] In some embodiments, such as Figure 3 As shown, wearable inertial sensing units typically include accelerometers, gyroscopes, and magnetometers. A triaxial accelerometer measures the appropriate linear acceleration of motion within a fixed three-dimensional (3D) frame, and the measured data includes both motion and gravitational components. For example... Figure 2 As shown, the subjects performed the Lobo test procedure, which involved closing their eyes for 30 seconds, bending over, and opening their eyes for 30 seconds, and the accelerometer data was recorded during the process.

[0064] S2. Process the acceleration data;

[0065] Specifically, the acceleration data during the Lobo test is retained, including the EC (eyes closed) and EO (eyes open) states. After removing the data during the action marking period, the beginning and end data, and erroneous data, the data in the middle 25 seconds of each state is retained and divided into n segments.

[0066] S3. Filter and normalize the processed acceleration data; specifically, perform a 5th-order Butterworth low-pass filter on the processed acceleration data with a cutoff frequency of 10Hz, and then perform normalization.

[0067] S4. Extract features from the processed acceleration data; specifically, the extracted features include the mean, standard deviation, maximum value, minimum value, average peak value (average of the peak values ​​of the acceleration data), average peak spacing (average of the spacing between adjacent acceleration peaks), and JERK parameters for each dimension. Root mean square and centroid frequency of the filtered acceleration data Features such as the mean / variance of acceleration spectrum slices (acceleration spectrum segmented according to percentiles 0, 15, 30, 45, 60, 75, 90, 100, and the mean and variance of each segment) and whether it is a stroke patient.

[0068] S5. Input the extracted features into the trained neural network model to predict the index value of the corresponding force measuring platform equipment.

[0069] In some embodiments, the neural network structure is as follows Figure 5 As shown, the neural network consists of five linear layers: an input layer (85 dimensions), a first hidden layer (128 dimensions), a second hidden layer (128 dimensions), a third hidden layer (64 dimensions), and an output layer (2 dimensions). The activation function is the ReLU function.

[0070] like Figure 6 As shown, training a neural network model includes the following steps:

[0071] S51. Record the acceleration data and force table equipment parameters during the subject's execution of the Lobo test procedure.

[0072] In some embodiments, wearable inertial sensing units typically include accelerometers, gyroscopes, and magnetometers. A triaxial accelerometer measures the appropriate linear acceleration of motion within a fixed three-dimensional (3D) frame, and the measured data includes motion and gravitational components. Figure 2 As shown, the subjects performed the Lobo test procedure, which involved closing their eyes for 30 seconds, bending over, and opening their eyes for 30 seconds. Accelerometer data and force table parameters were recorded during the test.

[0073] S52. Process the acceleration data; specifically, such as... Figure 7 As shown, it includes the following steps:

[0074] S521. Retain acceleration data during the Lobo test, including both EC (eyes closed) and EO (eyes open) states;

[0075] S522. After removing data during the action marking period, start and end data, and error data, retain the data in the middle 25 seconds of each state and divide it into n segments.

[0076] S53. The parameters of the force measuring table equipment are used as the indicators to be predicted; specifically, the indicators to be predicted are the elliptical area and the swing path length output by the force measuring table equipment (both indicators are divided into EC and EO stages).

[0077] S54. Combine the processed acceleration data with the index to be predicted to form a new sample; specifically, combine the acceleration data of each EC segment and EO segment with the index to be predicted to form two new samples, and predict two labels respectively, namely (EC... i +EO j EllipseArea), (EC i +EO j Sway Length).

[0078] Where i, j∈(1, 2, ..., n), and n=4, then for both indicator prediction tasks, the sample size becomes n times the original size. 2 times.

[0079] S55. Filter and normalize all samples; specifically, perform a 5th-order Butterworth low-pass filter on all samples with a cutoff frequency of 10Hz, and then perform normalization.

[0080] S56. Divide healthy samples and balanced risk samples by whether the true value of the indicator to be predicted is within the healthy range, and balance the number of positive and negative samples; for example Figure 8 As shown, the specific steps include:

[0081] S561. Divide healthy samples and balanced risk samples by the health reference experience range of the ellipse area index and the health reference experience range of the swing path length index, and balance the number of positive and negative samples; preferably, the health reference experience range of the ellipse area index is 1.0 to 2.96; the health reference experience range of the swing path length index is 1.0 to 1.73.

[0082] S562. Balance the number of samples that are outside the health reference experience range of the ellipse area index and the health reference experience range of the swing path length index, respectively, according to the ellipse area index and the swing path length index.

[0083] S57. Extract features from healthy samples and balanced risk samples; specifically, the extracted features include the mean, standard deviation, maximum value, minimum value, average peak value (average of the peak values ​​of acceleration data), average peak spacing (average of the spacing between adjacent acceleration peaks), and JERK parameters for each dimension. Root mean square and centroid frequency of the filtered acceleration data Features such as the mean / variance of acceleration spectrum slices (acceleration spectrum segmented according to percentiles 0, 15, 30, 45, 60, 75, 90, 100, and the mean and variance of each segment) and whether it is a stroke patient.

[0084] S58. Input the extracted features into the neural network model for training. For example... Figure 9 As shown, the specific steps include:

[0085] S581. Calculate the root mean square error and COD correlation parameters, and find the optimal weights for the root mean square error and COD correlation parameters through a grid search method; in this embodiment, a four-dimensional weight array (Wrms) is formed by combining integers from 1 to 7. a Wrms l Wcod a Wcod l ).

[0086] S582. The optimal weights are normalized and used as weights for calculating the overall loss; wherein, the COD correlation parameter is calculated by linear fitting prediction index and the true value output by the force table equipment.

[0087] S583. During the training process, the fitting slope is controlled within the first quadrant by controlling the weights, and the mean squared error loss weights are attenuated by the annealing strategy, thereby improving the linear fitting effect.

[0088] S584. For each weight, train the model and use the early stopping strategy to efficiently find the optimal weight combination and obtain the model with the optimal parameters, which is used to output the index value of the force measuring table equipment.

[0089] S585. The fit is evaluated by calculating the fitting parameters between the predicted and true values, the prediction classification accuracy, precision, recall, and F1 score.

[0090] This invention successfully predicted the output parameters of a force-measuring platform during static balance testing using data from an accelerometer. The fitted COD values ​​for the Ellipse area and Length parameters reached 0.804 and 0.812, respectively. When filtering the original signal using a 5th-order Butterworth low-pass filter, 10Hz was likely the optimal filter cutoff frequency. The model achieved its best results when the acceleration data duration was 6.25s.

[0091] Example 2

[0092] An electronic device, such as Figure 10 As shown, it includes: a memory storing program code; and a processor connected to the memory, which, when the program code is executed by the processor, implements a balance and stability evaluation method based on a flexible IMU sensor. For a detailed description of the method, please refer to the corresponding description in the above method embodiments, which will not be repeated here.

[0093] Example 3

[0094] A computer-readable storage medium, such as Figure 11 As shown, it stores program instructions, which, when executed, implement a balance and stability evaluation method based on a flexible IMU sensor. For a detailed description of the method, please refer to the corresponding description in the above method embodiments; it will not be repeated here.

[0095] The number of devices and processing scale described herein are for the purpose of simplifying the description of the invention. Applications, modifications, and variations of the invention will be readily apparent to those skilled in the art.

[0096] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

[0097] The apparatus, electronic device, and non-volatile computer storage medium and method provided in the embodiments of this specification are corresponding. Therefore, the apparatus, electronic device, and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, electronic device, and non-volatile computer storage medium will not be repeated here.

[0098] Those skilled in the art will also know that, besides implementing the controller in the form of purely computer-readable program code, the same functions can be achieved by logically programming the method steps, making the controller take the form of logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers (PLCs), and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the devices included within it for implementing various functions can also be considered structures within that hardware component. Alternatively, the devices for implementing various functions can be considered as both software units implementing the method and structures within a hardware component.

[0099] The systems, apparatuses, or units described in the above embodiments can be implemented by computer chips or physical entities, or by products with certain functions. For ease of description, the above apparatuses are described separately as various units based on their functions. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in one or more software and / or hardware.

[0100] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0102] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0103] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0104] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0105] This specification may be described in the general context of computer-executable instructions, such as program units, that are executed by a computer. Generally, program units include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification may also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program units may reside in local and remote computer storage media, including storage devices.

[0106] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0107] The above description is merely an embodiment of this specification and is not intended to limit the scope of one or more embodiments of this specification. Various modifications and variations can be made to one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of one or more embodiments of this specification.

Claims

1. A method for evaluating the balance and stability of a flexible IMU sensor, characterized in that, Includes the following steps: Acceleration data was recorded during the subjects' execution of the Lobo test procedure to obtain the acceleration data to be predicted; The acceleration data is processed; The processed acceleration data is then filtered and normalized. Extract features from the processed acceleration data; The extracted features are input into the trained neural network model to predict the index values ​​of the corresponding force measuring platform equipment. The training of the neural network model includes the following steps: Record acceleration data and force table equipment parameters during the subject's execution of the Lobo test procedure; The acceleration data is processed; The parameters of the force measuring table equipment are used as the indicators to be predicted. The processed acceleration data is combined with the index to be predicted to form a new sample; Filter and normalize all samples; The healthy samples and balanced risk samples are divided by whether the true value of the indicator to be predicted is within the healthy range, and the number of positive and negative samples is balanced. Extract features from healthy samples and balanced risk samples; The extracted features are input into the neural network model for training; The indicators to be predicted are the elliptical area and the swing path length output by the force measuring table equipment. The process of inputting the extracted features into the neural network model for training includes the following steps: Calculate the root mean square error and the COD correlation parameter, and find the optimal weights for the root mean square error and the COD correlation parameter through a grid search method; The optimal weights are normalized and used as weights for calculating the overall loss; wherein, the COD correlation parameter is calculated by linear fitting prediction index and the true value output by the force measuring table equipment; During training, the fitting slope is controlled within the first quadrant by controlling the weights, and the mean squared error loss weights are decayed by annealing strategy. For each weight, the model is trained, and the optimal weight combination is found by using an early stopping strategy. The model with the optimal parameters is then used to output the index values ​​of the force measuring platform equipment. The fit is evaluated by calculating the fitting parameters between the predicted and true values, the prediction classification accuracy, precision, recall, and F1 score.

2. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 1, characterized in that: The acceleration data is the linear acceleration of the motion within a fixed three-dimensional frame measured by a triaxial accelerometer.

3. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 2, characterized in that: The linear acceleration includes both motion and gravitational components.

4. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 1, characterized in that: The actions performed by the subjects in the Lobo test procedure include closing their eyes, bending over, and opening their eyes.

5. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 4, characterized in that: The processing of the acceleration data includes the following steps: Acceleration data for both closed and open eyes are retained; After removing data during the action marking period, start and end data, and error data, the data with a preset duration in the middle of each state is retained and divided into multiple segments.

6. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 1, characterized in that: The step of combining the processed acceleration data with the index to be predicted to form a new sample includes the following steps: Acceleration data for each closed-eye and open-eye phase, along with the indicators to be predicted, are combined into two new samples.

7. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 1, characterized in that: The process of dividing healthy samples and balanced risk samples based on whether the true value of the indicator to be predicted is within the healthy range, and balancing the number of positive and negative samples, includes the following steps: The healthy samples and balanced risk samples are divided by the health reference experience range of the ellipse area index and the health reference experience range of the swing path length index, and the number of positive and negative samples is balanced. The number of samples that fall within or outside the health reference range of the ellipse area index and the swing path length index are respectively balanced according to the ellipse area index and the swing path length index.

8. The balance and stability evaluation method based on a flexible IMU sensor as described in claim 7, characterized in that: The healthy reference range for the ellipse area index is 1.0 to 2.

96. The healthy reference range for the swing path length index is 1.0 to 1.

73.

9. The method for evaluating the balance and stability based on a flexible IMU sensor as described in claim 1, characterized in that: The extracted features include the mean, standard deviation, maximum, minimum, average peak value, average peak spacing, JERK parameters, root mean square, centroid frequency, mean / variance of acceleration spectrum slices, and whether the patient is a stroke patient.

10. An electronic device, characterized in that, include: A memory that stores program code; A processor connected to the memory, which, when the program code is executed by the processor, implements the method as described in any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that, It stores program instructions that, when executed, implement the method as described in any one of claims 1 to 9.

Images