WO2021184792A1

WO2021184792A1 - Dynamic balance assessment method and apparatus, and device and medium

Info

Publication number: WO2021184792A1
Application number: PCT/CN2020/129188
Authority: WO
Inventors: 吴剑煌; 刘旭辉
Original assignee: 中国科学院深圳先进技术研究院
Priority date: 2020-03-19
Filing date: 2020-11-16
Publication date: 2021-09-23
Also published as: CN111403039A

Abstract

A dynamic balance assessment method and apparatus, and a device and a medium. The method comprises: acquiring data to be assessed, and preprocessing the data to be assessed, so as to obtain standardized assessed data (S110); inputting the standardized assessed data into a pre-trained balance assessment model to obtain an output result of the balance assessment model (S120), wherein the balance assessment model is a time-series-based neural network model; and determining an assessment result according to the output result, and then outputting the assessment result (S130). By means of the dynamic balance assessment method, associated features in data to be assessed are deeply mined by means of taking a time-series-based neural network model as a balance assessment model, such that the accuracy of dynamic balance assessment is improved.

Description

A dynamic balance evaluation method, device, equipment and medium

Technical field

The embodiment of the present invention relates to the technical field of balance assessment, and in particular to a dynamic balance assessment method, device, equipment and medium.

Background technique

Fall injuries are one of the important factors that seriously endanger the health of middle-aged and elderly people. The main cause of falls is generally the poor balance of the middle-aged and the elderly or the deterioration of the body's muscle strength. Therefore, monitoring the balance problems of middle-aged and elderly people, and giving balance and muscle strength training early and timely, play an important role in preventing middle-aged and elderly people from falling.

Current methods for measuring human body dynamic balance are usually implemented based on statistical methods, and the evaluation results corresponding to the measurement data are obtained through data analysis and statistics on the measurement data. However, dynamic balance evaluation based on statistical methods relies on specific algorithms for data analysis and statistics, and the evaluation results are inaccurate.

Summary of the invention

The embodiments of the present invention provide a dynamic balance evaluation method, device, equipment, and medium, so as to improve the accuracy of dynamic balance evaluation.

In the first aspect, an embodiment of the present invention provides a dynamic balance evaluation method, including:

Obtain the data to be evaluated, preprocess the data to be evaluated, and obtain standardized evaluation data;

Input the standardized assessment data into the pre-trained balance assessment model to obtain the output result of the balance assessment model, where the balance assessment model is a neural network model based on time series;

Determine the evaluation result according to the output result, and output the evaluation result.

In the second aspect, an embodiment of the present invention also provides a dynamic balance evaluation device, including:

The standardized data module is used to obtain the data to be evaluated, and preprocess the data to be evaluated to obtain standardized evaluation data;

The model result acquisition module is used to input standardized evaluation data into the pre-trained balance evaluation model to obtain the output result of the dynamic balance evaluation, where the balance evaluation model is a neural network model based on time series;

The evaluation result output module is used to determine the evaluation result according to the output result and output the evaluation result.

In the third aspect, an embodiment of the present invention also provides a computer device, and the device includes:

One or more processors;

Storage device for storing one or more programs;

When one or more programs are executed by one or more processors, the one or more processors implement the dynamic balance evaluation method provided by any embodiment of the present invention.

In a fourth aspect, an embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the dynamic balance evaluation method as provided in any embodiment of the present invention is implemented.

In the embodiment of the present invention, standardized evaluation data is obtained by obtaining the data to be evaluated and preprocessing the data to be evaluated; inputting the standardized evaluation data into a pre-trained balance evaluation model to obtain the output result of the balance evaluation model, wherein the balance evaluation model It is a neural network model based on time series; the evaluation result is determined according to the output result, and the evaluation result is output. The neural network model based on the time series is used as a balance evaluation model to deeply mine the associated features in the data to be evaluated, which improves the dynamic balance evaluation Accuracy.

Description of the drawings

FIG. 1 is a flowchart of a dynamic balance evaluation method provided by Embodiment 1 of the present invention;

Figure 2a is a flowchart of a dynamic balance assessment method provided in the second embodiment of the present invention;

Fig. 2b is a schematic diagram of a basic neural network structure of an LSTM provided by the second embodiment of the present invention;

Fig. 2c is a schematic diagram of a chain structure of an LSTM provided by the second embodiment of the present invention;

2d is a schematic diagram of a state of an LSTM cell provided by the second embodiment of the present invention;

Fig. 2e is a schematic diagram of an LSTM forget gate provided by the second embodiment of the present invention;

Fig. 2f is a schematic diagram of an LSTM input gate provided by the second embodiment of the present invention;

Fig. 2g is a schematic diagram of an LSTM input gate provided by the second embodiment of the present invention;

FIG. 3a is a schematic diagram of a construction process of a balance evaluation model provided by Embodiment 3 of the present invention;

Fig. 3b is a schematic structural diagram of a balance evaluation model provided by the third embodiment of the present invention;

FIG. 3c is a schematic diagram of a data structure type output by an output layer provided by Embodiment 3 of the present invention; FIG.

4 is a schematic structural diagram of a dynamic balance evaluation device provided by the fourth embodiment of the present invention;

Fig. 5 is a schematic structural diagram of a computer device according to the fifth embodiment of the present invention.

Detailed ways

The present invention will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described here are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for ease of description, the drawings only show a part of the structure related to the present invention, but not all of the structure.

Example one

Fig. 1 is a flowchart of a dynamic balance assessment method provided by Embodiment 1 of the present invention. This embodiment can be applied to situations when dynamic balance evaluation is performed. The method can be executed by a dynamic balance evaluation device, which can be implemented by software and/or hardware, for example, the dynamic balance evaluation device can be configured in a computer device. As shown in Figure 1, the method includes:

S110. Obtain the data to be evaluated, and preprocess the data to be evaluated to obtain standardized evaluation data.

In this embodiment, the data to be evaluated may be dynamic balance test data of the subject. Among them, the dynamic balance test method is not limited here. Optionally, the dynamic balance test method can be a standing test, a balance beam test, a stepping test, and other test methods. Correspondingly, the data to be evaluated may be test data obtained by testing the subject using any one of the above-mentioned test methods.

In the dynamic balance test, the testee generally needs to complete the setting action, and collect the data of the setting index during the process of the testee performing the setting action, and obtain the dynamic balance test data containing multiple index collection data. In an embodiment of the present invention, the data to be evaluated includes index collection data under a set action, and preprocessing the data to be evaluated to obtain standardized evaluation data includes: normalizing the index collection data to obtain a standardized evaluation data. In the dynamic balance test, there may be multiple indexes corresponding to an action, and the dimensions of the test data obtained by different indexes may be different. In order to improve the accuracy of the balance evaluation results, it is necessary to normalize the test data of different indicators, and transform the dimensional test data into dimensionless data to obtain standardized evaluation data. Exemplarily, the StandarScaler method in the sklearn.preprocessing library may be used to take the index collection data corresponding to each set action as a unit, and scale the index collection data corresponding to each set action to obtain standardized evaluation data.

In an embodiment of the present invention, the data to be evaluated may be dynamic balance test data of the subject's standing test. In the standing test, the set action includes at least one of standing with eyes open on both feet, standing with eyes closed on both feet, standing on one foot with eyes open, and standing on one foot with eyes closed. The subject needs to complete at least one of the above actions to obtain the corresponding dynamic balance test data. In order to be able to more accurately and comprehensively evaluate the testee’s balance ability, standing with eyes open on both feet, standing with eyes closed on both feet, standing with eyes open on one foot, and standing on one foot with eyes closed as the set actions of the standing test, by combining the above The indicators of the four movements collect data to evaluate the balance ability of the subjects.

On the basis of the above scheme, the index collection data under the setting action includes the peripheral area of the eye open under the setting action, the trajectory length per unit area, the total trajectory length of shaking, the trajectory length in the first direction, the trajectory length in the second direction, and the first trajectory. At least one of the average center displacement in the direction, the average center displacement in the second direction, the maximum swing diameter in the first direction, the maximum swing diameter in the second direction, the average speed in the first direction, the average speed in the second direction, and the average swing speed.

When the data to be evaluated is the dynamic balance test data of the subject’s standing test, for each set action, the corresponding index collection data includes the peripheral area of the eyes, the length of the trajectory per unit area, the length of the total trajectory of shaking, and the trajectory in the first direction. Long, track length in the second direction, average center displacement in the first direction, average center displacement in the second direction, maximum swing diameter in the first direction, maximum swing diameter in the second direction, average speed in the first direction, average speed in the second direction, At least one of the average swing speeds. Wherein, the first direction and the second direction are perpendicular, the first direction may be a horizontal direction, and the second direction may be a vertical direction.

In order to be able to more accurately and comprehensively evaluate the subject’s balance ability, the peripheral area of the open eyes, the track length per unit area, the total track length of shaking, the track length in the first direction, the track length in the second direction, and the average center displacement in the first direction can be changed. , The average center displacement in the second direction, the maximum swing diameter in the first direction, the maximum swing diameter in the second direction, the average speed in the first direction, the average speed in the second direction, and the average swing speed are used as test indicators. Data is collected from two indicators (each set action corresponds to 12 indicators to collect data), and the testee’s balance ability is evaluated. Specifically, the 48 indicators corresponding to the four setting actions are shown in Table 1.

Table 1

S120. Input the standardized evaluation data into the pre-trained balance evaluation model to obtain an output result of the balance evaluation model.

The testee’s dynamic balance test data has a certain relevance. Taking the 48-dimensional data obtained from the standing test as an example, there is a temporal relevance between the set actions. Based on this, the 48-dimensional index collection data is also related to each other. There is a significant correlation, so there is also a certain correlation between the 48-dimensional standardized evaluation data based on the 48-dimensional index collection data. Therefore, a neural network model based on time series can be used as a balance evaluation model to mine the correlation characteristics between standardized evaluation data, and the standardized data can be input into the trained balance evaluation model to obtain the output result of the balance evaluation model.

Optionally, the output result of the balance evaluation model may be the category with the highest probability, or may be each category and the probability corresponding to each category. Among them, the categories in the output result of the balance assessment model may include low balance ability, medium balance ability, and high balance ability. The specific meaning of low balance ability, middle balance ability and high balance ability can refer to medical standards.

S130: Determine an evaluation result according to the output result, and output the evaluation result.

When the output result is the category with the highest probability, the output result can be directly output as the evaluation result. When the output result is each category and the probability corresponding to each category, the evaluation result can be determined and output according to the probability corresponding to each category.

In an embodiment of the present invention, determining the evaluation result according to the output result includes: obtaining the probability value of each category in the output result, and selecting a category from each category as the evaluation result according to the probability value of each category. Specifically, when the output result is each category and the probability corresponding to each category, the category corresponding to the maximum probability value is used as the evaluation result. Exemplarily, if the probability of category 1 (low balance ability) is 0.6, the probability of category 2 (medium balance ability) is 0.3, and the probability of category 3 (high balance ability) is 0.1, then category 1-low balance ability is taken as The evaluation result is output.

Example two

Fig. 2a is a flowchart of a dynamic balance assessment method provided in the second embodiment of the present invention. In this embodiment, on the basis of the foregoing embodiment, an operation of training the balance evaluation model is added. As shown in Figure 2a, the method includes:

S210. Obtain sample evaluation data and labels corresponding to the sample evaluation data, and generate training sample data according to the sample evaluation data and the labels corresponding to the sample evaluation data.

In this embodiment, the sample evaluation data may be standardized data obtained after preprocessing the sample test data, and the label corresponding to the sample evaluation data may be realized by manual labeling. Exemplarily, the doctor may label the sample evaluation data to ensure the accuracy and professionalism of the label of the sample evaluation data. Among them, the method of obtaining sample evaluation data from sample test data can refer to the method of obtaining standardized evaluation data from the data to be evaluated in the foregoing embodiment, which will not be repeated here.

It is understandable that the dynamic balance test method corresponding to the sample test data is the same as the dynamic balance test method corresponding to the original behavior data to be evaluated. Exemplarily, if the sample test data is the dynamic balance test data of the standing test, the data to be evaluated should also be the dynamic balance test data of the same standing test.

S220. Use the training sample data to train the pre-built balance evaluation model to obtain a trained balance evaluation model.

After obtaining the training sample data, use the training sample data to train the pre-built balance evaluation model to obtain a trained balance evaluation model.

In this embodiment, the pre-built balance evaluation model is a neural network model based on time series. Optionally, a balance evaluation model can be constructed based on a sequence model (such as a recurrent neural network or a variant of a recurrent neural network). In one embodiment, the pre-built balance evaluation model includes a feature extraction layer and a fully connected layer, where the feature extraction layer includes at least one layer of feature extraction network, and the feature extraction network is a cyclic neural network or a variant of the cyclic neural network. Among them, the variants of the recurrent neural network may be a long short-term memory network (Long Short-Term Memory, LSTM), a gated recurrent unit (Gated Recurrent Unit, GRU), and so on.

Specifically, the balance evaluation model includes an input layer, a feature extraction layer, a fully connected layer, and an output layer. Wherein, the feature extraction layer includes at least one layer of feature extraction network, and the number of layers of the fully connected layer may be at least one layer. The more layers of the feature extraction network and fully connected layers, the stronger the learning ability, but too many layers are likely to cause overfitting in the training process. Preferably, a three-layer LSTM can be selected as the feature extraction layer, the output of the first layer of the three-layer LSTM is used as the input of the second layer, and the output of the second layer is used as the input of the third layer. Through the combination of the three-layer LSTM network, It can further improve the learning ability of the model, dig deeper potential information, and avoid overfitting during model training.

LSTM is a variant of a special recurrent neural network. Fig. 2b is a schematic diagram of a basic neural network structure of an LSTM provided in the second embodiment of the present invention. In Figure 2b, the σ operation represents the sigmoid() function, the tanh operation represents the tanh() function, the × operation represents Pointwise multiplication, the + operation represents addition, X ₁ represents the data input at time _{t 1} _{, and H 1} represents the LSTM neuron’s output, C ₀ represents the cellular state, an output indicates the H _{_{_0, f t, i t, β}} , o t represent respective output of each calculation process.

LSTM is similar to the recurrent neural network. In the actual operation process, the network structure shown in Figure 2b is continuously repeated according to the time series of the input data, which is represented as a chain structure over time. A schematic diagram of the chain structure of a LSTM. Fig. 2c schematically shows the chain structure of the LSTM according to the time series of the input data, where X1, X2, and X3 represent the input of the input data in different time periods, H1, H2 and H3 represent the output at different moments.

The key of LSTM is the cell state. Figure 2d is a schematic diagram of the LSTM cell state provided in the second embodiment of the present invention. It can be seen from the dotted line in Figure 2d that the cell state runs through the entire neuron of the LSTM . LSTM adds or deletes information to the cell state through various "gate" structures. LSTM mainly includes three gate operations: forget gate, input gate, and output gate.

Figure 2e is a schematic diagram of an LSTM forgetting gate provided by the second embodiment of the present invention. The dotted line in Figure 2e represents the forgetting gate. The forgetting gate controls _{what information is thrown away in the cell state C 0. The} specific mathematical expression of the forgetting gate is: f _t =σ(W _f ·[H ₀ ,X ₁ ]+b _f ), where W _f and b _f represent the coefficient matrix and offset of the calculation process, respectively, and the _{final value of f t is} (0,1), 0 Means that the cell state information is completely lost, and 1 means that the cell state information is completely retained.

Figure 2f is a schematic diagram of an LSTM input gate provided by the second embodiment of the present invention. The dotted line in Figure 2f represents the input gate. The input gate determines which information is stored in the cell state. The specific mathematical expression of the input gate is: i _t =σ(W _i ·[H ₀ ,X ₁ ]+b _i ), β=tanh(W _c ·[H ₀ ,X ₁ ]+b _c ), where W _i and W _c represent the coefficient matrix of the calculation process , B _i and b _c are biases. After completing the calculation of the forget gate and the input gate, update C ₀ to C ₁ : C ₁ =C ₀ ×f _t +i _t ×β.

Fig. 2g is a schematic diagram of an LSTM input gate provided by the second embodiment of the present invention. The dotted line in Figure 2g represents the output gate. The output gate determines the output. The specific expression of the output gate is: O _t =σ(W ₀ ·[H ₀ ,X ₁ ]+b ₀ ), H ₁ ＝O _t ×tanh(C ₁ ), where W ₀ and b ₀ respectively represent the coefficient matrix and offset of the calculation process.

S230. Obtain the data to be evaluated, and preprocess the data to be evaluated to obtain standardized evaluation data.

S240. Input the standardized evaluation data into the pre-trained balance evaluation model to obtain an output result of the balance evaluation model.

S250: Determine an evaluation result according to the output result, and output the evaluation result.

The embodiment of the present invention generates training sample data according to the sample evaluation data and the label corresponding to the sample evaluation data by acquiring the sample evaluation data and the label corresponding to the sample evaluation data; using the training sample data to train the pre-built balance evaluation model, the training is obtained. The balance evaluation model is obtained by using the neural network model based on time series as the balance evaluation model, so that the construction of the balance evaluation model takes into account the correlation characteristics in the dynamic balance measurement data, and improves the dynamics of the balance evaluation model. Balance the accuracy of the assessment.

Example three

On the basis of the above-mentioned embodiment, this embodiment provides a preferred embodiment. In this embodiment, constructing a balance evaluation model based on LSTM is taken as an example to illustrate the construction and training of the balance evaluation model.

Fig. 3a is a schematic diagram of the construction process of a balance evaluation model provided by the third embodiment of the present invention. Fig. 3a schematically shows the process of constructing a balanced evaluation model based on the long and short-term memory network. As shown in Figure 3a, the construction of the balance assessment model mainly includes:

S310. Determine the dimension of the input data.

In the embodiment of the present invention, the German Bismarck (Bismarck International Group Inc) dynamic and static balancer can be used to measure the measurement data obtained by the subject, where the measurement data mainly include: standing with eyes open with both feet and eyes closed with both feet The 48-dimensional indicators of standing, standing on one foot with eyes open, and standing on one foot with eyes closed are shown in Table 1. That is, the 48-dimensional measurement data of the four actions is acquired as the source data.

S320. Data preprocessing.

The dimensions of the 48-dimensional data included in the measurement data are different, which is not conducive to the training of the deep learning model, and may cause the model to converge too slowly. In order to avoid the occurrence of the above phenomenon, it is necessary to standardize the original 48-dimensional data and unify the 48-dimensional data into the set data range. Optionally, the feature scaling (StandarScaler) method in the sklearn.preprocessing library can be used to scale the 48-dimensional data in columns (refer to the columns in Table 1), and the scaled data has a mean value of 0 and a variance of 1.

S330, three-layer LSTM deep neural network construction.

Since the 48-dimensional dynamic balance measurement indicators come from four actions, and there is a close correlation between the four actions, the neural network model training based on time series can get better training results.

In this embodiment, a three-layer LSTM neural network model can be used to construct a balanced evaluation model. Compared with a one-layer LSTM neural network model, a three-layer LSTM can dig deeper data information, and the learning ability is also better than that of a layer of neural network. The network model is good, and the choice of three-layer LSTM is not just because there are more layers, but also because the more layers, the greater the risk of overfitting. Fig. 3b is a schematic structural diagram of a balance evaluation model provided by the third embodiment of the present invention. As shown in FIG. 3b, the balance evaluation model includes an input layer 310, a three-layer LSTM network 320, a fully connected layer 330, and an output layer 340. Among them, the input layer 310 combines the 48-dimensional indicators according to four actions (each action has 12-dimensional indicators), respectively corresponding to M1, M2, M3, and M4 in the figure. The three-layer LSTM network 320 includes a first-layer LSTM network 321, a second-layer LSTM network 322, and a third-layer LSTM network 323. The output of the first-layer LSTM network 321 is used as the input of the second-layer LSTM network 322. The output of the LSTM network 322 is used as the input of the third-layer LSTM network 323. The combination of the three-layer LSTM network can further improve the learning ability of the model and dig deeper potential information. The fully connected layer 330 includes a first fully connected layer 331 and a second fully connected layer 332. The two fully connected layers fix the output of the neural network to three dimensions. Each dimension represents the probability of a classification result. The output layer 340 is used for Output the probability of each classification result.

FIG. 3c is a schematic diagram of a data structure type output by an output layer provided by Embodiment 3 of the present invention. As shown in Figure 3c, the first dimension represents low balance ability with a probability of 0.6, the second dimension represents medium balance ability with a probability of 0.3, and the third dimension represents high balance ability with a probability of 0.1, and the sum of the three probabilities is 1. . Among them, the probability of low balance ability is the largest, which means that the final result is low balance ability.

S340. Model training.

Optionally, a large amount of labeled dynamic balance data can be collected. The data format is: 48-dimensional balance index + label (balance level: low, medium, and high); the data is randomly divided into training set and test set, and the training set is used to train and predict balance Evaluate the model, use the test set to test the trained balance evaluation model.

S350. Model checking.

After using the training set to train the balance evaluation model, in order to verify the effectiveness of the balance evaluation model, the test set can be used to verify the balance evaluation model. Optionally, a variety of verification methods can be used to verify the effectiveness of the model, such as accuracy, sensitivity, ROC curve and other methods to verify the effectiveness of the model. If the detection result is that the effectiveness of the model is low, the model can be adjusted by adjusting the model structure (such as adjusting the number of layers of the basic network in the feature extraction layer), obtaining a larger number of samples for retraining, and so on.

It should be noted that the method provided in the embodiments of the present invention has been verified by experiments. By collecting more than 6000 physical examination data of the balance ability of middle-aged and elderly patients, the data set is divided into training and testing sets, and cross-validation and accuracy evaluation are adopted. The test results show that the accuracy rate of the evaluation model on the verification set is 86.67%, and the accuracy rate on the test set is 86.32%. If the amount of data for model training is further increased, the accuracy of the model will be further improved. It can be seen that the method for constructing a balance evaluation model for dynamic balance evaluation provided by the embodiment of the present invention has a higher accuracy rate.

In this embodiment, a deep neural network is used to construct a balance evaluation model. The model has better universal ability and improves the accuracy of balance evaluation. In addition, considering the correlation between the four actions in the evaluation data, a neural network model based on time series is used as The balanced evaluation model can dig out the potential relationships among the four actions, making the evaluation results more accurate.

Example four

Fig. 4 is a schematic structural diagram of a dynamic balance evaluation device provided in the fourth embodiment of the present invention. The dynamic balance evaluation device can be implemented by software and/or hardware. For example, the dynamic balance evaluation device can be configured in a computer device. As shown in Fig. 4, the device includes a standardized data module 410, a model result acquisition module 420, and an evaluation result output module 430, where:

The standardized data module 410 is used to obtain the data to be evaluated, preprocess the data to be evaluated, and obtain standardized evaluation data;

The model result acquisition module 420 is configured to input standardized evaluation data into a pre-trained balance evaluation model to obtain an output result of the dynamic balance evaluation, where the balance evaluation model is a neural network model based on time series;

The evaluation result output module 430 is configured to determine the evaluation result according to the output result, and output the evaluation result.

The embodiment of the present invention obtains the data to be evaluated through the standardized data module, and preprocesses the data to be evaluated to obtain standardized evaluation data; the model result acquisition module inputs the standardized evaluation data into the pre-trained balance evaluation model to obtain the output of the balance evaluation model As a result, the balance evaluation model is a neural network model based on time series; the evaluation result output module determines the evaluation result according to the output result, and outputs the evaluation result, and uses the neural network model based on time series as the balance evaluation model to deeply explore the pending evaluation The associated features in the data improve the accuracy of dynamic balance assessment.

Optionally, on the basis of the above solution, the device further includes a model training module for:

Before inputting the standardized evaluation data into the pre-trained balance evaluation model, obtain the sample evaluation data and the label corresponding to the sample evaluation data, and generate training sample data according to the sample evaluation data and the label corresponding to the sample evaluation data;

Use the training sample data to train the pre-built balance evaluation model to obtain a trained balance evaluation model.

Optionally, on the basis of the above solution, the pre-built balance evaluation model includes a feature extraction layer and a fully connected layer, where the feature extraction layer includes at least one layer of feature extraction network, and the feature extraction network is a cyclic neural network or a cyclic neural network Variants.

Optionally, on the basis of the foregoing solution, the evaluation result output module 430 is specifically configured to:

Obtain the probability value of each category in the output result, and select a category from each category as the evaluation result according to the probability value of each category.

Optionally, based on the above solution, the data to be evaluated includes index collection data under a set action, and the standardized data module 410 is specifically used for:

Normalize the index collection data to obtain standardized evaluation data.

Optionally, on the basis of the foregoing solution, the set action includes at least one of standing with eyes open on both feet, standing with eyes closed on both feet, standing with eyes open on one foot, and standing on one foot with eyes closed.

Optionally, on the basis of the above solution, the index collection data under the setting action includes the peripheral area of the eye open, the track length per unit area, the total trajectory length of shaking, the trajectory length in the first direction, and the trajectory in the second direction under the setting action. Length, the average center displacement in the first direction, the average center displacement in the second direction, the maximum swing diameter in the first direction, the maximum swing diameter in the second direction, the average speed in the first direction, the average speed in the second direction, and the average swing speed at least one.

The dynamic balance evaluation device provided by the embodiment of the present invention can execute the dynamic balance evaluation method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for the execution method.

Example five

Fig. 5 is a schematic structural diagram of a computer device according to the fifth embodiment of the present invention. Figure 5 shows a block diagram of an exemplary computer device 512 suitable for implementing embodiments of the present invention. The computer device 512 shown in FIG. 5 is only an example, and should not bring any limitation to the function and application scope of the embodiment of the present invention.

As shown in FIG. 5, the computer device 512 is represented in the form of a general-purpose computing device. The components of the computer device 512 may include, but are not limited to: one or more processors 516, a system memory 528, and a bus 518 connecting different system components (including the system memory 528 and the processor 516).

The bus 518 represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor 516, or a local bus using any bus structure among multiple bus structures. For example, these architectures include, but are not limited to, industry standard architecture (ISA) bus, microchannel architecture (MAC) bus, enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and peripheral component interconnection ( PCI) bus.

Computer device 512 typically includes a variety of computer system readable media. These media may be any available media that can be accessed by the computer device 512, including volatile and nonvolatile media, removable and non-removable media.

The system memory 528 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 530 and/or cache memory 532. The computer device 512 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. For example only, the storage device 534 may be used to read and write non-removable, non-volatile magnetic media (not shown in FIG. 5, usually referred to as a "hard drive"). Although not shown in FIG. 5, a disk drive for reading and writing to removable non-volatile disks (such as "floppy disks") and a removable non-volatile disk (such as CD-ROM, DVD-ROM) can be provided. Or other optical media) read and write optical disc drives. In these cases, each drive may be connected to the bus 518 through one or more data medium interfaces. The memory 528 may include at least one program product, the program product having a set (for example, at least one) of program modules, and these program modules are configured to perform the functions of the embodiments of the present invention.

A program/utility tool 540 having a set of (at least one) program module 542 may be stored in, for example, the memory 528. Such program module 542 includes but is not limited to an operating system, one or more application programs, other program modules, and program data Each of these examples or some combination may include the implementation of a network environment. The program module 542 generally executes the functions and/or methods in the described embodiments of the present invention.

The computer device 512 can also communicate with one or more external devices 514 (such as a keyboard, pointing device, display 524, etc.), and can also communicate with one or more devices that enable a user to interact with the computer device 512, and/or communicate with Any device (such as a network card, modem, etc.) that enables the computer device 512 to communicate with one or more other computing devices. Such communication can be performed through an input/output (I/O) interface 522. In addition, the computer device 512 may also communicate with one or more networks (for example, a local area network (LAN), a wide area network (WAN), and/or a public network, such as the Internet) through the network adapter 520. As shown in the figure, the network adapter 520 communicates with other modules of the computer device 512 through the bus 518. It should be understood that although not shown in the figure, other hardware and/or software modules can be used in conjunction with the computer device 512, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives And data backup storage system, etc.

The processor 516 executes various functional applications and data processing by running programs stored in the system memory 528, for example, to implement the dynamic balance evaluation method provided by the embodiment of the present invention, the method includes:

Of course, those skilled in the art can understand that the processor can also implement the technical solution of the dynamic balance evaluation method provided by any embodiment of the present invention.

Example Six

The sixth embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the dynamic balance evaluation method provided in the embodiment of the present invention is implemented, and the method includes:

Of course, in a computer-readable storage medium provided by an embodiment of the present invention, the computer program stored thereon is not limited to the above method operations, and can also perform related operations of the dynamic balance evaluation method provided by any embodiment of the present invention.

The computer storage medium of the embodiment of the present invention may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or a combination of any of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), Erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, apparatus, or device.

The computer-readable signal medium may include a data signal propagated in baseband or as a part of a carrier wave, and computer-readable program code is carried therein. This propagated data signal can take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The computer-readable signal medium may also be any computer-readable medium other than the computer-readable storage medium. The computer-readable medium may send, propagate, or transmit the program for use by or in combination with the instruction execution system, apparatus, or device .

The program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

The computer program code used to perform the operations of the present invention can be written in one or more programming languages or a combination thereof. The programming languages include object-oriented programming languages—such as Java, Smalltalk, C++, and also conventional Procedural programming language-such as "C" language or similar programming language. The program code can be executed entirely on the user's computer, partly on the user's computer, executed as an independent software package, partly on the user's computer and partly executed on a remote computer, or entirely executed on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (for example, using an Internet service provider to pass Internet connection).

Note that the above are only the preferred embodiments of the present invention and the applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments and substitutions can be made to those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention. The scope of is determined by the scope of the appended claims.

Claims

A dynamic balance assessment method, which is characterized in that it includes:

Acquiring data to be evaluated, preprocessing the data to be evaluated, and obtaining standardized evaluation data;

Inputting the standardized assessment data into a pre-trained balance assessment model to obtain an output result of the balance assessment model, wherein the balance assessment model is a neural network model based on a time series;

The evaluation result is determined according to the output result, and the evaluation result is output.
The method according to claim 1, wherein before inputting the standardized evaluation data into a pre-trained balance evaluation model, the method further comprises:

Acquiring sample evaluation data and a label corresponding to the sample evaluation data, and generating training sample data according to the sample evaluation data and the label corresponding to the sample evaluation data;

Use the training sample data to train the pre-built balance evaluation model to obtain a trained balance evaluation model.
The method according to claim 2, wherein the pre-built balance evaluation model includes a feature extraction layer and a fully connected layer, wherein the feature extraction layer includes at least one layer of feature extraction network, and the feature extraction network It is a cyclic neural network or a variant of a cyclic neural network.
The method according to claim 1, wherein the determining an evaluation result according to the output result comprises:

The probability value of each category in the output result is obtained, and a category is selected from each category as the evaluation result according to the probability value of each category.
The method according to claim 1, wherein the data to be evaluated includes index collection data under a set action, and the preprocessing of the data to be evaluated to obtain standardized evaluation data comprises:

The index collection data is normalized to obtain the standardized evaluation data.
The method according to claim 5, wherein the setting action comprises at least one of standing with eyes open on both feet, standing with eyes closed on both feet, standing with eyes open on one foot, and standing on one foot with eyes closed.
The method according to claim 6, wherein the index collection data under the setting action includes the peripheral area of the eye open, the length of the trajectory per unit area, the length of the total trajectory of shaking, and the trajectory in the first direction under the setting action. Long, the trajectory length in the second direction, the average center displacement in the first direction, the average center displacement in the second direction, the maximum swing diameter in the first direction, the maximum swing diameter in the second direction, the average speed in the first direction, the average speed in the second direction, At least one of the average swing speeds.
A dynamic balance assessment device, characterized in that it comprises:

The standardized data module is used to obtain the data to be evaluated, and preprocess the data to be evaluated to obtain standardized evaluation data;

A model result acquisition module, configured to input the standardized evaluation data into a pre-trained balance evaluation model to obtain an output result of the dynamic balance evaluation, wherein the balance evaluation model is a neural network model based on a time series;

The evaluation result output module is used to determine the evaluation result according to the output result, and output the evaluation result.
A computer device, characterized in that the device includes:

One or more processors;

Storage device for storing one or more programs;

When the one or more programs are executed by the one or more processors, the one or more processors implement the dynamic balance evaluation method according to any one of claims 1-7.
A computer-readable storage medium with a computer program stored thereon, wherein the program is executed by a processor to implement the dynamic balance evaluation method according to any one of claims 1-7.