Gait recognition method and device based on intelligent wearable sensor and storage medium

Abstract

The invention relates to the technical field of biological recognition, and provides a gait recognition method and device based on an intelligent wearable sensor and a storage medium, and the method comprises the steps: obtaining sensing data of a training target in seven dimensions through the intelligent wearable sensor, and determining original data according to the sensing data; performing data preprocessing on the original data to obtain training data; meanwhile, constructing a dual-channel hybrid neural network model, and training the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed; and performing gait recognition on a to-be-detected target based on the gait recognition model, and outputting a corresponding recognition result. The gait feature of the target person can be obtained through the gait recognition model, then the recognition information of the target person is obtained, and the method can be suitable for person information collection of all scenes.

Description

Technical Field

[0001] The present invention relates to the field of biometric technologies, and particularly to a gait recognition method, device, and computer-readable storage medium based on an intelligent wearable sensor. Background Art

[0002] The research on gait recognition has been a popular research field in recent years. Gait recognition technology integrates knowledge from disciplines such as kinematics, biology, and clinical medicine, has broad application prospects, and is of great significance to the development of technologies such as identity recognition, kinematic analysis, clinical medicine, and criminalistics.

[0003] Currently, gait recognition technologies can be classified into gait recognition methods based on inertial measurement units, computer vision, pressure sensors, and electromyographic signals according to the acquisition method of gait data. Some researchers fuse the above methods to improve the accuracy of gait recognition. Gait recognition based on pressure sensors collects information such as pressure, position, direction, and timestamp during gait through a pressure-sensitive floor. The pressure-sensitive floor is usually installed on the ground or a treadmill. However, this solution has high costs and disadvantages such as short measurement distance and poor flexibility. The acquisition of data is limited to the experimental site and it is difficult to be applied in real life.

[0004] In addition, gait recognition based on computer vision has been very popular in recent years. Thanks to the great success of deep learning algorithms in the field of image processing, gait recognition algorithms based on computer vision have also made great progress. However, gait recognition of such solutions is easily restricted by cameras and models. Cameras are usually fixedly placed, and the subjects must be within the camera shooting range. Moreover, it is easily affected by factors such as lighting conditions, perspective changes, occlusion, and clothing. In addition, the model training structure is simple, which affects the correct rate of activity recognition. Summary of the Invention

[0005] The present invention provides a gait recognition method, electronic device, and computer-readable storage medium based on an intelligent wearable sensor. Its main purpose is to obtain the gait characteristics of a target person based on a gait recognition model and perform gait recognition on it, which can be applied to scenarios such as identity recognition, kinematic analysis, clinical medicine, and criminalistics.

[0006] To achieve the above objective, the present invention provides a gait recognition method based on an intelligent wearable sensor, which is applied to a client. The method includes:

[0007] Applied to a client, characterized in that the method includes:

[0008] Obtain the sensing data of a training target in seven dimensions through an intelligent wearable sensor, and determine the original data according to the sensing data;

[0009] Perform data preprocessing on the original data to obtain training data; meanwhile,

[0010] Construct a dual-channel hybrid neural network model, and train the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed;

[0011] Perform gait recognition on the target to be detected based on the gait recognition model, and output the corresponding recognition result.

[0012] In addition, an optional technical solution is that the intelligent wearable sensor includes a fixed strap and an inertial sensor, a gyroscope, a tensile strain sensor, and a Bluetooth transmission module arranged on the fixed strap; the intelligent wearable sensor is arranged at the knee joint and / or ankle joint and / or hip joint of the training target through the fixed strap.

[0013] In addition, an optional technical solution is that the sensing data in the seven dimensions includes three-axis acceleration data collected by the inertial sensor, three-axis angular velocity data collected by the gyroscope, and tensile strain data collected by the tensile strain sensor.

[0014] In addition, an optional technical solution is to obtain the sensing data of the training target in seven dimensions as the original data through the intelligent wearable sensor, including:

[0015] Develop a supporting host computer program for the lower computer serial port data through a data acquisition module to perform data point sampling on the movements of different training targets at regular intervals, and obtain the sensing data;

[0016] Perform unified identity annotation on the sensing data through a data annotation module to determine the annotated data;

[0017] Summarize the annotated data of all training targets into an npy file through a dataset synthesis module, and replace the identity with the category to obtain the original data.

[0018] In addition, an optional technical solution is to perform data preprocessing on the original data to obtain training data, including:

[0019] Perform filtering and denoising processing on the original data;

[0020] Perform segmentation processing on the original data after filtering and denoising in a sliding window manner;

[0021] Perform normalization processing on the segmented original data to form the training data.

[0022] In addition, an optional technical solution is that the dual-channel hybrid neural network model includes a first channel, a second channel, and a multi-layer perception mechanism; among them,

[0023] The first channel includes a first convolutional neural network and a first recurrent neural network that are connected to each other;

[0024] The second channel includes a Fourier transform mechanism, a second convolutional neural network, and a second recurrent neural network that are connected to each other;

[0025] The outputs of the first recurrent neural network and the second recurrent neural network are combined and input into the multi-layer perceptron mechanism, and the recognition result is output through the multi-layer perceptron mechanism.

[0026] In addition, an optional technical solution is that the training data is divided into a training data set and a test data set according to a ratio of 4:1;

[0027] The training data set is used to train the dual-channel hybrid neural network model, and the test data set is used to test the trained dual-channel hybrid neural network model, and the gait recognition model is determined according to the test result.

[0028] In addition, an optional technical solution is that training the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed includes:

[0029] Training the dual-channel hybrid neural network model based on the training data until the loss function of the dual-channel hybrid neural network model is less than a preset threshold to form the gait recognition model; where

[0030] The loss function includes mean absolute error, mean square error, and cross-entropy function;

[0031] The expression formula of the mean absolute error is:

[0032]

[0033] The expression formula of the mean square error is:

[0034]

[0035] where n represents the number of the training data, y i represents the true value of the i-th training data, and y i p represents the predicted value of the i-th training data;

[0036] The expression formula of the cross-entropy function is:

[0037]

[0038] where y represents the true value of the input training data, Represents the predicted value corresponding to the said y.

[0039] To achieve the above object, the present invention further provides an electronic device, which includes: a memory and a processor. The memory includes a gait recognition program based on intelligent wearable sensors. When the gait recognition program based on intelligent wearable sensors is executed by the processor, the following steps are implemented:

[0040] Obtain the sensing data of the training target in seven dimensions through intelligent wearable sensors respectively, and determine the original data according to the sensing data;

[0041] Perform data preprocessing on the original data to obtain training data; at the same time,

[0042] Construct a dual-channel hybrid neural network model, and train the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed;

[0043] Perform gait recognition on the target to be detected based on the gait recognition model, and output the corresponding recognition result.

[0044] To achieve the above object, the present invention further provides a computer-readable storage medium. The computer-readable storage medium includes a gait recognition program based on intelligent wearable sensors. When the gait recognition program based on intelligent wearable sensors is executed by a processor, the steps of the gait recognition method based on intelligent wearable sensors as described above are implemented.

[0045] The gait recognition method, electronic device and computer-readable storage medium based on intelligent wearable sensors proposed by the present invention can detect the gait characteristics of the human body according to the actually collected sensing data, and then achieve gait recognition of the user based on this, with high recognition accuracy and wide application range. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the application environment of the preferred embodiment of the gait recognition method based on intelligent wearable sensors of the present invention;

[0047] Figure 2 It is a logic diagram of the gait recognition system based on intelligent wearable sensors according to an embodiment of the present invention;

[0048] Figure 3 It is a flowchart of the gait recognition method based on intelligent wearable sensors according to an embodiment of the present invention;

[0049] Figure 4 It is a schematic structure of the dual-channel hybrid neural network model according to an embodiment of the present invention.

[0050] The realization, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0051] It should be understood that the specific embodiments described herein are merely used to explain the present invention and are not used to limit the present invention.

[0052] The present invention provides a gait recognition method based on an intelligent wearable sensor, which is applied to an electronic device (or client) 1. Refer to Figure 1 As shown, it is a schematic diagram of an application environment of a preferred embodiment of the gait recognition method based on an intelligent wearable sensor according to the present invention.

[0053] In this embodiment, the electronic device 1 may be a terminal device with computing functions such as a server, a smart phone, a tablet computer, a portable computer, a desktop computer, etc.

[0054] The electronic device 1 includes: a processor 12, a memory 11, a camera device 13, a network interface 14, and a communication bus 15.

[0055] The memory 11 includes at least one type of readable storage medium. The at least one type of readable storage medium may be a non-volatile storage medium such as a flash memory, a hard disk, a multimedia card, a card-type memory 11, etc. In some embodiments, the readable storage medium may be an internal storage unit of the electronic device 1, such as the hard disk of the electronic device 1. In other embodiments, the readable storage medium may also be an external memory 11 of the electronic device 1, such as a plug-in hard disk equipped on the electronic device 1, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0056] In this embodiment, the readable storage medium of the memory 11 is generally used to store the gait recognition program 10 based on an intelligent wearable sensor installed in the electronic device 1, etc. The memory 11 may also be used to temporarily store data that has been output or will be output.

[0057] The processor 12 may be a central processing unit (CPU), a microprocessor, or other data processing chips in some embodiments, and is used to run the program code stored in the memory 11 or process data, such as executing the gait recognition program 10 based on an intelligent wearable sensor, etc.

[0058] The network interface 14 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), and is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0059] The communication bus 15 is used to implement the connection and communication between these components.

[0060] Figure 1 Only the electronic device 1 with components 11 - 15 is shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented.

[0061] Optionally, the electronic device 1 may further include a user interface, and the user interface may include an input unit such as a keyboard, a voice input device such as a microphone and other devices with voice recognition functions, a voice output device such as a speaker, headphones, etc. Optionally, the user interface may further include a standard wired interface and a wireless interface.

[0062] Optionally, the electronic device 1 may further include a display, and the display may also be referred to as a display screen or a display unit. In some embodiments, it may be an LED display, a liquid crystal display, a touch liquid crystal display, an organic light-emitting diode (OLED) toucher, etc. The display is used to display the information processed in the electronic device 1 and to display a visual user interface.

[0063] Optionally, the electronic device 1 further includes a touch sensor. The area provided by the touch sensor for the user to perform a touch operation is called a touch area. In addition, the touch sensor described here may be a resistive touch sensor, a capacitive touch sensor, etc. Moreover, the touch sensor includes not only a contact type touch sensor, but also a proximity type touch sensor, etc. In addition, the touch sensor may be a single sensor, or may be a plurality of sensors arranged in an array, for example.

[0064] In addition, the area of the display of the electronic device 1 may be the same as or different from the area of the touch sensor. Optionally, the display and the touch sensor are stacked to form a touch display screen. The device detects the touch operation triggered by the user based on the touch display screen.

[0065] Optionally, the electronic device 1 may further include a radio frequency (RF) circuit, sensors, an audio circuit, etc., which will not be elaborated here.

[0066] In Figure 1 In the shown device embodiment, in the memory 11 which is a kind of computer storage medium, an operating system and a gait recognition program 10 based on intelligent wearable sensors may be included; when the processor 12 executes the gait recognition program 10 stored in the memory 11, the following steps are implemented:

[0067] Obtain the sensing data of the training target in seven dimensions through intelligent wearable sensors respectively, and determine the original data according to the sensing data;

[0068] Perform data preprocessing on the original data to obtain training data; meanwhile,

[0069] Construct a dual-channel hybrid neural network model, and train the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed;

[0070] Perform gait recognition on the target to be detected based on the gait recognition model, and output the corresponding recognition result.

[0071] Corresponding to the above-mentioned gait recognition electronic device based on intelligent wearable sensors, the present invention also provides a gait recognition system based on intelligent wearable sensors.

[0072] As Figure 2 As shown in the logic block diagram of the preferred embodiment of the gait recognition system based on intelligent wearable sensors of the present invention, the gait recognition system 100 based on intelligent wearable sensors in the embodiment of the present invention includes:

[0073] An original data acquisition unit 101, configured to obtain the sensing data of the training target in seven dimensions through intelligent wearable sensors respectively, and determine the original data according to the sensing data;

[0074] A training data acquisition unit 102, configured to perform data preprocessing on the original data to obtain training data; meanwhile,

[0075] A recognition model formation unit 103, configured to construct a dual-channel hybrid neural network model, and train the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed;

[0076] A recognition unit 104, configured to perform gait recognition on the target to be detected based on the gait recognition model, and output the corresponding recognition result.

[0077] In addition, the present invention also provides a gait recognition method based on intelligent wearable sensors. Referring to Figure 3 As shown in the flowchart of the preferred embodiment of the gait recognition method based on intelligent wearable sensors of the present invention. This method can be executed by a device, and the device can be implemented by software and / or hardware.

[0078] In this embodiment, the gait recognition method based on intelligent wearable sensors includes:

[0079] S110: Obtain the sensing data of the training target in seven dimensions through intelligent wearable sensors respectively, and determine the original data according to the sensing data.

[0080] Specifically, the smart wearable sensor can adopt a textile knee pad structure, a belt structure or a wristband structure. The smart wearable sensor includes a fixed strap and an inertial sensor, a gyroscope, a tensile strain sensor and a Bluetooth transmission module arranged on the fixed strap. The smart wearable sensor can be set at the knee joint and / or ankle joint and / or hip joint of the training target through the fixed strap to collect sensor data related to the target gait.

[0081] Among them, the sensor data in the above seven dimensions further include three-axis acceleration data collected by inertial sensors, three-axis angular velocity data collected by gyroscopes, and tensile strain data collected by tensile strain sensors, etc. The collected sensor data can be transmitted through a Bluetooth transmission module, and then processed and analyzed by a host computer. In addition, the smart wearable sensor of the embodiment of the present invention may include a six-axis inertial measurement unit (IMU) and a tensile strain sensor, a Bluetooth to serial wireless communication module, a Bluetooth receiver, and a host computer processing program. The client can receive sensor data in real time through Bluetooth transmission, and output corresponding recognition results through model processing.

[0082] In addition, in order to ensure the integrity, diversity and effectiveness of the original data, the training targets (multiple users wearing sensors) need to collect data in the most natural posture. The experimenters wear wearable sensors and connect to and output data with the host computer through the sensor's wireless communication module. At the same time, the PC host computer collects sensor data in seven dimensions through the serial port for data processing.

[0083] In a specific embodiment of the present invention, the sensor data of the training target in seven dimensions are respectively obtained as raw data through the intelligent wearable sensor, including:

[0084] 1. Use the data acquisition module to formulate a matching upper computer program for the lower computer serial port data, so as to sample the data points of different training targets at regular intervals and obtain sensor data;

[0085] 2. Use the data annotation module to uniformly annotate the sensor data and determine the annotated data;

[0086] 3. Use the data set synthesis module to aggregate the labeled data of all training targets into npy files, and replace the identities with categories to obtain the original data.

[0087] S120: Perform data preprocessing on the original data to obtain training data.

[0088] Among them, data preprocessing is performed on the original data to obtain training data, which at least includes: performing filtering and denoising processing on the original data; performing segmentation processing on the original data after filtering and denoising by adopting a sliding window method; performing normalization processing on the segmented original data to form training data.

[0089] Specifically, in the process of eliminating outliers from the original data, mainly the outliers in the data are eliminated. The outliers mainly refer to the data collected by the experimenter that exceeds the transmission range of the sensor, with NaN values or null values appearing, and this data needs to be removed.

[0090] In addition, for denoising the data, mainly the average filtering algorithm is used to reduce the noise of the data. The average filtering algorithm refers to using the average value of the data within a certain time range to smooth the signal to reduce the influence of noise. This filtering algorithm is applicable to processing signals with periodic or random noise. The specific filtering formula is expressed as:

[0091]

[0092] Among them, y(i) represents the filtered output value, M is the size of the filtering window, x(j) is the data point within the window, and thus part of the noise is effectively removed through the average filtering algorithm.

[0093] In addition, data segmentation mainly means that when the processed data is long-sequence data and cannot be input into the network model, it is necessary to adopt a sliding window method to segment the data. The length of one window can be set to 50 frames, and the step size is 25 frames.

[0094] Furthermore, the data can be normalized to bring one dimension of the data within a certain range and conform to the normal distribution. The expression formula is:

[0095]

[0096] Among them, x is the original data value, μ is the mean of this dimension, σ is the standard deviation of this dimension, and x′ is the data after normalization.

[0097] S130: Construct a dual-channel hybrid neural network model, and train the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed.

[0098] Specifically, Figure 4 shows the framework structure of the dual-channel hybrid neural network model according to an embodiment of the present invention, as Figure 4As shown in the figure, the channel mixing neural network model includes a first channel, a second channel, and a multi-layer perceptron mechanism. Among them, the first channel includes a first convolutional neural network and a first recurrent neural network connected to each other; the second channel includes a Fourier transform mechanism, a second convolutional neural network, and a second recurrent neural network connected to each other. The outputs of the first recurrent neural network and the second recurrent neural network are combined and input into the multi-layer perceptron mechanism, and the recognition result is output through the multi-layer perceptron mechanism.

[0099] It can be seen that the raw data is the time-series data collected by the sensor. After FFT (Fourier) transformation, the time-domain data can be converted into frequency-domain data. This dual-channel design can obtain more feature data. After the dual-channel data passes through one-dimensional convolution and recurrent neural networks, feature fusion is performed, and two features of the time domain and frequency domain of the data can be captured. Then, it is input into the multi-layer perceptron for classification to achieve the effect of personnel recognition.

[0100] As a specific example, the following will take one-dimensional convolution to extract data features as an example for detailed description.

[0101] Among them, the input layer is usually set as the 0th layer: M represents the size of the time window after data preprocessing, and the output of the convolutional layer is expressed as:

[0102]

[0103] In the formula, f(·) is the activation function; b is the bias term; q is the one-dimensional convolution kernel vector; is the length of q.

[0104] Then, the data is sent to the pooling layer MP to reduce the dimensionality of the feature space and suppress overfitting. The output of MP is expressed as:

[0105] P j =max(c (j-1)R ,…,c jR ) j=1,…,M / R

[0106] In the formula, R represents the size of the pooling window. The output of the pooling layer is the feature map P output by the convolutional layer. The data after convolution is input into the recurrent network to extract the time-series features in the data.

[0107] It can be seen that RNN is a recurrent neural network used to process sequence data. Different from the traditional feedforward neural network, RNN has recurrent connections, which enables it to maintain a memory state when processing sequences. When the data is input into the recurrent neural network, it will go through the following calculation process, and the formula is as follows:

[0108] Assume that at time step t, the input is x t , and the hidden state is h t, the output is y t , then there is

[0109] h t = σ(W hx x t + W hh h t-1 + b h )

[0110] y t = W yh h t + b y

[0111] where, W hx is the weight matrix from the input to the hidden state, W hh is the weight matrix from the hidden state to the hidden state, W yh is the weight matrix from the hidden state to the output, b h and b y are the bias terms of the hidden state and the output respectively, and σ is the activation function (such as tanh or ReLU).

[0112] Finally, the data passing through the recurrent neural network will go through a multi-layer perceptron layer again. The basic working principle of the multi-layer perceptron is to pass the input data through a series of linear and non-linear transformations to the output layer to achieve a complex non-linear mapping of the input data. During the training process, the multi-layer perceptron adjusts the weights and biases in the network through the backpropagation algorithm to minimize the error between the predicted output and the actual output. The specific formula is as follows:

[0113] For the hidden layer, assuming there are (N) neurons, the input is (x), the weight is (W), the bias is (b), and the activation function is (f), then the output (h) of the hidden layer can be expressed as:

[0114] h = f(Wx + b)

[0115] And for the output layer, assuming there are (M) neurons, the output of the hidden layer is (h), the weight of the output layer is (V), the bias is (c), and the activation function of the output layer is (g), then the output (y) of the output layer can be expressed as:

[0116] y = g(Vh + c)

[0117] After obtaining the data with extracted features through the multi-layer perceptron module, it is classified through softmax. The formula of the softmax function is as follows:

[0118]

[0119] In the formula, x is the input of softmax, that is, the output of the fully connected layer; K is the number of recognizable identities. The result of softmax is the probability of each category, and their sum is 1.

[0120] S140: Perform gait recognition on the target to be detected based on the gait recognition model, and output the corresponding recognition result.

[0121] Specifically, in the real-time activity recognition stage, for the newly input sensor data, start counting from the 50th frame. After data preprocessing, input it into the entire neural network model (i.e., the dual-channel hybrid neural network model), and then the gait of the experimental personnel can be recognized to obtain the personnel identity and display it on the PC interface. Write a QT script to implement the human-computer interaction interface. During the application process, the personnel perform corresponding activities, which are handed over to the algorithm model for recognition, and the recognition result is displayed on the interface. The specific display result can be an activity or a disease diagnosis result, etc.

[0122] In another specific embodiment of the present invention, the training data can be divided into a training data set and a test data set according to a ratio of 4:1; the training data set is used to train the dual-channel hybrid neural network model, and the test data set is used to test the trained dual-channel hybrid neural network model, and determine the gait recognition model according to the test results.

[0123] Furthermore, training the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed includes: training the dual-channel hybrid neural network model based on the training data until the loss function of the dual-channel hybrid neural network model is less than a preset threshold to form a gait recognition model; wherein,

[0124] The loss function includes mean absolute error, mean square error, and cross-entropy function;

[0125] The expression formula of the mean absolute error is:

[0126]

[0127] The expression formula of the mean square error is:

[0128]

[0129] Wherein, n represents the number of training data, y i represents the true value of the i-th training data, y i p represents the predicted value of the i-th training data;

[0130] The expression formula of the cross-entropy function is:

[0131]

[0132] Among them, y represents the true value of the input training data, represents the predicted value corresponding to y.

[0133] In addition, an embodiment of the present invention further provides a computer-readable storage medium, which includes a gait recognition program based on an intelligent wearable sensor. When the gait recognition program based on the intelligent wearable sensor is executed by a processor, the operations shown in the above method are implemented.

[0134] The specific implementation manner of the computer-readable storage medium of the present invention is substantially the same as the specific implementation manners of the above-mentioned gait recognition method based on an intelligent wearable sensor and the electronic device, and will not be described in detail here.

[0135] It should be noted that in this article, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of additional identical elements in the process, device, article or method including that element.

[0136] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments. Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiment methods can be implemented by means of software plus a necessary general hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation manner. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions for causing a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in various embodiments of the present invention.

[0137] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the description of the present invention and the contents of the drawings, or directly or indirectly applied to other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. A gait recognition method based on intelligent wearable sensors, applied to a client, characterized in that, The method includes: Obtaining sensing data of the training target in seven dimensions through intelligent wearable sensors respectively, and determining the original data according to the sensing data; Performing data preprocessing on the original data to obtain training data; meanwhile, Constructing a dual-channel hybrid neural network model, and training the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed; Performing gait recognition on the target to be detected based on the gait recognition model, and outputting the corresponding recognition result.

2. The gait recognition method based on intelligent wearable sensors according to claim 1, wherein The intelligent wearable sensor includes a fixed strap and an inertial sensor, a gyroscope, a tensile strain sensor and a Bluetooth transmission module arranged on the fixed strap; The intelligent wearable sensor is arranged at the knee joint and / or ankle joint and / or hip joint of the training target through the fixed strap.

3. The gait recognition method based on intelligent wearable sensors according to claim 2, wherein The sensing data in the seven dimensions includes three-axis acceleration data collected by the inertial sensor, three-axis angular velocity data collected by the gyroscope, and tensile strain data collected by the tensile strain sensor.

4. The gait recognition method based on an intelligent wearable sensor according to claim 1, wherein Obtaining sensing data of the training target in seven dimensions as the original data through intelligent wearable sensors respectively, including: Formulating a supporting host computer program for the lower computer serial port data through a data acquisition module to perform data point sampling on the movements of different training targets at regular intervals and obtain the sensing data; Performing unified identity annotation on the sensing data through a data annotation module to determine the annotated data; Aggregating the annotated data of all training targets into an npy file through a data set synthesis module, and replacing the identity with the category to obtain the original data.

5. The gait recognition method based on an intelligent wearable sensor according to claim 1, characterized in that Performing data preprocessing on the original data to obtain training data, including: Performing filtering and denoising processing on the original data; Performing segmentation processing on the original data after filtering and denoising processing by adopting a sliding window method; Performing normalization processing on the segmented original data to form the training data.

6. The gait recognition method based on an intelligent wearable sensor according to claim 1, characterized in that The dual-channel hybrid neural network model includes a first channel, a second channel and a multi-layer perception mechanism; wherein, The first channel includes a first convolutional neural network and a first recurrent neural network connected to each other; The second channel includes a Fourier transform mechanism, a second convolutional neural network and a second recurrent neural network connected to each other; The outputs of the first recurrent neural network and the second recurrent neural network are combined and input into the multi-layer perception mechanism, and the recognition result is output through the multi-layer perception mechanism.

7. The gait recognition method based on intelligent wearable sensors according to claim 1, wherein The training data is divided into a training data set and a test data set according to a ratio of 4:1; The training data set is used to train the dual-channel hybrid neural network model, and the test data set is used to test the trained dual-channel hybrid neural network model, and the gait recognition model is determined according to the test results.

8. The gait recognition method based on an intelligent wearable sensor according to claim 1, wherein Training the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed, including: Training the dual-channel hybrid neural network model based on the training data until the loss function of the dual-channel hybrid neural network model is less than a preset threshold to form the gait recognition model; wherein, The loss function includes mean absolute error, mean square error and cross-entropy function; The expression formula of the mean absolute error is: The expression formula of the mean square error is: where n represents the number of the training data, y i represents the true value of the i-th training data, y i p represents the predicted value of the i-th training data; The expression formula of the cross-entropy function is: Among them, y represents the true value of the input training data, represents the predicted value corresponding to the said y.

9. An electronic device, characterized in that, The electronic device includes: a memory and a processor. The memory includes a gait recognition program based on an intelligent wearable sensor. When the gait recognition program based on the intelligent wearable sensor is executed by the processor, the following steps are implemented: Obtaining sensing data of a training target in seven dimensions through an intelligent wearable sensor, and determining original data according to the sensing data; Performing data preprocessing on the original data to obtain training data; meanwhile, Constructing a dual-channel hybrid neural network model, and training the dual-channel hybrid neural network model based on the training data until a gait recognition model is formed; Performing gait recognition on a target to be detected based on the gait recognition model, and outputting a corresponding recognition result.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a gait recognition program based on an intelligent wearable sensor. When the gait recognition program based on the intelligent wearable sensor is executed by a processor, the steps of the gait recognition method based on an intelligent wearable sensor according to any one of claims 1 to 8 are implemented.

Images