A method, device and electronic device for obtaining human behavior recognition signals

Abstract

The present application relates to a method for obtaining human behavior recognition signals, and the method includes: obtaining an initial phase difference according to the initial phase values in an initial phase data sequence; determining a region where a phase mutation occurs according to the calculated initial phase difference and an initialization step size; in this region, calculating a target phase difference according to the count value of a dynamic counter, and determining a target phase mutation value according to the target phase difference; calculating the target phase mutation value according to the target phase difference to obtain a target phase data sequence; performing normalization processing on the target phase data sequence, the received channel state data, and the received signal strength data; obtaining a converted phase data sequence according to the calculated distance function difference; obtaining a signal strength data difference according to the acquired signal strength data; and fusing the converted phase data sequence, the signal strength data difference, and the channel state data after the normalization processing is completed.

Description

Technical Field

[0001] The present application relates to the fields of Internet of Things technology and human behavior recognition technology, and particularly to a method, device and electronic device for obtaining human behavior recognition signals. Background Art

[0002] With the rapid development of Internet of Things technology, intelligent recognition and monitoring of human behavior have become a research hotspot. Human behavior recognition technology has shown important application values in multiple fields, including but not limited to security monitoring, smart home, medical monitoring, autonomous driving, and work safety. Existing research work on human behavior recognition mainly includes vision-based human behavior recognition technology and wearable sensor-based human behavior recognition technology. The specific methods are as follows:

[0003] 1. Vision-based human behavior recognition technology: This type of technology captures images or videos containing different human behaviors through cameras, extracts behavior features using image processing algorithms, and then inputs them into machine learning models for classification. Although this technology has achieved high recognition accuracy, it has problems such as being sensitive to environmental lighting conditions, high risk of privacy leakage, and strict hardware installation requirements. In addition, the position and angle of the camera will also affect the final recognition effect.

[0004] 2. Wearable sensor-based human behavior recognition technology: This method mainly uses various sensors to collect information such as acceleration and pressure to identify different human behaviors. Although this technology avoids the problem of privacy leakage, wearing sensors for a long time may cause discomfort to users. In addition, most sensors have high costs and require regular charging and maintenance, which increases the usage cost and inconvenience for users.

[0005] 3. RFID-based human behavior recognition technology: In recent years, researchers have begun to explore the possibility of using RFID technology for human behavior recognition. RFID technology has advantages such as low cost, easy deployment, and strong signal penetration, and has been widely used in fields such as supply chain management, inventory control, and three-dimensional positioning. The basic principle of using RFID technology for behavior recognition is to identify human behavior by reading the phase and RSSI data change patterns generated by RFID tags during human movement. However, existing RFID-based behavior recognition methods still face some challenges, such as the impact of tag deployment schemes on recognition accuracy and data instability caused by environmental and device interference.

[0006] 4. Human Behavior Recognition Technology Based on Wi-Fi CSI: Wi-Fi Channel State Information (CSI) contains rich spatial and temporal characteristics, can provide finer-grained information than traditional Received Signal Strength Indication (RSSI), and is suitable for contactless behavior recognition. It can be obtained only through the existing Wi-Fi infrastructure without the need for additional hardware, so it has the advantages of low cost and easy deployment. However, Wi-Fi CSI also faces data stability problems caused by multipath effects, noise, and other wireless communication interferences.

[0007] Therefore, no matter which of the above modes, it will face the problems of low accuracy, unstable data, and limited types of human behaviors that can be recognized. Summary of the Invention

[0008] Embodiments of the present invention provide a method, device, and electronic device for obtaining human behavior recognition signals, which are used to solve the problems of low accuracy and unstable data when recognizing human behaviors in the prior art. The method specifically includes:

[0009] Calculate an initial phase difference according to the initial phase values in the obtained initial phase data sequence; the initial phase data sequence includes multiple initial phase values for identifying the behaviors of different parts of the human body;

[0010] Determine the region where phase mutation occurs from the initial phase data sequence according to the initial phase difference and the initialization step size;

[0011] In the region where phase mutation occurs, calculate the target phase difference between two adjacent initial phase values according to the count value of the dynamic counter, and determine the target phase mutation value according to the target phase difference; calculate the target phase mutation value according to the target phase difference to obtain the target phase data sequence;

[0012] Perform normalization processing on the target phase data sequence, the received channel state data, and the received signal strength data;

[0013] Calculate the distance function difference generated by relative motion between the transmitting module that transmits the human behavior recognition signal and the receiving module that receives the human behavior recognition signal;

[0014] Obtain the converted phase data sequence corresponding to the target phase data sequence after normalization processing according to the distance function difference;

[0015] Obtain the signal strength data difference according to the two adjacent signal strength data after normalization processing obtained before and after;

[0016] Fuse the converted phase data sequence, the signal strength data difference, and the channel state data after normalization to obtain a fused human behavior recognition signal.

[0017] Among them, determining the region where phase mutation occurs from the initial phase data sequence according to the initial phase difference and the initialization step length includes:

[0018] The initialization step length includes a first preset step length and a second preset step length, and the first preset step length is greater than the second preset step length;

[0019] Find two adjacent initial phase values in the initial phase data sequence whose initial phase difference is greater than the first preset phase difference. When the found initial phase values are in a continuous region, use the second preset step length to find the region where phase mutation occurs in this continuous region;

[0020] Find two adjacent initial phase values in the initial phase data sequence whose initial phase difference is less than the second preset phase difference. When the found initial phase values are in a continuous region, use the first preset step length to find the region where phase mutation occurs in this continuous region.

[0021] Among them, calculating the target phase difference between two adjacent initial phase values according to the count value of the dynamic counter includes:

[0022] Starting from the second initial phase value in the initial phase data sequence, add the values calculated based on the current counter respectively to obtain multiple intermediate phase values;

[0023] Take the difference between two adjacent intermediate phase values as the target phase difference.

[0024] Among them, determining the target phase mutation value according to the target phase difference includes:

[0025] Use the second preset step length to traverse the initial phase values in the region where phase mutation occurs in turn;

[0026] Find two adjacent intermediate phase values whose target phase difference is greater than the first preset target phase difference. When the found intermediate phase values are in a continuous region, reduce the second preset step length by a preset value and search this continuous region to determine the target phase mutation value;

[0027] Find two adjacent intermediate phase values whose target phase difference is less than the second preset target phase difference. When the found intermediate phase values are in a continuous region, increase the second preset step length by a preset value and search this continuous region to determine the target phase mutation value.

[0028] Among them, calculating the target phase mutation value according to the target phase difference includes:

[0029] When the target phase difference value is greater than the preset target phase difference value, subtract the preset target value from the current target phase mutation value;

[0030] When the target phase difference value is less than the preset target phase difference value, add the preset target value to the current target phase mutation value.

[0031] Among them, calculating the difference in the distance function generated by the relative motion between the transmitting module that transmits the human behavior recognition signal and the receiving module that receives the human behavior recognition signal includes:

[0032] Calculate the distance function according to the distance between the receiving module and the transmitting module, the phase offset value between the receiving module and the transmitting module, and the wavelength of the receiving module;

[0033] Take the difference between the distance function calculated by the receiving module at the current moment and the distance function calculated at the previous moment as the distance function difference.

[0034] Among them, calculating the distance function according to the distance between the receiving module and the transmitting module, the phase offset value between the receiving module and the transmitting module, and the wavelength of the receiving module includes:

[0035] Calculate the distance function according to the following formula :

[0036] ;

[0037] Among them, is the distance between the receiving module and the transmitting module; is the wavelength of the receiving module; is the phase offset between the receiving module and the transmitting module.

[0038] Among them, obtaining the converted phase data sequence corresponding to the normalized target phase data sequence according to the distance function difference includes:

[0039] Calculate the converted phase data of each data in the normalized target phase data sequence , to obtain the converted phase data sequence;

[0040] ;

[0041] Among them, is the time difference generated when the transmitting module and the receiving module move relative to each other; The distance function difference generated within the time difference.

[0042] An embodiment of the present invention further provides a device for obtaining a human behavior recognition signal, and the device includes:

[0043] A first calculation module, configured to calculate an initial phase difference according to the initial phase values in the obtained initial phase data sequence; the initial phase data sequence includes a plurality of initial phase values for identifying the behaviors of different parts of the human body;

[0044] A first determination module, configured to determine a region where a phase mutation occurs from the initial phase data sequence according to the initial phase difference and the initialization step size;

[0045] A second calculation module, configured to calculate a target phase difference between two adjacent initial phase values according to the count value of a dynamic counter in the region where the phase mutation occurs, and determine a target phase mutation value according to the target phase difference; calculate the target phase mutation value according to the target phase difference to obtain a target phase data sequence;

[0046] A first processing module, configured to perform normalization processing on the target phase data sequence, the channel state data, and the signal strength data;

[0047] A third calculation module, configured to calculate a distance function difference generated due to relative motion between a transmitting module that transmits the human behavior recognition signal and a receiving module that receives the human behavior recognition signal;

[0048] A second processing module, configured to obtain a converted phase data sequence corresponding to the target phase data sequence after normalization processing according to the distance function difference;

[0049] A third processing module, configured to obtain a signal strength data difference according to two adjacent signal strength data after normalization processing;

[0050] A fourth processing module, configured to fuse the converted phase data sequence, the signal strength data difference, and the channel state data after normalization processing to obtain a fused human behavior recognition signal.

[0051] An embodiment of the present invention further provides an electronic device, characterized in that the electronic device includes a processor, a memory, and a program or instruction stored on the memory and executable on the processor, and the program or instruction is executed by the processor to perform the steps of the above method.

[0052] The method provided by the embodiment of the present invention processes the received human body signals and integrates the wifi signals at the same time to form a fusion signal, which can improve the recognition accuracy and ensure data stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a schematic flowchart of a method for obtaining human behavior recognition signals provided by the embodiment of the present invention;

[0054] Figure 2 It is a schematic structural diagram of integrating the ECA mechanism into the residual network model;

[0055] Figure 3 It is a schematic structural diagram of the residual module of the ECA mechanism;

[0056] Figure 4 It is a schematic structural diagram of a deep residual neural network model based on residual blocks;

[0057] Figure 5 It is a schematic structural diagram of a method for obtaining human behavior recognition signals provided by the embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] To make the above objects, features, and advantages of the present application more obvious and understandable, the following describes the specific embodiments of the present application in detail with reference to the accompanying drawings. Many specific details are set forth in the following description to fully understand the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present application. Therefore, the present application is not limited by the specific embodiments disclosed below.

[0059] In the description of the present application, it should be understood that if terms such as "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. appear, the orientation or positional relationship indicated by these terms is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present application.

[0060] In addition, if the terms "first" and "second" appear, these terms are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include at least one such feature. In the description of this application, if the term "plurality" appears, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0061] An embodiment of the present invention provides a method for obtaining a human behavior recognition signal. This method is applied to a system for obtaining a human behavior recognition signal, and is mainly used to solve the problems of low accuracy, unstable data, and limited types when obtaining human behavior information in the prior art.

[0062] This system includes a transmitting module that transmits a human behavior recognition signal and a receiving module that receives the human behavior recognition signal, and cooperates with Wi-Fi for signal transmission. In this application, the transmitting module is deployed at key parts of the human body to maximize the signal capture effect and cooperate with Wi-Fi data for accurate recognition. The specific deployment is as follows: One transmitting module is placed at each of the two arms, one transmitting module is placed at the waist, one transmitting module is placed on the back, and one transmitting module is placed at each of the two thighs. The transmitting module can be an RFID tag. In terms of WIFI, a TP-Link wireless router and a laptop equipped with a Wi-Fi network card are used as experimental devices. The position spacing between each transmitting module is greater than 10 cm. As Figure 1 shown, the specific process is as follows:

[0063] Step 11, calculate an initial phase difference based on the initial phase values in the obtained initial phase data sequence; the initial phase data sequence includes a plurality of initial phase values used to identify the behaviors of different parts of the human body.

[0064] The initial phase data sequence is composed of human behavior data obtained by transmitting modules placed at different parts of the human body.

[0065] First, quickly scan the entire initial phase data sequence, calculate the difference between adjacent two initial phase values, and use this difference as the initial phase difference.

[0066] Step 12, determine the region where a phase mutation occurs from the initial phase data sequence according to the initial phase difference and the initialization step size. The specific process of this step is as follows:

[0067] The initialization step size includes a first preset step size and a second preset step size, and the first preset step size is greater than the second preset step size;

[0068] Find two adjacent initial phase values in the initial phase data sequence whose initial phase difference is greater than a first preset phase difference. When the found initial phase values form a continuous region, use a second preset step size to find the region where phase mutation occurs within this continuous region;

[0069] Find two adjacent initial phase values in the initial phase data sequence whose initial phase difference is less than a second preset phase difference. When the found initial phase values form a continuous region, use a first preset step size to find the region where phase mutation occurs within this continuous region.

[0070] The above first preset phase difference is greater than the second preset phase difference.

[0071] Illustrate the above content with an example:

[0072] Mark all initial phase values in the initial phase data sequence whose initial phase differences are greater than the first preset phase difference and less than the second preset phase difference. Since the initial phase difference is calculated from two adjacent initial phase values, the initial phase values marked in this step can be either a continuous region or a discontinuous region.

[0073] Since the signals used to describe normal human behaviors are generally relatively smooth signals, if a large phase difference occurs, it indicates that a phase mutation has occurred in the signal. Therefore, in this step, it is necessary to find the region where the phase difference changes significantly, and this region indicates that a phase mutation has occurred. However, in the actual operation process, due to relying solely on this method, errors may occur. Therefore, in order to more accurately find the phase mutation, it is necessary to dynamically adjust the step size. The specific method is as follows:

[0074] Set the initial step size Δ0 = 1, set the maximum step size, i.e., the first preset step size, as Δmax = 5, and set the minimum step size, i.e., the second preset step size, as Δmin = 1;

[0075] When the initial phase difference is less than the second preset phase difference, use the first preset step size Δmax = 5 to find the phase mutation within this continuous region. When a phase mutation is detected, use the second preset step size Δmin = 1 to continue the search to ensure the accuracy of the search. After that, if no phase mutation occurs for several consecutive initial phase values, the step size can be gradually increased but not exceeding Δmax = 5 at most.

[0076] When the initial phase difference is greater than the first preset phase difference, use the second preset step size Δmin = 1 to find the phase mutation within this continuous region. If no phase mutation occurs for several consecutive initial phase values, the step size can be gradually increased but not exceeding Δmax = 5 at most.

[0077] In this step, the initial step size, the first preset step size, and the second preset step size can be set according to the actual detection object.

[0078] Step 13, in the region where the phase mutation occurs, calculate the target phase difference between two adjacent initial phase values according to the count value of the dynamic counter, calculate the target phase difference between two adjacent initial phase values, and determine the target phase mutation value according to the target phase difference; calculate the target phase mutation value according to the target phase difference to obtain a target phase data sequence.

[0079] In this step, the dynamic counter is used to record the situations of phase sudden decrease and phase sudden increase. It is used to record the number of times of phase sudden decrease. It is to record the number of times of phase sudden increase. In the initial situation .

[0080] In this step, the method for calculating the target phase difference between two adjacent initial phase values according to the count value of the dynamic counter is as follows:

[0081] Starting from the second initial phase value in the initial phase data sequence, add the values calculated based on the current counter respectively to obtain multiple intermediate phase values; take the difference between two adjacent intermediate phase values as the target phase difference. The calculation method of the value calculated based on the current counter is as follows:

[0082] First, calculate the counter difference , and starting from the second initial phase value, add to each initial phase value as the intermediate phase value.

[0083] In this step, the specific method for determining the target phase mutation value according to the target phase difference is as follows:

[0084] Traverse the initial phase values in the region where the phase mutation occurs in turn using the second preset step size;

[0085] Find two adjacent intermediate phase values whose target phase difference is greater than the first preset target phase difference. When the found intermediate phase values are in a continuous region, reduce the second preset step size by a preset value and search this continuous region to determine the target phase mutation value;

[0086] Find two adjacent intermediate phase values whose target phase difference is less than the second preset target phase difference. When the found intermediate phase values are in a continuous region, increase the second preset step size by a preset value and search this continuous region to determine the target phase mutation value.

[0087] Calculating the target phase mutation value according to the target phase difference includes:

[0088] When the target phase difference is greater than the preset target phase difference, subtract the preset target value from the current target phase value;

[0089] When the target phase difference is less than the preset target phase difference, add the preset target value to the current target phase value.

[0090] For example:

[0091] If the target phase difference is greater than the first preset target phase difference , it indicates that the phase has a sudden increase, then subtract the preset target value from the current target phase mutation value , add 1;

[0092] If the target phase difference is less than , it indicates that the phase has a sudden decrease, add the preset target value to the current target phase mutation value , add 1, and update the counter machine counter difference .

[0093] So far, the target phase data sequence is obtained.

[0094] In the method provided in this application, step 13 can be directly executed, that is, taking step 13 as the initial step. When step 13 is the initial step, there is no need to first determine the area where the phase mutation occurs, and the process of "calculating the target phase difference between two adjacent initial phase values according to the count value of the dynamic counter, and determining the target phase mutation value according to the target phase difference" can be directly carried out. Considering the premise of saving calculation time and improving calculation efficiency, step 11 and step 12 can be executed first and then step 13.

[0095] Step 14, perform normalization processing on the target phase data sequence, the received channel state data, and the received signal strength data. Specifically, this step includes:

[0096] Among them, the signal strength data is used to represent the received signal strength of the receiving module; the channel state data is used to represent the Wi-Fi channel state information, which contains rich spatial and temporal characteristics and can provide finer-grained information than the signal strength data of the receiving module, and is suitable for contactless behavior recognition.

[0097] In order to ensure that the three parts of the data sequence: the phase data sequence, the signal strength data sequence of the receiving module, and the data sequence of the channel data module that generates the channel state data, have the same length within the same sampling period, it is necessary to perform normalization processing on these three parts of the data sequence respectively. For the current data in each data sequence, calculate according to the following formula (1):

[0098] Formula (1);

[0099] where is the current sampling time for normalization processing. and are adjacent sampling time points to the current sampling time. is the calculation result of the current data in the data sequence, that is, the phase value data or signal strength data value or channel state data at time .

[0100] To reduce the interference of noise on the experimental results, a Kalman filter can also be used to smooth the data that has completed the above processing process.

[0101] Step 15, calculate the difference in the distance function generated due to relative motion between the transmitting module that transmits the human behavior recognition signal and the receiving module that receives the human behavior recognition signal; the specific process of this step is as follows:

[0102] According to the distance d between the receiving module and the transmitting module, the phase offset value between the receiving module and the transmitting module and the wavelength of the receiving module calculate the distance function ; calculate the formula (2) of the distance function as follows:

[0103] Formula (2);

[0104] According to the distance function calculated by the receiving module at the current moment, and the difference from the distance function calculated at the previous moment, as the distance function difference , where i in formula (2) is the data arranged in order.

[0105] Step 16, obtain the converted phase data sequence corresponding to the target phase data sequence after normalization processing according to the distance function difference, and obtain the converted phase data sequence through formula (3) :

[0106] Formula (3);

[0107] where is the sampling time difference.

[0108] Since the read / write rate of the transmitting module is very fast, at time and The moving speed v of the transmitting module within a moment can be regarded as constant. Then, within the moment to the moment, the change in the relative distance between the transmitting module and the receiving module can be obtained according to formula (4):

[0109] Formula (4);

[0110] wherein is the included angle between the transmitting module and the receiving module; from formulas (2) and (4), it can be obtained that and = The relationship between them is formula (5):

[0111] Formula (5);

[0112] From formulas (4) and (5), formula (6) can be obtained:

[0113] (Formula 6);

[0114] When the included angle between the receiving module and the transmitting module is , and the receiving module moves at a speed v, the converted phase data sequence is calculated by the following formula (7):

[0115] (Formula 7);

[0116] According to formulas (6) and (7), formula (3) can be obtained.

[0117] Step 17: According to the two adjacent signal strength data obtained after normalization before and after, the signal strength data difference is obtained, where i in formula (8) represents the time point, and formula (8) is as follows:

[0118] (Formula 8);

[0119] After the above process is completed, the data can also be smoothed to effectively suppress background noise, thereby reducing the training time and improving the accuracy of behavior classification.

[0120] Step 18: Fuse the converted phase data sequence, the signal strength data difference, and the channel state data obtained after normalization to obtain a fused human behavior recognition signal.

[0121] In this step, to achieve the effective fusion of data between the transmitting module and the receiving module, the transformed phase data sequence, the difference in signal strength data, and the channel state data after normalization are concatenated in the channel dimension, and these time series data are fused. The data obtained from this fusion process will be used for subsequent feature extraction and classification analysis, providing more accurate support for human behavior recognition. The fusion method adopted in the present invention belongs to a multi-channel data input fusion strategy, by combining the data source characteristics of different sensors.

[0122] In this application, feature extraction and behavior recognition can be further performed, specifically including:

[0123] In this application, by integrating the ECA mechanism into the residual network model, different weights are assigned to each channel, thus significantly improving the performance of the CNN model. Different from many existing methods that often increase network complexity when optimizing performance, the ECA module can enhance cross-channel information interaction and improve model performance while only introducing a small number of additional parameters.

[0124] As Figure 2 shown in the flowchart of the ECA mechanism, the working process of the ECA module first performs global average pooling (Global Average Pooling, GAP) on all channels while keeping the number of channels unchanged. Subsequently, the principle of weight sharing and one-dimensional convolution operation of size k are used to generate channel attention. This process not only emphasizes the importance of each channel but also particularly focuses on the interaction between adjacent channels, ensuring the effective transmission and utilization of information between channels.

[0125] In the ECA-ResNet architecture, the residual module structure of the ECA mechanism is as Figure 3 shown, and the performance of the residual network is improved by introducing the ECA mechanism. Each basic unit is composed of a series of components connected in cascade, including a convolutional layer (Conv), batch normalization (Batch Normalization, BN), an activation function, and an attention module. To address the problems of gradient disappearance and network degradation, skip connections (Shortcut Connections, SC) are adopted inside these units.

[0126] Specifically, the ECA attention module is embedded in two different types of residual bottleneck structures Block, forming the basic units of ECA-ResNet, which are respectively labeled as Block1 and Block2:

[0127] Block1: When the number of input and output channels is inconsistent, in order to match the dimensions of the feature map, an additional convolutional layer is added to the skip connection. This ensures that even when the network depth increases, the information flow can be smoothly transmitted, thus maintaining the effectiveness of the network.

[0128] Block2: For the case where the number of input and output channels is the same, the skip connection can directly add the input to the output, simplifying the structure and reducing the computational burden.

[0129] This design not only enhances the model's attention to important channels but also improves the gradient propagation problem during training through an optimized information flow path, thereby enhancing the model's learning ability and final performance. In this way, ECA-ResNet achieves more efficient feature extraction and higher recognition accuracy while maintaining a low complexity.

[0130] In the behavior recognition part, the fused data needs to be input into the residual network structure with the fused ECA mechanism for identifying and distinguishing different behavior changes. The three input channels are the transformed phase data sequence after phase conversion, the difference in signal strength data, and the channel state data after normalization processing. Finally, a unified behavior recognition result is output through the fully connected layer. The structure of the deep residual neural network model based on residual blocks constructed in this application is as Figure 4 shown.

[0131] The structure of the deep residual neural network model based on residual blocks. First, the dimension of the feature map is adjusted through Block1, and then 3 Block2 modules are stacked to form a large Block module for extracting deep features. To improve the feature extraction ability, 3×3 convolutional kernels and 2×2 average pooling are used to replace the traditional 7×7 convolutional kernels, avoiding the limitation of large convolutional kernels on key feature extraction. Then, 3 large Block modules are used for convolution and attention training. Through an effective attention mechanism, the selectivity of the network to important features is enhanced. After passing through the convolution and attention modules, average pooling and max pooling are used to extract global features, and behavior classification (standing, sitting, squatting, sleeping, drinking water) is performed through the fully connected layer and the Softmax function. The dataset is randomly divided into a training set, a validation set, and a test set in a ratio of 6:2:2 to complete the training and evaluation of the model. The improved ECA-ResNet model effectively improves the recognition accuracy of human behaviors by optimizing the convolution design and introducing an attention mechanism.

[0132] As Figure 5 shown, an embodiment of the present invention provides a device for obtaining human behavior recognition signals, and the device includes:

[0133] A first calculation module 51, configured to calculate an initial phase difference according to the initial phase values in the obtained initial phase data sequence; the initial phase data sequence includes multiple initial phase values for identifying behaviors of different parts of the human body;

[0134] The first determination module 52 is configured to determine, according to the initial phase difference and the initialization step size, a region where a phase mutation occurs from the initial phase data sequence;

[0135] The second calculation module 53 is configured to, in the region where the phase mutation occurs, calculate a target phase difference between two adjacent initial phase values according to the count value of the dynamic counter, and determine a target phase mutation value according to the target phase difference; calculate the target phase mutation value according to the target phase difference to obtain a target phase data sequence;

[0136] The first processing module 54 is configured to perform normalization processing on the target phase data sequence, the received channel state data, and the received signal strength data;

[0137] The third calculation module 55 is configured to calculate a distance function difference generated due to relative movement between a transmitting module that transmits the human behavior recognition signal and a receiving module that receives the human behavior recognition signal;

[0138] The second processing module 56 is configured to obtain a converted phase data sequence corresponding to the target phase data sequence after normalization processing according to the distance function difference;

[0139] The third processing module 57 is configured to obtain a signal strength data difference according to two adjacent received signal strength data after normalization processing;

[0140] The fourth processing module 58 is configured to fuse the converted phase data sequence, the signal strength data difference, and the channel state data after normalization processing to obtain a fused human behavior recognition signal.

[0141] The first determination module 52 is further configured to determine, according to the initial phase difference and the initialization step size, a region where a phase mutation occurs from the initial phase data sequence, including:

[0142] The initialization step size includes a first preset step size and a second preset step size, and the first preset step size is greater than the second preset step size;

[0143] Search for two adjacent initial phase values in the initial phase data sequence whose initial phase difference is greater than a first preset phase difference. When the found initial phase values are in a continuous region, use the second preset step size to search for a region where a phase mutation occurs in this continuous region;

[0144] Search for two adjacent initial phase values in the initial phase data sequence whose initial phase difference is less than a second preset phase difference. When the found initial phase values are in a continuous region, use the first preset step size to search for a region where a phase mutation occurs in this continuous region.

[0145] The second calculation module 53 is further configured to start from the second initial phase value in the initial phase data sequence, and respectively add the values calculated based on the current counter to obtain a plurality of intermediate phase values;

[0146] Take the difference between two adjacent intermediate phase values as the target phase difference.

[0147] Among them, determining the target phase mutation value according to the target phase difference includes:

[0148] Use a second preset step size to sequentially traverse the initial phase values in the area where the phase mutation occurs;

[0149] Find two adjacent intermediate phase values whose target phase difference is greater than the first preset target phase difference. When the found intermediate phase values are in a continuous area, reduce the second preset step size by a preset value and then search this continuous area to determine the target phase mutation value;

[0150] Find two adjacent intermediate phase values whose target phase difference is less than the second preset target phase difference. When the found intermediate phase values are in a continuous area, increase the second preset step size by a preset value and then search this continuous area to determine the target phase mutation value.

[0151] Among them, calculating the target phase mutation value according to the target phase difference includes:

[0152] When the target phase difference is greater than the preset target phase difference, subtract the preset target value from the current target phase mutation value;

[0153] When the target phase difference is less than the preset target phase difference, add the preset target value to the current target phase mutation value.

[0154] The third calculation module 55 is further configured to calculate the distance function according to the distance between the receiving module and the transmitting module, the phase offset value between the receiving module and the transmitting module, and the wavelength of the receiving module;

[0155] Take the difference between the distance function calculated by the receiving module at the current moment and the distance function calculated at the previous moment as the distance function difference.

[0156] Among them, calculating the distance function according to the distance between the receiving module and the transmitting module, the phase offset value between the receiving module and the transmitting module, and the wavelength of the receiving module includes:

[0157] Calculate the distance function according to the following formula :

[0158]

[0159] wherein, is the distance between the receiving module and the transmitting module; is the wavelength of the receiving module; is the phase offset between the receiving module and the transmitting module.

[0160] wherein, obtaining the converted phase data sequence corresponding to the target phase data sequence after normalization processing according to the distance function difference includes:

[0161] calculating the converted phase data of each data in the target phase data sequence after normalization processing , to obtain the converted phase data sequence;

[0162]

[0163] wherein, is the time difference generated when the transmitting module and the receiving module move relative to each other; is the distance function difference generated within the time difference.

[0164] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program or instruction stored on the memory and executable on the processor, and the program or instruction is executed by the processor to perform the steps of the above method for obtaining a human behavior recognition signal.

[0165] In this application, unless otherwise clearly specified and limited, if terms such as "installation", "connection", "connection", "fixation", etc. appear, these terms should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components or the interaction relationship between two components, unless otherwise clearly limited. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific situations.

[0166] In this application, unless otherwise clearly stipulated and defined, when a first feature is described as being "on" or "under" a second feature or similar descriptions, it may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, the first feature being "above", "over" and "on top of" the second feature may mean that the first feature is directly above or diagonally above the second feature, or simply indicates that the first feature has a higher horizontal height than the second feature. The first feature being "under", "beneath" and "underneath" the second feature may be that the first feature is directly below or diagonally below the second feature, or simply indicates that the first feature has a lower horizontal height than the second feature.

[0167] It should be noted that if an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or there may also be an intermediate element. If an element is considered to be "connected to" another element, it can be directly connected to the other element or there may be an intermediate element at the same time. If so, the terms "vertical", "horizontal", "upper", "lower", "left", "right" and similar expressions used in this application are only for the purpose of illustration and do not represent the only implementation. The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0168] The above-described embodiments merely represent several implementation manners of this application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent application. It should be pointed out that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application patent shall be subject to the appended claims.

Claims

1. A method for obtaining a human behavior recognition signal, characterized in that: The method comprises: An initial phase difference value is calculated based on an initial phase value in an acquired initial phase data sequence; the initial phase data sequence includes a plurality of initial phase values ​​for identifying behaviors of different parts of a human body; According to the initial phase difference value and the initialization step size, determining the region where the phase mutation occurs from the initial phase data sequence; In the region where the phase mutation occurs, a target phase difference between two adjacent initial phase values ​​is calculated according to the count value of the dynamic counter, and a target phase mutation value is determined according to the target phase difference; the target phase mutation value is calculated according to the target phase difference to obtain a target phase data sequence; Normalizing the target phase data sequence, the received channel state data, and the received signal strength data; According to the distance between the receiving module receiving the human behavior recognition signal and the transmitting module transmitting the human behavior recognition signal, the phase offset value between the receiving module and the transmitting module, and the wavelength of the receiving module, the distance function θ is calculated according to the following formula; Wherein, d is the distance between the receiving module and the transmitting module; λ is the wavelength of the receiving module; Ω is the phase offset between the receiving module and the transmitting module; The difference between the distance function calculated by the receiving module at the current moment and the distance function calculated at the previous moment is used as the distance function difference generated by the relative motion between the transmitting module and the receiving module; The conversion phase data f of each data in the target phase data sequence after normalization is calculated according to the following formula: d , and obtain the conversion phase data sequence; Wherein, Δt is the time difference generated when the transmitting module and the receiving module move relative to each other; Δθ is the distance function difference generated within the Δt time difference; Obtaining a signal strength data difference according to the acquired two adjacent normalized signal strength data; The conversion phase data sequence, the signal strength data difference and the channel state data after normalization processing are fused to obtain a fused human behavior recognition signal.

2. The method according to claim 1, characterized in that The step of determining the region where the phase mutation occurs from the initial phase data sequence according to the initial phase difference value and the initialization step size includes: The initialization step length includes a first preset step length and a second preset step length, and the first preset step length is greater than the second preset step length; Find two adjacent initial phase values ​​whose initial phase difference is greater than the first preset phase difference from the initial phase data sequence, and when the found initial phase values ​​are continuous areas, use the second preset step size to find an area where a sudden phase change occurs in the continuous area; Find two adjacent initial phase values ​​whose initial phase difference is less than the second preset phase difference from the initial phase data sequence, and when the found initial phase values ​​are in a continuous area, use the first preset step size to find an area where a phase mutation occurs in the continuous area.

3. The method according to claim 1, characterized in that The step of calculating the target phase difference between two adjacent initial phase values ​​according to the count value of the dynamic counter includes: Starting from the second initial phase value in the initial phase data sequence, respectively adding the values ​​calculated based on the current counter to obtain a plurality of intermediate phase values; The difference between two adjacent intermediate phase values ​​is used as the target phase difference.

4. The method according to claim 2, characterized in that Determining a target phase mutation value according to the target phase difference value includes: Using a second preset step size, the initial phase values ​​in the region where the phase mutation occurs are sequentially traversed; Find two adjacent intermediate phase values ​​where the target phase difference value is greater than the first preset target phase difference value, and when the intermediate phase value found is a continuous area, reduce the second preset step size by a preset value and then search the continuous area to determine the target phase mutation value; Find two adjacent intermediate phase values ​​where the target phase difference value is less than the second preset target phase difference value. When the intermediate phase value found is a continuous area, increase the second preset step size by a preset value and then search the continuous area to determine the target phase mutation value.

5. The method according to claim 4, characterized in that Calculating the target phase mutation value according to the target phase difference includes: When the target phase difference value is greater than the preset target phase difference value, the preset target value is subtracted from the current target phase mutation value; When the target phase difference value is less than the preset target phase difference value, the preset target value is added to the current target phase mutation value.

6. A device for obtaining human behavior recognition signals, characterized in that: The device comprises: A first calculation module, used to calculate an initial phase difference value according to an initial phase value in an acquired initial phase data sequence; the initial phase data sequence includes a plurality of initial phase values ​​for identifying behaviors of different parts of a human body; A first determination module is used to determine the region where the phase mutation occurs from the initial phase data sequence according to the initial phase difference value and the initialization step size; A second calculation module is used to calculate a target phase difference between two adjacent initial phase values ​​according to the count value of the dynamic counter in the region where the phase mutation occurs, determine a target phase mutation value according to the target phase difference; and calculate the target phase mutation value according to the target phase difference to obtain a target phase data sequence; A first processing module, used for normalizing the target phase data sequence, the received channel state data, and the received signal strength data; A third calculation module is used to calculate the distance function θ according to the following formula based on the distance between the receiving module receiving the human behavior recognition signal and the transmitting module transmitting the human behavior recognition signal, the phase offset value between the receiving module and the transmitting module, and the wavelength of the receiving module; Wherein, d is the distance between the receiving module and the transmitting module; λ is the wavelength of the receiving module; Ω is the phase offset between the receiving module and the transmitting module; The difference between the distance function calculated by the receiving module at the current moment and the distance function calculated at the previous moment is used as the distance function difference generated by the relative motion between the transmitting module and the receiving module; The second processing module is used to calculate the conversion phase data f of each data in the target phase data sequence after the normalization processing according to the following formula d , and obtain the conversion phase data sequence; Wherein, Δt is the time difference generated when the transmitting module and the receiving module move relative to each other; Δθ is the distance function difference generated within the Δt time difference; A third processing module is used to obtain a signal strength data difference according to the acquired two adjacent normalized signal strength data; The fourth processing module is used to fuse the conversion phase data sequence, the signal strength data difference and the channel state data after normalization processing to obtain a fused human behavior recognition signal.

7. An electronic device, characterized in that: The electronic device includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor. The program or instruction is executed by the processor to perform the steps of any one of the methods for obtaining a human behavior recognition signal in claims 1-5.

Images