Intelligent human body behavior sensing system based on Wi-Fi

Abstract

The invention provides an intelligent human body behavior perception system based on Wi-Fi, and relates to the field of behavior perception, the system comprises a signal acquisition module, a data processing module, a deep learning module and a behavior analysis module; the signal acquisition module is used for extracting CSI data information to form a high-dimensional signal data matrix; the data processing module is used for performing singular value decomposition (SVD) denoising, selecting an optimal dimension reduction strategy, executing low-rank matrix decomposition, and filling missing data in combination with nuclear norm minimization to obtain a feature vector; the deep learning module is used for outputting preliminary prediction information; and the behavior analysis module is used for calculating a classification posterior probability in combination with Bayesian reasoning so as to carry out probability correction on the preliminary prediction information. The system can support dynamic behavior recognition in a complex scene, and is high in robustness and adaptability; besides, the system is excellent in real-time performance, collected data can be rapidly processed, behavior feedback can be generated, and the requirements of real-time applications such as smart home and health monitoring are met.

Description

Technical Field

[0001] The present application relates to the field of behavior perception technology, and in particular to an intelligent human behavior perception system based on Wi-Fi. Background Art

[0002] At present, Wi-Fi technology has become an indispensable wireless communication means in daily life, and its application scope has gradually expanded from traditional data transmission to the field of environmental perception. As a technology with high penetration rate, low cost and non-intrusiveness, Wi-Fi signals provide new possibilities for human behavior perception through their propagation characteristics. Studies have shown that human activities can cause changes in the multipath propagation, reflection, diffraction and other characteristics of wireless signals. These changes can be captured by analyzing the received signal strength indicator (RSSI) or channel state information (CSI), thereby realizing non-contact human behavior perception.

[0003] Among related technologies, compared with traditional behavior sensing technologies based on cameras or wearable devices, Wi-Fi behavior sensing technology has the following advantages: (1) Non-intrusiveness and privacy protection: No cameras or wearable sensors are required, avoiding direct infringement of user privacy. (2) Low cost and high coverage: Relying on existing Wi-Fi infrastructure, it is easy to deploy and cost-effective. (3) Wide range of environmental adaptability: Wi-Fi signals can penetrate walls and are also effective in complex indoor environments.

[0004] However, despite its significant advantages, Wi-Fi behavior sensing technology still faces the following challenges: (1) Multipath effect and noise interference: Multipath propagation in indoor environments increases the complexity of signals, and noise further reduces the reliability of recognition. (2) Real-time and dynamic scene adaptation: In multi-user dynamic scenes, how to efficiently process time-varying signals and classify them in real time is an urgent problem to be solved. (3) Fine-grained behavior recognition: The detection of refined behaviors (such as gestures, breathing, etc.) requires higher data resolution and modeling capabilities. (4) Cross-environment versatility: Existing models are usually too dependent on specific environments, and their performance drops significantly when the scene switches. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present application provides an intelligent human behavior perception system based on Wi-Fi, which solves the problems of low recognition accuracy, poor environmental adaptability and insufficient real-time performance in the existing Wi-Fi signal behavior perception technology.

[0006] To achieve the above objectives, this application is implemented through the following technical solutions:

[0007] In the first aspect, an embodiment of the present application provides an intelligent human behavior perception system based on Wi-Fi, which includes a signal acquisition module, a data processing module, a deep learning module and a behavior analysis module; wherein the signal acquisition module is used to use a preset Wi-Fi device to collect channel state information CSI in multiple frequency bands to detect wireless signal changes, and extract CSI data information to form a high-dimensional signal data matrix; the data processing module is used to perform singular value decomposition SVD denoising on the signal data matrix, select the optimal dimensionality reduction strategy through deep reinforcement learning DQN and perform low-rank matrix decomposition, and fill in the missing data in combination with nuclear norm minimization to improve signal stability, and obtain a feature vector.

[0008] Furthermore, the deep learning module is used to process the feature vector in combination with the convolutional neural network CNN to extract the spatial features of the CSI data information, and to perform time series modeling and training in combination with the long short-term memory network LSTM to output preliminary prediction information representing user behavior; the behavior analysis module is used to calculate the classification posterior probability in combination with Bayesian reasoning to perform probability correction on the preliminary prediction information and obtain the user's behavior pattern classification result; and anomaly detection algorithm is used to evaluate the behavior pattern, support adaptive learning, and adjust the anomaly detection threshold according to the user's long-term behavior pattern to improve the classification confidence.

[0009] According to the first aspect of the embodiment of the present application, the intelligent human behavior perception system adopts multiple-input multiple-output MIMO technology and supports IEEE802.11n / ac protocol, and introduces large-scale antenna arrays and beamforming technology to improve the directionality and anti-interference ability of the signal; the Wi-Fi device is an Intel 5300 wireless network card and supports 2.4GHz and 5GHz frequency bands, among which 2.4GHz uses a 40MHz channel and 5GHz uses an 80MHz channel.

[0010] According to the first aspect of the embodiment of the present application, the signal data matrix is ​​subjected to singular value decomposition (SVD) denoising, the optimal dimensionality reduction strategy is selected through deep reinforcement learning (DQN) and low-rank matrix decomposition is performed, and the missing data is filled in combination with nuclear norm minimization to improve signal stability and obtain a feature vector. Specifically, the following steps may be included: the signal data matrix is ​​subjected to singular value decomposition (SVD), and the noise matrix is ​​calculated to remove low-energy noise and improve signal quality; the optimal dimensionality reduction strategy is selected using deep reinforcement learning (DQN), and low-rank matrix decomposition is performed in combination with a gradient optimization method to further reduce the data dimension; the missing values ​​in the CSI data information are filled in combination with an error-first data filling method through deep reinforcement learning (DQN) and nuclear norm minimization to improve signal stability and data integrity and obtain a feature vector.

[0011] According to the first aspect of the embodiment of the present application, the aforementioned deep learning module includes a timing analysis unit, a classification unit and a training output unit; wherein the timing analysis unit is used to capture time series dependencies using a long short-term memory network LSTM according to feature vectors, use Transformer for time series modeling, extract human behavior patterns across time dimensions, and adapt to the behavior recognition needs of different scenarios through online updates and transfer learning; the classification unit is used to perform behavior classification based on the analysis results of the timing analysis unit using a fully connected layer and a Softmax function; the training output unit is used to perform supervised learning using cross entropy loss during the model training phase, and use the Adam optimizer to optimize model parameters; and perform real-time inference after model training to output preliminary prediction information characterizing user behavior.

[0012] According to the first aspect of the embodiment of the present application, the aforementioned method of using a long short-term memory network LSTM to capture time series dependencies based on feature vectors, using Transformer to perform time series modeling, and extracting human behavior patterns across time dimensions can specifically include the following steps: when the feature vector output by the data processing module maintains a certain spatial structure, using 1D-CNN to extract local features, and then inputting them into Transformer for time series modeling; when the feature vector output by the data processing module is a 2D matrix, using 1D-CNN to process local features, and then inputting them into Transformer for behavior modeling.

[0013] According to the first aspect of the embodiment of the present application, the aforementioned behavior analysis module includes: a classification result optimization unit, an intelligent anomaly detection unit and an analysis result output unit; the classification result optimization unit is used to use Bayesian reasoning to calculate the posterior probability of behavior classification and optimize the preliminary prediction information; the intelligent anomaly detection unit is used to construct a sliding window statistical model, calculate the time series change rate of the behavior characteristics, and use the isolation forest algorithm or the local anomaly factor LOF for intelligent anomaly detection to intelligently evaluate human behavior and obtain a behavior pattern classification result; the analysis result output unit is used to trigger a real-time feedback and abnormal alarm mechanism according to the behavior pattern classification result; for normal behavior, the behavior pattern classification result is directly sent to the user interface or the Internet of Things terminal; for detected abnormal behavior, the alarm mechanism is triggered and combined with smart home devices for linkage to improve safety.

[0014] According to the first aspect of the embodiment of the present application, the typical behavior recognition range supported by the aforementioned intelligent human behavior perception system includes typical indoor behaviors, special gestures and breathing detection; typical indoor behaviors include standing, walking, sitting and falling; special gestures include waving movements and commanding movements; breathing detection is used to monitor the user's micro-movements and vital signs.

[0015] In a second aspect, an embodiment of the present application provides an intelligent human behavior perception method based on Wi-Fi, which includes: using a preset Wi-Fi device to collect channel state information CSI in multiple frequency bands through a signal acquisition module to detect wireless signal changes, and extracting CSI data information to form a high-dimensional signal data matrix; performing singular value decomposition SVD denoising on the signal data matrix through a data processing module, selecting the optimal dimensionality reduction strategy and performing low-rank matrix decomposition through deep reinforcement learning DQN, and filling missing data in combination with nuclear norm minimization to improve signal stability, and obtaining a feature vector; processing the feature vector through a deep learning module combined with a convolutional neural network CNN to extract the spatial characteristics of the CSI data information, and combining a long short-term memory network LSTM to perform time series modeling and training to output preliminary prediction information representing user behavior; calculating the classification posterior probability through a behavior analysis module combined with Bayesian reasoning to perform probability correction on the preliminary prediction information to obtain the user's behavior pattern classification result; and using an anomaly detection algorithm to evaluate the behavior pattern, support adaptive learning, and adjust the anomaly detection threshold according to the user's long-term behavior pattern to improve the classification confidence.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, an intelligent human behavior perception method based on Wi-Fi in the aforementioned second aspect is implemented.

[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, an intelligent human behavior perception method based on Wi-Fi in the aforementioned second aspect is implemented.

[0018] This application provides an intelligent human behavior perception system based on Wi-Fi. Compared with the prior art, it has the following beneficial effects:

[0019] This application directly uses Wi-Fi devices to collect channel state information CSI in multiple frequency bands. The system has the characteristics of non-intrusion and privacy protection. It can realize human behavior perception without cameras or wearable devices, avoiding direct infringement of user privacy. The system relies on existing Wi-Fi devices, is low-cost and easy to deploy, and has wide applicability. After collecting data to obtain CSI data information and forming a signal data matrix, the system extracts spatial features and performs time series modeling to judge user behavior patterns by combining deep learning technology. The system can support dynamic behavior recognition in complex scenarios, with strong robustness and high adaptability. In addition, the system performs well in real-time, can quickly process the collected data and generate behavior feedback, and meet the needs of real-time applications such as smart homes and health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 It is a flow chart of an intelligent human behavior perception method based on Wi-Fi provided in an embodiment of the present application;

[0022] Figure 2 It is a structural diagram of a Wi-Fi-based intelligent human behavior perception system provided in an embodiment of the present application;

[0023] Figure 3 This is a module architecture diagram of a Wi-Fi-based intelligent human behavior perception system provided in an embodiment of the present application;

[0024] Figure 4 It is a schematic diagram of a hierarchical architecture of a Wi-Fi-based intelligent human behavior perception system provided in an embodiment of the present application;

[0025] Figure 5 It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0028] The embodiments of the present application provide an intelligent human behavior perception system based on Wi-Fi, thereby solving the problems of low recognition accuracy, poor environmental adaptability and insufficient real-time performance in the existing Wi-Fi signal behavior perception technology.

[0029] The technical solution in the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:

[0030] Wi-Fi technology has become an indispensable wireless communication method in daily life, and its application scope has gradually expanded from traditional data transmission to the field of environmental perception. As a technology with high penetration rate, low cost and non-intrusiveness, Wi-Fi signals provide new possibilities for human behavior perception through their propagation characteristics. Studies have shown that human activities can cause changes in the multipath propagation, reflection, diffraction and other characteristics of wireless signals. These changes can be captured by analyzing the received signal strength indicator (RSSI) or channel state information (CSI), thereby realizing non-contact human behavior perception.

[0031] Compared with traditional behavior sensing technologies based on cameras or wearable devices, Wi-Fi behavior sensing technology has the following advantages: (1) Non-intrusiveness and privacy protection: No cameras or wearable sensors are required, avoiding direct infringement of user privacy. (2) Low cost and high coverage: Relying on existing Wi-Fi infrastructure, it is easy to deploy and cost-effective. (3) Wide environmental adaptability: Wi-Fi signals can penetrate walls and are also effective in complex indoor environments. However, despite the significant advantages of Wi-Fi behavior sensing technology, Wi-Fi behavior sensing technology still faces the following challenges: (1) Multipath effect and noise interference: Multipath propagation in indoor environments increases the complexity of signals, and noise further reduces the reliability of recognition. (2) Real-time and dynamic scene adaptation: In multi-user dynamic scenes, how to efficiently process time-varying signals and classify them in real time is an urgent problem to be solved. (3) Fine-grained behavior recognition: The detection of refined behaviors (such as gestures, breathing, etc.) requires higher data resolution and modeling capabilities. (4) Cross-environment versatility: Existing models are usually too dependent on specific environments, and their performance decreases significantly when scenes switch.

[0032] In recent years, the rapid development of deep learning technology has provided an effective means to solve the above problems. Combining Wi-Fi CSI data with deep learning models, such as convolutional neural networks (CNN) and long short-term memory networks (LSTM), can extract higher-level features from the signal, greatly improving the accuracy and robustness of behavior perception. For example, CNN is good at processing spatial features, while LSTM is able to capture dynamic changes in time series. This combination significantly enhances the classification ability of complex behaviors. In addition, Wi-Fi behavior sensing technology has shown broad application prospects in smart homes, health monitoring, industrial safety, and motion analysis. For example, using Wi-Fi technology to monitor falls of the elderly can provide real-time alarm services; in industrial environments, sensing workers' behavior and posture can help reduce safety accidents; in sports scenes, real-time capture of motion trajectories can provide support for personalized training plans.

[0033] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0034] The following first introduces a Wi-Fi-based intelligent human behavior perception system provided in an embodiment of the present application.

[0035] Please refer to the schematic diagram of a Wi-Fi-based intelligent human behavior perception system 100 provided in the embodiment of the present application. Figure 2 and Figure 3 , the Wi-Fi-based intelligent human behavior perception system 100 may include:

[0036] The signal acquisition module 110 is used to collect channel state information CSI in multiple frequency bands using a preset Wi-Fi device to detect wireless signal changes, and extract CSI data information to form a high-dimensional signal data matrix.

[0037] The data processing module 120 is used to perform singular value decomposition (SVD) denoising on the signal data matrix, select the optimal dimensionality reduction strategy and perform low-rank matrix decomposition through deep reinforcement learning (DQN), and fill in the missing data in combination with nuclear norm minimization to improve signal stability, thereby obtaining a feature vector.

[0038] The deep learning module 130 is used to process the feature vector in combination with the convolutional neural network CNN to extract the spatial features of the CSI data information, and to perform time series modeling and training in combination with the long short-term memory network LSTM to output preliminary prediction information representing user behavior.

[0039] The behavior analysis module 140 is used to combine Bayesian reasoning to calculate the classification posterior probability to make probability corrections to the preliminary prediction information to obtain the user's behavior pattern classification result; and use anomaly detection algorithms to evaluate behavior patterns, support adaptive learning, and adjust anomaly detection thresholds according to users' long-term behavior patterns to improve classification confidence.

[0040] According to an embodiment of the present application, any multiple modules among the signal acquisition module 110, the data processing module 120, the deep learning module 130 and the behavior analysis module 140 can be combined into one module for implementation, or any one of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module.

[0041] The above is a specific implementation of an intelligent human behavior perception system based on Wi-Fi provided in an embodiment of the present application. It should be noted that the CSI data information contains the amplitude and phase information of multiple subcarriers, which can reflect the dynamic impact of human body movements on the signal propagation path in the environment. In order to improve the directionality and accuracy of the collected data, the system combines a multi-antenna configuration to cover a larger environmental range and enhance the capture capability of specific path signals.

[0042] It can be understood that the present application directly uses Wi-Fi devices to collect channel state information CSI in multiple frequency bands. The system is non-invasive and privacy-protecting. It can realize human behavior perception without cameras or wearable devices, avoiding direct infringement of user privacy. The system relies on existing Wi-Fi devices, is low-cost and easy to deploy, and has wide applicability. After collecting data to obtain CSI data information and forming a signal data matrix, the system extracts spatial features and performs time series modeling to judge user behavior patterns by combining deep learning technology. The system can support dynamic behavior recognition in complex scenarios, with strong robustness and high adaptability. In addition, the system performs well in real-time, can quickly process the collected data and generate behavior feedback, and meet the needs of real-time applications such as smart homes and health monitoring.

[0043] The technical solution of this application provides an efficient and accurate solution for intelligent human behavior perception, which is particularly suitable for application scenarios that require real-time monitoring. The system can realize automatic control of equipment in smart homes by identifying user behavior; in the field of health monitoring, timely alarms can be provided by detecting behaviors such as falls and abnormal activities; in industrial scenarios, operational risks can be reduced by monitoring workers' movements and postures; in motion analysis, personalized training suggestions can be provided by capturing users' movement trajectories. This application not only improves the practicality of behavior perception technology, but also provides technical support for future smart life and public safety.

[0044] In one example, the intelligent human behavior perception system adopts multiple-input multiple-output MIMO technology and supports IEEE802.11n / ac protocol. It introduces large-scale antenna arrays and beamforming technology to improve the signal's directionality and anti-interference ability. The Wi-Fi device is an Intel5300 wireless network card and supports 2.4GHz and 5GHz frequency bands, of which 2.4GHz uses a 40MHz channel and 5GHz uses an 80MHz channel.

[0045] In the embodiment of the present application, it can be understood that the system of the present application supports 64-subcarrier CSI extraction, can realize dynamic behavior perception in complex environments, and detect wireless signal changes caused by human activities. In order to improve the overall performance, the system adopts multiple-input multiple-output MIMO technology in hardware implementation, supports 802.11n / ac protocol, and improves the signal directionality and anti-interference ability by introducing large-scale antenna arrays and beamforming technology.

[0046] It should be noted that the Wi-Fi-based intelligent human behavior perception system provided by this application supports real-time optimization and adopts parallel computing and hardware acceleration technologies, such as CUDA-based GPU parallel computing and FPGA hardware acceleration framework, to ensure real-time behavior perception capabilities in complex scenarios. By designing a distributed architecture, the system can coordinate processing between edge devices, edge computing nodes and the cloud to improve response speed and expansion capabilities.

[0047] It should also be noted that the signal acquisition module 110 collects CSI changes caused by human activities in the environment through commercial Wi-Fi devices, the data processing module 120 pre-processes the collected CSI data information to extract effective information with behavioral characteristics, the deep learning module performs behavior recognition and classification based on the processed feature data, and the behavior analysis module 140 applies processing to the recognition results.

[0048] In some embodiments, the above-mentioned singular value decomposition SVD denoising is performed on the signal data matrix, the optimal dimensionality reduction strategy is selected through deep reinforcement learning DQN and low-rank matrix decomposition is performed, and the missing data is filled in combination with nuclear norm minimization to improve signal stability, and the feature vector is obtained, which can specifically include the following steps:

[0049] S210, performing singular value decomposition (SVD) on the signal data matrix, calculating the noise matrix to remove low energy noise and improve signal quality;

[0050] S220, using deep reinforcement learning DQN to select the optimal dimensionality reduction strategy, and combined with the gradient optimization method, perform low-rank matrix decomposition to further reduce the data dimension;

[0051] S230. Through deep reinforcement learning DQN combined with the error-first data filling method, the missing values ​​in the CSI data information are filled using nuclear norm minimization to improve signal stability and data integrity and obtain a feature vector.

[0052] In the embodiment of the present application, it can be understood that in terms of software implementation, the data processing module 120 of the present application adopts a method combining singular value decomposition SVD with deep reinforcement learning DQN optimization matrix decomposition to perform signal denoising to enhance signal quality and environmental adaptability.

[0053] In some embodiments, the aforementioned deep learning module 130 may include:

[0054] The time series analysis unit 310 is used to capture time series dependencies using a long short-term memory network (LSTM) based on feature vectors, use Transformer to perform time series modeling, extract human behavior patterns across time dimensions, and adapt to behavior recognition requirements in different scenarios through online updating and transfer learning.

[0055] The classification unit 320 is used to perform behavior classification based on the analysis result of the time series analysis unit using a fully connected layer and a Softmax function.

[0056] The training output unit 330 is used to use cross entropy loss to perform supervised learning during the model training phase and use the Adam optimizer to optimize model parameters; and to perform real-time reasoning after model training to output preliminary prediction information that characterizes user behavior.

[0057] It is understandable that the deep learning module 130 uses a multi-layer convolutional neural network CNN to extract spatial features from the preprocessed data. CNN extracts local features of the signal through convolution kernels, especially the signal change pattern caused by human body movements on multiple subcarriers. In addition, the system integrates a long short-term memory network LSTM to capture the time series characteristics of CSI data information, further improving the ability to recognize dynamic behaviors. In order to optimize the extraction effect of key features, the deep learning module 130 can also introduce an attention mechanism to highlight key signal features through weighted processing.

[0058] It should be noted that in the training phase, the Wi-Fi-based intelligent human behavior perception system provided by this application uses cross entropy loss for supervised learning and uses the Adam optimizer to optimize model parameters. After the training is completed, the module can directly skip this step and enter real-time reasoning. This application can effectively enhance the environmental adaptability of Wi-Fi CSI signals and improve the stability of the system in different rooms, furniture layouts or changes in Wi-Fi devices.

[0059] In some embodiments, the above method uses a long short-term memory network LSTM to capture time series dependencies based on feature vectors, uses Transformer to perform time series modeling, and extracts human behavior patterns across time dimensions, which may specifically include the following steps:

[0060] S410. When the feature vector output by the data processing module maintains a certain spatial structure, 1D-CNN is used to extract local features, and then the features are input into Transformer for time series modeling.

[0061] S420: When the feature vector output by the data processing module is a 2D matrix, 1D-CNN is used to process local features, and then the features are input into Transformer for behavior modeling.

[0062] In the embodiments of the present application, it can be understood that the present application adopts Transformer / LSTM for time series modeling to extract behavior patterns across time dimensions, thereby improving the adaptability of CSI data information to complex scenarios; when the CSI data information fluctuates greatly, the system will adaptively optimize the behavior category according to the spatiotemporal modeling results, thereby improving the recognition accuracy.

[0063] In some embodiments, the aforementioned behavior analysis module 140 may include:

[0064] A classification result optimization unit 510, for calculating the posterior probability of the behavior classification using Bayesian reasoning to optimize the preliminary prediction information;

[0065] The intelligent anomaly detection unit 520 is used to construct a sliding window statistical model, calculate the time series change rate of the behavior characteristics, and use the isolation forest algorithm or the local anomaly factor LOF to perform intelligent anomaly detection to intelligently evaluate human behavior and obtain a behavior pattern classification result;

[0066] The analysis result output unit 530 is used to trigger real-time feedback and abnormal alarm mechanism according to the behavior pattern classification result; for normal behavior, the behavior pattern classification result is directly sent to the user interface or the Internet of Things terminal; for detected abnormal behavior, the alarm mechanism is triggered and linked with smart home devices to improve safety.

[0067] In the embodiments of the present application, it can be understood that the present application performs probability correction on the CNN / LSTM prediction results and further optimizes the classification results output by the deep learning model to improve the reliability, robustness and environmental adaptability of the system and improve the classification confidence. The present application combines Bayesian reasoning to calculate the classification posterior probability, and uses anomaly detection algorithms to evaluate behavior patterns to reduce false positives, and supports adaptive learning, adjusts anomaly detection thresholds according to users' long-term behavior patterns, and improves personalized analysis capabilities.

[0068] Specifically: In the behavior analysis module 140 of the present application, the system combines classification result optimization and intelligent anomaly detection to perform intelligent analysis of human behavior, and improves the practicality of the system through real-time feedback and abnormal alarm mechanism. The entire behavior analysis process includes three main steps, namely, classification result optimization, intelligent anomaly detection, and output of behavior analysis results, to ensure that the system can accurately identify human behavior in different environments and application scenarios.

[0069] First, the behavior analysis module obtains the classification output of the deep learning model (CNN+LSTM+Transformer), including the behavior category and its confidence. Since the deep learning model may have low classification confidence due to environmental changes or data noise, the system will further calculate the posterior probability of the behavior category and adjust the preliminary classification results in combination with Bayesian reasoning to improve the accuracy and stability of recognition. In addition, in order to prevent the low-confidence classification results from affecting the final decision, the system sets a confidence threshold (such as 70%). If the confidence of the classification result is lower than the threshold, it will enter the anomaly detection step, otherwise it will directly output the classification result.

[0070] Secondly, in order to further improve the ability to detect abnormal behaviors, the behavior analysis module adopts an intelligent anomaly detection mechanism to calculate the changing trend of behavioral features based on a sliding window and evaluate the degree of abnormality of the behavior. The system calculates the mean and standard deviation of the features through historical behavior data, and analyzes the current behavior pattern in combination with anomaly detection algorithms (such as Isolation Forest or LOF). When the anomaly score of the behavior feature exceeds the set threshold, the system determines it as abnormal behavior and triggers the alarm mechanism. If the anomaly score is low, the system determines that the behavior is within the normal range and enters the next step of optimization analysis.

[0071] Finally, the system triggers real-time feedback and abnormal alarm mechanisms based on the analysis results. For normal behavior, the system directly sends the classification results to the user interface or IoT terminal to achieve instant interaction. For abnormal behavior detected, the system will trigger an alarm mechanism and can be linked with smart home devices (such as cameras, door locks, etc.) to improve security. At the same time, abnormal behavior data will be stored and used for subsequent model optimization, so that the system can continuously improve recognition accuracy and environmental adaptability during long-term operation.

[0072] In one example, Bayesian inference is calculated as:

[0073]

[0074] Where P(A) is the prior probability of the behavior category, P(B|A) is the probability predicted by the deep learning classifier, P(A|B) represents the final behavior judgment probability after inference optimization, and P(B) represents the total probability of observing CSI signal feature B under all possible behaviors.

[0075] The above time series change rate satisfies the expression:

[0076]

[0077] Where n represents the total number of sampling times, Trend represents the rate of change of the time series; x i represents the feature vector of the i-th CSI sampling, xi-1 The feature vector representing the i-1th CSI sampling; if the behavior characteristics at a certain moment deviate from the historical behavior pattern by more than the set threshold, the system determines that the behavior is abnormal.

[0078] The anomaly score of intelligent anomaly detection satisfies the expression:

[0079]

[0080] Among them, S t represents the anomaly score at the current moment, x t is the data that characterizes the behavior characteristics at the current moment, μ and σ are the mean and standard deviation of the historical behavior characteristics; if S t If the set threshold is exceeded, the abnormal alarm mechanism will be triggered.

[0081] In one example, the typical behavior recognition range supported by the aforementioned intelligent human behavior perception system includes typical indoor behaviors, special gestures and breathing detection; typical indoor behaviors include standing, walking, sitting and falling; special gestures include waving and commanding movements; breathing detection is used to monitor the user's micro-movements and vital signs.

[0082] In some embodiments, Figure 1 As shown, the present application provides an intelligent human behavior perception method based on Wi-Fi, and the intelligent human behavior perception method may include the following steps:

[0083] S610, using a preset Wi-Fi device to collect channel state information CSI in multiple frequency bands through a signal collection module to detect wireless signal changes, and extracting CSI data information to form a high-dimensional signal data matrix;

[0084] S620, performing singular value decomposition (SVD) denoising on the signal data matrix through the data processing module, selecting the optimal dimensionality reduction strategy and performing low-rank matrix decomposition through deep reinforcement learning (DQN), and filling in missing data in combination with nuclear norm minimization to improve signal stability, thereby obtaining a feature vector;

[0085] S630, processing the feature vector through a deep learning module combined with a convolutional neural network (CNN) to extract spatial features of the CSI data information, and combining a long short-term memory network (LSTM) to perform time series modeling and training to output preliminary prediction information representing user behavior;

[0086] S640. The classification posterior probability is calculated through the behavior analysis module combined with Bayesian reasoning to perform probability correction on the preliminary prediction information to obtain the user's behavior pattern classification result; and anomaly detection algorithm is used to evaluate the behavior pattern, support adaptive learning, and adjust the anomaly detection threshold according to the user's long-term behavior pattern to improve the classification confidence.

[0087] Figure 1 The method shown has the functions of realizing each module in the aforementioned Wi-Fi-based intelligent human behavior perception system and can achieve its corresponding technical effects. For the sake of concise description, it will not be repeated here.

[0088] Through the above steps, this application can effectively reduce false alarms, improve the accuracy of human behavior recognition, and enhance the stability and adaptability of CSI data information in complex environments. This method is not only suitable for smart home scenarios, but can also be widely used in medical health monitoring, industrial safety monitoring and other fields, providing reliable technical support for contactless behavior recognition in different scenarios.

[0089] Figure 4 The schematic diagram of the layered architecture of signal processing and behavior recognition of this application is shown, including four main parts: physical layer, network layer, computing layer and application layer. Each layer plays an important role in the process of human behavior perception to improve the recognition accuracy and environmental adaptability of the system. The specific functions of each layer are as follows:

[0090] At the physical layer, the system uses Wi-Fi transmitters and receivers for signal collection. The Wi-Fi transmitter sends wireless signals, which are affected by factors such as human reflection and furniture obstruction during propagation and eventually reach the Wi-Fi receiver. The signals include direct-view signals and reflected signals, of which the reflected signals carry information about human activities. By analyzing the channel state information, human behavior characteristics can be extracted to provide data support for subsequent processing.

[0091] At the network layer, the system is responsible for the collection and transmission of CSI data information. The Wi-Fi receiver collects the original CSI data information and transmits the data to the computing layer for processing through the local network or wirelessly. In this process, the network layer is mainly responsible for data formatting and basic encapsulation of CSI data information to ensure that the subsequent computing layer can correctly parse and use the data for in-depth analysis.

[0092] At the computing layer, the system preprocesses, extracts features, and classifies CSI data information. First, SVD noise reduction, matrix decomposition, and data filling techniques are used to enhance signal stability and reduce data noise. In addition, the system introduces deep reinforcement learning DQN to automatically select the optimal matrix decomposition strategy to optimize the processing effect of CSI data information. Then, the system further performs time series modeling on the reduced CSI features to extract behavioral patterns across time dimensions.

[0093] At the application layer, the system combines Bayesian reasoning, intelligent anomaly detection and behavioral decision-making to analyze human behavior. First, the system calculates the posterior probability of the classification result based on Bayesian reasoning, optimizes the preliminary classification results of deep learning, and improves the classification confidence. Then, the system uses an intelligent anomaly detection method to analyze the behavioral characteristics. The system calculates the changing trend of behavioral characteristics through a sliding window statistical model, and combines the isolated forest algorithm or the local anomaly factor LOF for intelligent anomaly detection. If the anomaly score exceeds the threshold, the system triggers an abnormal alarm. Finally, the system makes real-time feedback and behavioral decisions based on the analysis results. If the behavior is normal, the classification results are sent to the user interface or the IoT terminal to achieve interactive control; if abnormal behavior (such as falling) is detected, the alarm mechanism is triggered, and smart home devices (such as cameras and door locks) can be linked for early warning. In addition, the system supports adaptive optimization, stores abnormal behavior data, and uses it for model updates to improve long-term stability and recognition accuracy.

[0094] In some embodiments, the present application provides an electronic device, the structure diagram of the electronic device is as follows Figure 5 shown.

[0095] The electronic device may include a processor 710 and a memory 720 storing computer program instructions.

[0096] Specifically, the processor 710 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0097] The memory 720 may include a large capacity memory for data or instructions. By way of example and not limitation, the memory 720 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive or a combination of two or more of these. In appropriate cases, the memory 720 may include a removable or non-removable (or fixed) medium. In appropriate cases, the memory 720 may be inside or outside the integrated gateway disaster recovery device. In a specific embodiment, the memory 720 is a non-volatile solid-state memory.

[0098] The memory 720 may include a read-only memory (ROM), a random access memory (RAM), a disk storage medium device, an optical storage medium device, a flash memory device, an electrical, optical or other physical / tangible memory storage device. Therefore, generally, the memory 720 includes one or more tangible (non-transitory) computer-readable storage media (e.g., a memory device) encoded with software including computer executable instructions, and when the software is executed (e.g., by one or more processors), it can perform the operations described in the above-mentioned embodiment of a Wi-Fi-based intelligent human behavior perception method.

[0099] The processor 710 implements a Wi-Fi-based intelligent human behavior perception method in the above embodiment by reading and executing computer program instructions stored in the memory 720 .

[0100] In one example, the electronic device may further include a communication interface 730 and a bus 700. Figure 5 As shown, the processor 710, the memory 720, and the communication interface 730 are connected via a bus 700 and communicate with each other.

[0101] The communication interface 730 is mainly used to implement communication between various modules, devices, units and / or equipment in the embodiments of the present application.

[0102] Bus 700 includes hardware, software or both, and the parts of online data flow billing equipment are coupled to each other. For example, but not limitation, bus may include accelerated graphics port (AGP) or other graphics bus, enhanced industrial standard architecture (EISA) bus, front-end bus (FSB), hypertransport (HT) interconnection, industrial standard architecture (ISA) bus, infinite bandwidth interconnection, low pin count (LPC) bus, memory bus, micro channel architecture (MCA) bus, peripheral component interconnection (PCI) bus, PCI-Express (PCI-X) bus, serial advanced technology attachment (SATA) bus, video electronics standard association local (VLB) bus or other suitable bus or two or more of these combinations. In appropriate cases, bus 700 may include one or more buses. Although the present application embodiment describes and shows a specific bus, the present application considers any suitable bus or interconnection.

[0103] In addition, in combination with the intelligent human behavior perception method based on Wi-Fi in the above embodiment, the embodiment of the present application can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when the computer program instructions are executed by the processor, the intelligent human behavior perception method based on Wi-Fi in the above embodiment is implemented.

[0104] It should be clear that the present application is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present application.

[0105] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present application are programs or code segments that are used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link by a data signal carried in a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, optical fiber media, radio frequency (RF) links, etc. The code segment can be downloaded via a computer network such as the Internet, an intranet, etc.

[0106] It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, this application is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiment, or in a different order from the embodiment, or several steps can be performed simultaneously.

[0107] Aspects of the present disclosure are described above with reference to the flowchart and / or block diagram of the method, device (system) and computer program product according to the embodiment of the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device to produce a machine so that these instructions executed by the processor of the computer or other programmable data processing device enable the implementation of the function / action specified in one or more boxes of the flowchart and / or block diagram. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field programmable logic circuit. It can also be understood that each box in the block diagram and / or flowchart and the combination of boxes in the block diagram and / or flowchart can also be implemented by dedicated hardware that performs a specified function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0108] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent human behavior perception system based on Wi-Fi, characterized in that: include: The signal acquisition module is used to collect channel state information CSI in multiple frequency bands using a preset Wi-Fi device to detect wireless signal changes, and extract CSI data information to form a high-dimensional signal data matrix; A data processing module is used to perform singular value decomposition (SVD) denoising on the signal data matrix, select the optimal dimensionality reduction strategy and perform low-rank matrix decomposition through deep reinforcement learning (DQN), and fill in missing data in combination with nuclear norm minimization to improve signal stability, thereby obtaining a feature vector; A deep learning module, used to process the feature vector in combination with a convolutional neural network (CNN) to extract the spatial features of the CSI data information, and to perform time series modeling and training in combination with a long short-term memory network (LSTM) to output preliminary prediction information representing user behavior; A behavior analysis module, used to calculate the classification posterior probability in combination with Bayesian reasoning to perform probability correction on the preliminary prediction information and obtain the user's behavior pattern classification result; It also uses anomaly detection algorithms to evaluate behavioral patterns, supports adaptive learning, and adjusts anomaly detection thresholds based on users' long-term behavioral patterns to improve classification confidence.

2. The Wi-Fi-based intelligent human behavior perception system as claimed in claim 1, characterized in that ,The intelligent human behavior perception system adopts multiple input multiple output MIMO technology and supports IEEE802.11n / ac protocol, and introduces large-scale antenna array and beamforming technology to improve the signal directivity and anti-interference capability; The Wi-Fi device is an Intel 5300 wireless network card and supports 2.4 GHz and 5 GHz frequency bands, wherein 2.4 GHz uses a 40 MHz channel and 5 GHz uses an 80 MHz channel.

3. The Wi-Fi-based intelligent human behavior perception system according to claim 1, characterized in that: The signal data matrix is ​​subjected to singular value decomposition (SVD) denoising, the optimal dimensionality reduction strategy is selected through deep reinforcement learning (DQN) and low-rank matrix decomposition is performed, and the missing data is filled in with nuclear norm minimization to improve signal stability. Get the feature vector, including: Performing singular value decomposition (SVD) on the signal data matrix and calculating a noise matrix to remove low energy noise and improve signal quality; Deep reinforcement learning DQN is used to select the optimal dimensionality reduction strategy, and combined with gradient optimization methods, low-rank matrix decomposition is performed to further reduce the data dimension; Through deep reinforcement learning DQN combined with the error-first data filling method, the missing values ​​in the CSI data information are filled by minimizing the nuclear norm to improve signal stability and data integrity and obtain a feature vector.

4. The Wi-Fi-based intelligent human behavior perception system according to any one of claims 1 to 3, characterized in that: The deep learning module includes: A time series analysis unit is used to capture time series dependencies using a long short-term memory network (LSTM) according to the feature vector, perform time series modeling using a Transformer, extract human behavior patterns across time dimensions, and adapt to behavior recognition requirements in different scenarios through online updating and transfer learning; A classification unit, used for classifying behaviors using a fully connected layer and a Softmax function based on the analysis result of the time series analysis unit; The training output unit is used to perform supervised learning using cross entropy loss during the model training phase and optimize model parameters using the Adam optimizer; and to perform real-time inference after model training to output preliminary prediction information that characterizes user behavior.

5. The Wi-Fi-based intelligent human behavior perception system as claimed in claim 4, characterized in that: The method of using a long short-term memory network (LSTM) to capture time series dependencies based on the feature vector and using Transformer to perform time series modeling to extract human behavior patterns across time dimensions includes: When the feature vector output by the data processing module maintains a certain spatial structure, 1D-CNN is used to extract local features, and then the features are input into Transformer for time series modeling; When the feature vector output by the data processing module is a 2D matrix, 1D-CNN is used to process local features, and then the features are input into Transformer for behavior modeling.

6. The Wi-Fi-based intelligent human behavior perception system according to any one of claims 1 to 3, characterized in that: The behavior analysis module includes: A classification result optimization unit, used to calculate the posterior probability of the behavior classification using Bayesian reasoning to optimize the preliminary prediction information; Intelligent anomaly detection unit, used to build a sliding window statistical model, calculate the time series change rate of behavioral characteristics, and use the isolation forest algorithm or local anomaly factor LOF to perform intelligent anomaly detection to intelligently evaluate human behavior and obtain behavioral pattern classification results; The analysis result output unit is used to trigger real-time feedback and abnormal alarm mechanism according to the behavior pattern classification result; for normal behavior, the behavior pattern classification result is directly sent to the user interface or the Internet of Things terminal; for detected abnormal behavior, the alarm mechanism is triggered and linked with smart home devices to improve safety.

7. The Wi-Fi-based intelligent human behavior perception system according to any one of claims 1 to 3, characterized in that: The typical behavior recognition range supported by the intelligent human behavior perception system includes typical indoor behaviors, special gestures and breathing detection; the typical indoor behaviors include standing, walking, sitting and falling; the special gestures include waving and commanding; the breathing detection is used to monitor the user's micro-movements and vital signs.

8. A Wi-Fi-based intelligent human behavior perception method, characterized in that: include: The signal acquisition module uses a preset Wi-Fi device to collect channel state information CSI in multiple frequency bands to detect wireless signal changes, and extracts CSI data information to form a high-dimensional signal data matrix; The signal data matrix is ​​subjected to singular value decomposition (SVD) denoising through a data processing module, an optimal dimensionality reduction strategy is selected through deep reinforcement learning (DQN) and low-rank matrix decomposition is performed, and missing data is filled in combination with nuclear norm minimization to improve signal stability, thereby obtaining a feature vector; The feature vector is processed by a deep learning module in combination with a convolutional neural network (CNN) to extract the spatial features of the CSI data information, and a long short-term memory network (LSTM) is used to perform time series modeling and training to output preliminary prediction information representing user behavior; The behavior analysis module is combined with Bayesian reasoning to calculate the classification posterior probability to perform probability correction on the preliminary prediction information to obtain the user's behavior pattern classification result; It also uses anomaly detection algorithms to evaluate behavioral patterns, supports adaptive learning, and adjusts anomaly detection thresholds based on users' long-term behavioral patterns to improve classification confidence.

9. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, the intelligent human behavior perception method based on Wi-Fi as claimed in claim 8 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the intelligent human behavior perception method based on Wi-Fi as claimed in claim 8 is implemented.

Images