Security and protection method and system based on smart home

Abstract

The invention relates to the technical field of smart home security and protection, and discloses a smart home security and protection system based on a body, and the method comprises the steps: collecting real-time sensing data through Wi-Fi multipath scattered waves, constructing a body sensing network, recognizing behavior feature information, optimizing a preset security and protection model decision architecture, and obtaining an optimized security and protection process. The method comprises the following steps: firstly, generating a security response path, determining a personnel activity space-time distribution state, calculating a security efficiency value, then determining an optimization direction according to the security efficiency value, generating a dynamic protection organization unit, calculating a misjudgment attenuation rate, and finally determining a security state and analyzing a security dimension according to the misjudgment attenuation rate, thereby constructing a more accurate and intimate intelligent security scheme. According to the invention, the intelligent level of autonomous decision making and active protection of home security can be improved.

Description

Technical Field

[0001] The present invention relates to a method and system for embodied smart home security, belonging to the technical field of smart home security. Background Art

[0002] Embodied smart home security refers to the combination of multimodal sensing devices embedded in the environment (such as visual sensors, infrared radars, sound recognition modules, etc.) with autonomous decision-making systems to achieve real-time security protection in home scenarios.

[0003] At present, existing technologies mainly rely on passive monitoring of static security equipment (such as cameras, door and window sensors), or trigger alarms based on a single rule (such as infrared motion detection). They lack multi-dimensional dynamic analysis of complex home scenarios (such as behavioral intention recognition and multi-device collaborative response). For example, the system may mistakenly trigger an alarm due to pet activities, or be unable to distinguish between elderly falls and intrusions. At the same time, traditional solutions rely on cloud-based decision-making, resulting in local response delays and failure in network disconnection scenarios. This limitation reduces the accuracy and real-time performance of the security system, easily causing false alarms and missed alarms, and cannot achieve anthropomorphic situational awareness decision-making. For example, it is impossible to dynamically adjust deployment strategies according to family member identities and work and rest patterns, or identify complex scenarios such as forced door opening. Therefore, a method based on embodied smart home security is needed to improve the intelligence level of home security systems through environment-embedded multimodal perception and localized decision-making. Summary of the Invention

[0004] The present invention provides a method and system for home security based on embodied smart technology, the main purpose of which is to improve the security level of the home security system.

[0005] To achieve the above objectives, the present invention provides a security method based on embodied smart home, comprising:

[0006] Acquiring real-time perception data corresponding to a residential environment, collecting behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, analyzing dynamic activity curves corresponding to the behavioral characteristic signals, and constructing an embodied perception network corresponding to the residential environment based on the dynamic activity curves;

[0007] Based on the embodied perception network, behavioral feature information in the residential environment is identified; based on the behavioral feature information, a decision architecture in a preset security model is analyzed; behavioral matching relationships in the decision architecture are queried; the behavioral matching relationships are adaptively optimized to obtain an optimized security process;

[0008] generating a security response path corresponding to the residential environment based on the optimized security process, determining a spatiotemporal distribution of human activities in the residential environment based on the security response path, and calculating a security efficiency value corresponding to the residential environment based on the spatiotemporal distribution;

[0009] Determining a security optimization direction corresponding to the residential environment based on the security efficiency value, generating a dynamic deployment unit corresponding to the residential environment based on the security optimization direction, extracting key execution parameters from the dynamic deployment unit, and calculating a false positive attenuation rate corresponding to the residential environment based on the key execution parameters;

[0010] Based on the misjudgment attenuation rate, the security status corresponding to the residential environment is determined, the security dimension of the residential environment corresponding to the security status is analyzed, and based on the security dimension, an intelligent security solution corresponding to the residential environment is constructed.

[0011] Optionally, constructing an embodied perception network corresponding to the residential environment according to the dynamic activity curve includes:

[0012] Extracting a time series feature vector from the dynamic activity curve;

[0013] Collecting spatial topological data in the residential environment and generating a spatial perception matrix corresponding to the spatial topological data;

[0014] identifying a responsive topological link in the spatial perception matrix;

[0015] Extracting a link response node in the response topology link;

[0016] An embodied perception network corresponding to the residential environment is constructed based on the link response node.

[0017] Optionally, identifying behavioral characteristic information in the residential environment based on the embodied perception network includes:

[0018] Obtaining an original behavior signal output by the embodied perception network;

[0019] Segmenting and calibrating the original behavior signal to obtain behavior time sequence segments;

[0020] Performing cross-modal fusion on the behavior time sequence segments to obtain a behavior semantic vector;

[0021] Performing behavior matching on the behavior semantic vector and a preset behavior template library to obtain a behavior matching label;

[0022] Identify the behavior feature information in the behavior matching tag.

[0023] Optionally, the adaptively optimizing the behavior matching relationship to obtain an optimized security process includes:

[0024] Query the historical security event set corresponding to the behavior matching relationship;

[0025] Extracting abnormal behavior features corresponding to each event in the historical security event set;

[0026] Analyzing the behavioral correlation mechanism between the abnormal behavioral characteristics;

[0027] Analyzing the mechanism linkage data corresponding to the behavior association mechanism;

[0028] Based on the mechanism linkage data, the behavior matching relationship is adaptively optimized to obtain an optimized security process.

[0029] Optionally, generating a security response path corresponding to the residential environment based on the optimized security process includes:

[0030] Analyze the security requirement level corresponding to the optimized security process;

[0031] Based on the security requirement level, collecting real-time security signals in the residential environment;

[0032] Performing channel mapping on the real-time security signal to obtain a signal response channel;

[0033] collecting key response points in the signal response channel;

[0034] Based on the key response points, a security response path corresponding to the residential environment is generated.

[0035] Optionally, calculating the security efficiency value corresponding to the residential environment according to the spatiotemporal distribution state includes:

[0036] The security efficiency value corresponding to the residential environment is calculated using the following formula:

[0037]

[0038] Where EZ represents the security efficiency value corresponding to the residential environment, n represents the number of spatial areas divided by the residential area, i represents the number index corresponding to the spatial area, C i represents the actual coverage efficiency of the i-th spatial region, T i represents the total length of time a person stays in the i-th spatial area during the statistical period, m represents the total number of security events during the statistical period, j represents the number index corresponding to the security event, and R j represents the response efficiency index corresponding to the j-th security event, β represents the event priority weight, and D irepresents the real-time monitoring density of the i-th spatial region, and Ω represents the time dimension weight corresponding to the spatiotemporal distribution state.

[0039] Optionally, determining a security optimization direction corresponding to the residential environment based on the security efficiency value includes:

[0040] Analyze the weak link parameters corresponding to the security efficiency value;

[0041] Locating inefficient coverage areas of security equipment in the residential environment based on the weak link parameters;

[0042] Extracting security log data corresponding to the inefficient coverage area;

[0043] Locating a security optimization area corresponding to the residential environment based on the security log data and historical intrusion events;

[0044] Determine the security optimization direction corresponding to the security optimization area.

[0045] Optionally, calculating the misjudgment attenuation rate corresponding to the residential environment based on the key execution parameter includes:

[0046] The false positive attenuation rate corresponding to the residential environment is calculated using the following formula:

[0047]

[0048] Among them, R fad represents the false positive decay rate corresponding to the residential environment, l represents the total number of false positive types, k represents the number index corresponding to the false positive type, α k represents the adjustment coefficient of the key execution parameter corresponding to the kth false alarm type, γ k Indicates the sensitivity coefficient corresponding to the kth false alarm type, ERR k represents the false alarm rate of the kth false alarm type before the key execution parameters are adjusted, W k Indicates the weight coefficient corresponding to the k-th false alarm type.

[0049] Optionally, determining the security status corresponding to the residential environment based on the misjudgment attenuation rate includes:

[0050] Setting a dynamic monitoring threshold corresponding to the residential environment according to the false positive attenuation rate;

[0051] Configuring the signal sampling frequency corresponding to the dynamic monitoring threshold;

[0052] screening abnormal signal segments in the residential environment based on the signal sampling frequency;

[0053] generating a warning signal instruction corresponding to the abnormal signal segment;

[0054] Based on the early warning signal instruction, the security status corresponding to the residential environment is determined.

[0055] In order to solve the above problems, the present invention further provides an embodied smart home security system, the system comprising:

[0056] a network construction module for acquiring real-time perception data corresponding to a residential environment, collecting behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, analyzing dynamic activity curves corresponding to the behavioral characteristic signals, and constructing an embodied perception network corresponding to the residential environment based on the dynamic activity curves;

[0057] a process optimization module for identifying behavioral characteristic information in the residential environment based on the embodied perception network, analyzing a decision architecture in a preset security model based on the behavioral characteristic information, querying behavioral matching relationships in the decision architecture, and adaptively optimizing the behavioral matching relationships to obtain an optimized security process;

[0058] an efficiency value calculation module, configured to generate a security response path corresponding to the residential environment based on the optimized security process, determine a spatiotemporal distribution of human activities in the residential environment based on the security response path, and calculate a security efficiency value corresponding to the residential environment based on the spatiotemporal distribution;

[0059] a decay rate calculation module, configured to determine a security optimization direction corresponding to the residential environment based on the security efficiency value, generate a dynamic deployment unit corresponding to the residential environment based on the security optimization direction, extract key execution parameters from the dynamic deployment unit, and calculate a false positive decay rate corresponding to the residential environment based on the key execution parameters;

[0060] A solution construction module is used to determine the security status corresponding to the residential environment based on the misjudgment attenuation rate, analyze the security dimensions of the residential environment corresponding to the security status, and construct an intelligent security solution corresponding to the residential environment based on the security dimensions.

[0061] Compared with the problems described in the background technology, the present invention obtains real-time perception data corresponding to the residential environment and collects behavioral feature signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves. It can realize camera-free monitoring under the premise of protecting privacy and accurately capture subtle abnormal behaviors such as climbing and rummaging; multi-dimensional data fusion can effectively distinguish scenes such as pet activities and elderly falls, and significantly reduce the false alarm rate of the security system. The present invention is based on the embodied perception network to identify behavioral feature information in the residential environment. It can use dynamic graph structure modeling to collaboratively perceive data of multiple devices and accurately capture cross-regional behavioral associations. The real-time updated network topology can adapt to environmental changes, improve the robustness and timeliness of behavior recognition in complex scenarios, and provide more reliable feature basis for security decision-making. Furthermore, based on the optimized security process, the present invention generates a security response path corresponding to the residential environment. The shortest execution link can be automatically planned based on the dynamic rule priority, and the path reliability can be verified through a rehearsal mechanism to ensure that responses are not missed in multiple scenarios, forming a precise protection system with a full-process closed loop. Furthermore, the present invention determines the security optimization direction corresponding to the residential environment based on the security efficiency value, and can accurately locate the weak links of the system. For example, if the coverage efficiency of a certain area is low or the response time is long, resources can be reasonably allocated accordingly, sensors can be added or equipment can be upgraded. The security strategy can be dynamically adjusted according to the weights of different time periods to achieve efficient resource utilization and improve the overall security level. Finally, the present invention determines the security status corresponding to the residential environment based on the false positive attenuation rate, and can intuitively quantify the improvement of the false alarm of the security system, accurately evaluate the reliability of the security system, and timely discover security loopholes. For example, if the security status of a certain area is poor due to frequent false positives, it is convenient to make targeted adjustments and optimizations to improve security efficiency. Therefore, the embodiment of the present invention provides a security method and system based on embodied smart home, which can improve the security level of the home security system. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 A schematic diagram of a process of implementing an embodied smart home security method according to an embodiment of the present invention;

[0063] Figure 2 A schematic diagram of an embodied perception network in an embodied smart home security method provided by one embodiment of the present invention;

[0064] Figure 3 A schematic diagram of modules for implementing an embodied smart home security system provided in one embodiment of the present invention.

[0065] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0066] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0067] The embodiments of the present application provide a method for embodied smart home security. The method can be executed by at least one of a server, a terminal, or other electronic device capable of executing the method provided by the embodiments of the present application. In other words, the method can be executed by software or hardware installed on a terminal device or a server device. The server can include, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0068] Example 1:

[0069] Reference Figure 1 FIG. 1 is a flow chart of a method for embodied smart home security provided by an embodiment of the present invention. In this embodiment, the method for embodied smart home security includes:

[0070] S1. Acquire real-time perception data corresponding to a residential environment, collect behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, analyze dynamic activity curves corresponding to the behavioral characteristic signals, and construct an embodied perception network corresponding to the residential environment based on the dynamic activity curves.

[0071] By acquiring real-time perception data corresponding to the residential environment and collecting behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, the present invention can achieve camera-free monitoring while protecting privacy, and accurately capture subtle abnormal behaviors such as climbing and rummaging; multi-dimensional data fusion can effectively distinguish between scenes such as pet activities and elderly falls, significantly reducing the false alarm rate of the security system.

[0072] Among them, the residential environment refers to the physical environment within the family living space, covering areas such as the living room, bedroom, kitchen, balcony, etc., including static layouts such as furniture placement, door and window status, and distribution of electrical equipment, as well as dynamic scenes such as human activities, pet movement, and changes in the location of items. For example, the position of the sofa and coffee table in the living room, whether the windows are closed, the walking trajectory of the elderly in the bedroom, the activities of pets in the kitchen, etc., all belong to the category of residential environment; the real-time perception data refers to environmental information data collected in real time by various sensors deployed in the house (such as visual, infrared, sound, gas sensors, etc.), including dynamic or static information such as images, temperature, humidity, sound waveforms, and human movement trajectories. For example, indoor scenes captured by cameras, human movement speed and direction monitored by infrared radars, and glass breaking sounds or cries for help collected by microphones are all real-time perception data; the wireless Wi-Fi multipath scattered waves refer to multiple propagation path signals formed by reflection, refraction, and diffraction of Wi-Fi signals during propagation due to obstacles such as human bodies and furniture. These signals carry the position, shape, and movement status of objects in the environment. State information, for example, when someone walks in a room, their body reflects Wi-Fi signals, causing slight changes in the phase and intensity of the signal received by the receiving end. These changes constitute multipath scattered waves. The behavioral characteristic signal refers to a characteristic signal extracted from real-time perception data and Wi-Fi multipath scattered waves, which can characterize the behavior pattern of people or objects, such as movement amplitude, speed, trajectory, signal frequency domain changes, etc. For example, by analyzing the Doppler frequency shift of the Wi-Fi signal, the high-frequency disturbance characteristics corresponding to the behavior of "climbing a window" can be extracted; from infrared radar data, the sudden acceleration change characteristics when "an elderly person falls" can be extracted, both of which are behavioral characteristic signals. Optionally, the acquisition of real-time perception data corresponding to the residential environment can be achieved through a sensor network, such as an Internet of Things node composed of temperature sensors, humidity sensors, infrared sensors, sound sensors, light sensors, etc., to ultimately obtain real-time perception data. The collection of the behavioral characteristic signal corresponding to the real-time perception data can be achieved through a signal processing algorithm, such as Fourier transform, wavelet transform, Kalman filter, etc., to ultimately obtain the behavioral characteristic signal.

[0073] Specifically, by integrating multimodal data through the Transformer architecture and based on reinforcement learning algorithms, a disposal plan can be generated without a preset program. Action strategies can be optimized through experience accumulation, and multiple devices can share data to form swarm intelligence.

[0074] Furthermore, by analyzing the dynamic activity curves corresponding to the behavioral characteristic signals, the present invention can accurately capture subtle changes in behavioral patterns (such as speed mutation points of abnormal movements) through visual time series features, effectively distinguishing complex scenarios such as intrusions and falls from daily activities, thereby significantly improving the accuracy and real-time performance of abnormal behavior identification and reducing false alarms and missed alerts.

[0075] The dynamic activity curve refers to a visualization curve generated by unfolding behavioral characteristic signals (such as Wi-Fi signal phase changes, displacement data monitored by infrared radar, etc.) in a time series and performing mathematical transformations (such as Fourier transform and sliding average) to reflect the dynamic characteristics of the behavior in the time domain (time-amplitude) or frequency domain (frequency-energy). For example, by analyzing the phase fluctuations of multipath scattered waves of Wi-Fi signals, a "time-phase offset" curve is plotted. During normal walking, the curve exhibits low-frequency stable fluctuations, while when climbing a window, the curve exhibits high-frequency violent oscillations. This curve is the dynamic activity curve representing the climbing behavior. Optionally, analyzing the dynamic activity curve corresponding to the behavioral characteristic signal can be achieved using time series analysis methods, such as sliding window statistics, dynamic time warping (DTW), autoregressive integrated moving average model (ARIMA), and other algorithms to ultimately obtain the dynamic activity curve.

[0076] Furthermore, the present invention constructs an embodied perception network corresponding to the residential environment based on the dynamic activity curve, which can convert scattered multi-source sensor data into a structured environmental model, and dynamically map spatial topology and behavioral associations through graph neural networks (such as sudden changes in signals in door and window areas to enhance monitoring by nearby sensors), thereby achieving cross-device collaborative perception and improving the accuracy of behavioral understanding in complex scenarios.

[0077] Among them, the embodied perception network refers to a dynamic graph network constructed with link response nodes as graph structure nodes and response topology links as edges. It can map the interactive relationship between behavior and space in the environment in real time, and support multi-device collaborative perception and behavior pattern reasoning. For example, a network composed of nodes and signal links such as kitchen gas sensors, living room cameras, and smart gateways can dynamically adjust the node connection weights according to the concentration changes and personnel movement trajectories during gas leaks to achieve collaborative risk perception.

[0078] As an embodiment of the present invention, constructing an embodied perception network corresponding to the residential environment based on the dynamic activity curve includes: extracting the time series feature vector in the dynamic activity curve; collecting spatial topology data in the residential environment and generating a spatial perception matrix corresponding to the spatial topology data; identifying the response topology link in the spatial perception matrix; extracting the link response node in the response topology link; and constructing the embodied perception network corresponding to the residential environment based on the link response node.

[0079] Among them, the time series feature vector refers to the set of time series features extracted from the dynamic activity curve, which describes the time domain and frequency domain variation law of the behavior in the form of a numerical vector, including the time series distribution characteristics of parameters such as amplitude, frequency, and energy. For example, from the human movement curve monitored by infrared radar, data such as the displacement change value per second and the speed fluctuation frequency are extracted to form vectors with dimensions such as 0.5m / s, 2Hz, and 15J, which are the time series feature vectors of walking behavior. The spatial topology data refers to the structured data of the spatial position, layout relationship, and signal transmission properties of physical entities in a residential environment, covering geometric and physical information such as sensor coordinates, furniture distribution, and signal path obstruction. For example, the camera in the bedroom is located at point (2,3) and the air conditioner is located at point (4,1). There is a wall between the two and the Wi-Fi router, and the signal transmission distances are 4m and 5.5m respectively. Both belong to the category of spatial topology data. The spatial perception matrix refers to a mathematical matrix constructed based on spatial topology data, which is used to quantify the perceptual association strength between entities in the environment. The matrix elements are usually real-time interaction parameters such as signal strength, transmission delay, and attenuation coefficient. For example, 4×4 In the spatial perception matrix N, N[2,3]=-75dB means that the signal strength between sensor 2 and sensor 3 is -75dB, and N[3,4]=8ms means that the transmission delay of the signal from sensor 3 to sensor 4 is 8ms, reflecting the perception linkage of spatial entities. The response topology link refers to the combination of transmission paths in the spatial perception matrix that produces a significant signal response to a specific behavior, which is manifested as a set of links with abnormal parameter changes in the matrix (such as sudden changes in strength and sudden increases in delay). For example, when someone pries the door or window, the signal strength between the door or window sensor and the nearby infrared module changes from - The signal level drops from 65dB to -95dB. This link and the adjacent Wi-Fi signal link together form an "intrusion response" topology link. The link response node refers to the sensor or device node in the response topology link that plays a decisive role in the change of behavioral signals. It is usually the key endpoint for signal parameter mutation and undertakes the core functions of data collection and feature extraction. For example, in the "rummaging through the drawer" behavior, the drawer's built-in pressure sensor (detecting opening and closing actions) and the upper Wi-Fi signal receiver (monitoring hand movement scattering) serve as the nodes at both ends of the link and become link response nodes due to signal anomalies.

[0080] Furthermore, the extraction of the time series feature vectors in the dynamic activity curve can be achieved through a time series analysis method, such as: using wavelet transform combined with the signal.wavelet function of the SciPy library to calculate multi-scale time-frequency features, and finally obtaining a time series feature vector containing time domain and frequency domain information (RSSI, delay, phase, spectrum energy, wavelet standard deviation); the collection of spatial topology data in the residential environment can be achieved through three-dimensional scanning technology, such as: using a LiDAR sensor combined with the point cloud processing function of the Open3D library to obtain the room geometry, and finally obtaining spatial topology data with spatial coordinate relationships; the generation of the spatial perception matrix corresponding to the spatial topology data can be achieved through a matrix decomposition algorithm, such as: based on non-negative matrix decomposition combined with the NMF model of the Scikit-learn library to reduce the dimensionality of spatial features, and finally obtain a spatial topology data reflecting the environment. A spatial perception matrix with a regular distribution pattern; the identification of the response topology link in the spatial perception matrix can be achieved by a graph search algorithm, such as: using the Dijkstra algorithm in combination with Python's igraph library to calculate the shortest path set, and finally obtaining the response topology link that describes the spatial connectivity; the extraction of the link response node in the response topology link can be achieved by a key point detection method, such as: using Harris corner detection in combination with the cornerHarris function of the OpenCV library to locate the topological turning point, and finally obtaining the link response node that represents the path change; the construction of the embodied perception network corresponding to the residential environment can be achieved by complex network modeling technology, such as: learning node representation based on the graph embedding algorithm combined with TensorFlow's GraphSAGE framework, and finally obtaining an embodied perception network that integrates spatial cognitive logic.

[0081] Specifically, extracting the timing feature vector from the dynamic activity curve includes reading the RSSI value output by the Wi-Fi module, or obtaining multiple RSSI values ​​through a multi-antenna system, obtaining the delay component in the channel state information (CSI), or using a multipath propagation model to estimate the delay, detect signal attenuation, and estimate the distance.

[0082] Specifically, to further understand the structure of the embodied perception network in this application, please refer to Figure 2 The image shown is a schematic diagram of the embodied perception network provided by the present invention. It should be noted that, in the present invention, Figure 2The schematic diagram presented is only used to show the structural architecture of the embodied perception network, in which the embodied perception network receives the MFCC feature sequence input to the GRU network, and receives the dynamic activity curve graph input to the CNN network. The output of the GRU layer and the output of the CNN layer are jointly input to the fully connected layer 1 and the fully connected layer 2, and then output through the softmax layer. This network structure is closely related to the process of constructing the embodied perception network by extracting the time series feature vectors in the dynamic activity curve and collecting spatial topological data as described above. It is an intuitive presentation of the above construction process at the network structure level, and is not limited to monitoring the behavior analysis of the embodied perception network in different actual application scenarios.

[0083] Specifically, the embodied perception network uses scattered wave body imaging technology to achieve biometric recognition and behavioral intention analysis. Specifically, when loitering behavior is detected (e.g., a stranger lingers for more than five minutes), the system automatically triggers a Level 1 alarm and pushes a warning message. If the person is identified as a registered family member, a multi-factor authentication process (such as preset gestures and voiceprint verification) is initiated to avoid the risk of coercion caused by directly opening the door. This technology complements Wi-Fi multipath scattered wave sensing: the former is used for macro-behavioral pattern recognition, while the latter is used for micro-motion feature extraction.

[0084] S2. Based on the embodied perception network, behavioral feature information in the residential environment is identified; based on the behavioral feature information, a decision architecture in a preset security model is analyzed; and the behavioral matching relationship in the decision architecture is queried, and the behavioral matching relationship is adaptively optimized to obtain an optimized security process.

[0085] The present invention is based on the embodied perception network to identify behavioral feature information in the residential environment. It can model multi-device collaborative perception data through dynamic graph structure, accurately capture cross-regional behavioral associations, and the real-time updated network topology can adapt to environmental changes, improve the robustness and timeliness of behavior recognition in complex scenarios, and provide a more reliable feature basis for security decision-making.

[0086] The behavioral feature information refers to the specific behavioral attributes extracted from the behavior matching tag, including the behavior type (such as intrusion, abnormal movement), occurrence location, timestamp, and associated environmental parameters (such as the sensor nodes involved and the change in signal strength). For example, the behavioral feature information can be expressed as "At 14:30 on XX day of XX month in 2025, an intrusion behavior was detected in the balcony area, associated with Wi-Fi node R1 and infrared sensor S2, the signal strength dropped sharply by 20dB, and the template 'climbing the window' was matched."

[0087] As an embodiment of the present invention, the identification of behavioral feature information in the residential environment based on the embodied perception network includes: obtaining the original behavior signal output by the embodied perception network; segmenting and calibrating the original behavior signal to obtain behavior time series segments; cross-modally fusing the behavior time series segments to obtain behavior semantic vectors; performing behavior matching on the behavior semantic vectors with a preset behavior template library to obtain behavior matching tags; and identifying the behavioral feature information in the behavior matching tags.

[0088] The original behavior signal refers to the unprocessed signal collected and transmitted in real time by each node (such as sensors and smart devices) in the embodied perception network, which contains physical layer data such as voltage fluctuation, phase offset, and frequency change, and directly reflects the original perception information in the environment. For example, the signal phase fluctuation sequence collected by the Wi-Fi receiver (such as 0.1rad, -0.3rad, 0.2rad) and the pulse signal of the human body movement distance output by the infrared sensor (such as 50cm, 30cm, 45cm) are all original behavior signals; the behavior time sequence segment refers to the original behavior signal divided into time windows and marked Continuous data segments with the start and end times of the behavior are annotated to focus on behavioral pattern analysis within a specific time period. For example, the Wi-Fi signal phase fluctuation data lasting 10 seconds is divided into 10 segments per second, where the 3rd to 5th second segments are annotated as "hand rummaging action" and the 7th to 9th second segments are annotated as "footstep movement", which are the behavioral time series segments; the behavioral semantic vector refers to the conversion of behavioral time series segments into high-dimensional vectors with semantic meaning through cross-modal fusion (such as splicing visual image features, Wi-Fi signal features, and sound features) to represent the comprehensive characteristics of the behavior. For example, the fusion camera The extracted "human posture key point coordinate vector", "frequency domain energy distribution vector" of Wi-Fi signal and "voiceprint feature vector" of microphone are used to generate two-dimensional semantic vectors of 0.8 (climbing probability) and 0.2 (normal activity) to describe the semantic attributes of the current behavior; the preset behavior template library refers to a set of pre-stored standard behavior feature vectors, which contains typical patterns of various normal and abnormal behaviors (such as "intrusion", "fall", "pet running", etc.), which are used to match and compare with real-time semantic vectors. For example, the "window intrusion" template in the template library contains the high-frequency disturbance vector of Wi-Fi signal (frequency>10Hz). ), the rapid vertical movement vector of the infrared radar (speed > 1.5m / s), and the opening signal vector of the door and window sensor (state = 1) to form a multi-dimensional feature combination; the behavior matching label refers to the behavior category label and matching confidence output after comparing the behavior semantic vector with the preset template library, which is used to identify the type and credibility of the current behavior. For example, if the matching confidence of the real-time semantic vector and the "elderly fall" template reaches 92%, the output label is "fall (confidence 92%)"; if the matching degree with the "pet running" template is 65% and there is no higher match, the output is "suspected pet activity (confidence 65%)".

[0089] Furthermore, the acquisition of the original behavior signal output by the embodied perception network can be achieved through a multimodal sensor fusion method, such as: using an inertial measurement unit combined with the Topic subscription mechanism of the ROS framework to collect motion data streams, and finally obtaining the original behavior signal containing acceleration and angular velocity information; the segmentation and calibration of the original behavior signal can be achieved through an event boundary detection algorithm, such as: using a sliding window variance method combined with the rolling function of the Pandas library to divide significant behavior intervals, and finally obtaining behavior time series segments with timestamps; the cross-modal fusion of the behavior time series segments can be achieved through an attention mechanism encoding method, such as: based on the Transformer architecture combined with TensorFlow The MultiHeadAttention layer fuses visual inertial features to ultimately obtain a behavioral semantic vector embedded in a spatiotemporal context; the behavioral matching of the behavioral semantic vector with a preset behavioral template library can be achieved through a similarity measurement method, such as using cosine similarity calculation combined with the IndexFlatIP interface of the Faiss library to retrieve the nearest neighbor template, and ultimately obtaining a behavioral matching label that describes the behavior category; the identification of behavioral feature information in the behavioral matching label can be achieved through a feature importance analysis method, such as extracting key dimensions based on the SHAP value interpretation model combined with the feature_importance function of the LightGBM framework, and ultimately obtaining behavioral feature information that reflects behavioral differences.

[0090] Based on this behavioral feature information, the present invention analyzes the decision-making framework within a pre-set security model and queries the behavioral matching relationships within the decision-making framework. This allows for rapid identification of the corresponding protection strategy for the current behavior through structured rules, avoiding the real-time computational delays of complex algorithms and reducing redundant judgment steps. This results in more accurate and efficient security responses, particularly ensuring reliable decision-making in scenarios involving network outages or edge computing.

[0091] Among them, the preset security model refers to a pre-built multi-level security strategy set, which defines a logical framework for behavior recognition, risk assessment and response mechanism based on historical data and security rules, including a behavior classifier, a risk level assessment module and a response strategy library. For example, the model can define "when an intrusion is detected and the doors and windows are open, a level one alarm is triggered and the electronic locks are closed; if an elderly person falls, a level two alarm is triggered and the emergency contact is notified", forming a standardized decision-making process; the decision architecture refers to the hierarchical relationship and interaction logic of each functional module in the preset security model, which is usually organized in the form of a decision tree, a state machine, etc., to clarify the judgment path from behavior feature input to response instruction output. For example, the decision architecture can be designed as a three-layer tree structure of "behavioral feature → risk level judgment (low / medium / high) → response strategy selection (prompt / alarm / active intervention)". For example, when "gas leakage concentration > threshold" is detected, it jumps directly to the execution node of "close valve + open window"; the behavior matching relationship refers to the relationship between the behavior feature information defined in the preset security model and the risk level, response The mapping rules between response strategies are established through the "feature-level-strategy" triplet to ensure that each behavior type corresponds to a unique or combined protection measure. For example, the behavior matching relationship can be expressed as "feature = 'Wi-Fi signal high-frequency disturbance + door and window displacement' → risk level = high → strategy = 'trigger alarm + record video + push positioning'", or "feature = 'pet activity' → risk level = low → strategy = 'ignore'", to achieve differentiated responses. Optionally, the analysis of the decision architecture in the preset security model can be implemented through a model structure analysis method, such as: using computational graph visualization technology combined with the graph interface of the TensorBoard tool to display the network hierarchy, and finally obtaining a decision architecture containing input and output nodes; the query of the behavior matching relationship in the decision architecture can be implemented through an association rule mining algorithm, such as: discovering frequent itemsets based on the Apriori algorithm combined with the association_rules function of the MLxtend library, and finally obtaining a behavior matching relationship that describes the correspondence between conditional behaviors.

[0092] The present invention adaptively optimizes the behavior matching relationship to obtain an optimized security process, continuously iterates rule parameters, dynamically corrects the matching logic of false alarm and missed alarm scenarios, and reduces edge computing resource consumption, ultimately forming a precise security process that is more suitable for dynamic residential scenarios.

[0093] Among them, the optimized security process refers to the security strategy execution process formed by adjusting the original behavior matching rules through analyzing historical data and association mechanisms, which includes more accurate feature matching logic, optimized response sequence and dynamic weight allocation. For example, the optimization process can adjust the infrared trigger threshold of the "pet activity" scene from "moving distance > 50cm" to "moving speed > 1.2m / s and duration > 10s", while lowering the alarm priority of the area and reducing false alarms; adding a joint verification of the "sound call for help" feature to the "elderly fall" scene to improve the recognition accuracy.

[0094] As an embodiment of the present invention, the adaptive optimization of the behavior matching relationship to obtain an optimized security process includes: querying the historical security event set corresponding to the behavior matching relationship; extracting the abnormal behavior characteristics corresponding to each event in the historical security event set; analyzing the behavior association mechanism between the abnormal behavior characteristics; parsing the mechanism linkage data corresponding to the behavior association mechanism; and adaptively optimizing the behavior matching relationship based on the mechanism linkage data to obtain an optimized security process.

[0095] Among them, the historical security event set refers to a collection of historical abnormal event records stored in the system, including data such as the time and location of the event, the triggered behavioral characteristics, the response strategy and the processing results, which are used to analyze the correlation between behavioral patterns and security effects. For example, the historical event set may include records such as "false alarm in the living room on X / XX / 2025 (pet running triggered an infrared alarm)" and "balcony intrusion on X / XX / 2025 (Wi-Fi signal abnormality + doors and windows opened)", with sensor data and response logs attached; the abnormal behavior characteristics refer to key behavioral attributes extracted from historical security events that are different from normal activities, including quantitative indicators such as signal strength mutation value, behavior duration, and spatial location distribution. For example, in the "electrical overload" event, the abnormal characteristics can be manifested as a current value greater than 10A (exceeding the rated value), a duration greater than 5 minutes, and an occurrence location in the kitchen socket area, forming a multidimensional feature group. The behavioral association mechanism refers to the inherent logical connection between abnormal behavioral features, revealing how different features work together to lead to specific security incidents, usually manifested as a causal relationship or a collaborative relationship. For example, the association mechanism of the "gas leak" incident is "gas concentration sensor value rises (cause) → linked window ventilation (response) + triggering an alarm (result)", which may be accompanied by "personnel rapid movement trajectory (collaborative feature)" indicating an attempt to deal with the leak; the mechanism linkage data refers to data that describes the interactive relationship between each link in the behavioral association mechanism, including feature triggering sequence, response delay time, equipment linkage status, etc., which is used to quantify the execution efficiency and reliability of the mechanism. For example, the mechanism linkage data can record "door and window sensor triggering (t=0s) → smart gateway receiving signal within 100ms → electronic lock starting and closing within 300ms → alarm sounding after 500ms", reflecting the full process time chain from detection to response.

[0096] Furthermore, the query of the historical security event set corresponding to the behavior matching relationship can be implemented through a database retrieval algorithm, such as: using the aggregation query function of Elasticsearch in combination with timestamps and behavior tags to perform multi-condition filtering, and finally obtaining the historical security event set; the extraction of abnormal behavior features corresponding to each event in the historical security event set can be implemented through feature engineering technology, such as: using Scikit-learn's feature selection module in combination with the random forest algorithm to extract keyword vectors in the event report, and finally obtaining abnormal behavior features; the analysis of the behavior association mechanism between the abnormal behavior features can be implemented through an association rule mining algorithm, such as: calculating the behavior co-occurrence frequency in Python's MLxtend library based on the Apriori algorithm, and finally obtaining abnormal behavior features; the analysis of the mechanism linkage data corresponding to the behavior association mechanism can be implemented through graph embedding technology, such as: using the Node2Vec algorithm in combination with NetworkX to construct a behavior association graph and generate a low-dimensional vector representation, and finally obtaining the behavior association mechanism; the adaptive optimization of the behavior matching relationship can be implemented through a reinforcement learning framework, such as: using TensorFlow's deep Q network to train the intelligent agent to dynamically adjust the matching strategy according to security feedback, and finally obtaining an optimized security process.

[0097] Specifically, the optimized security process should include a dynamic deployment strategy that is adaptive in time and space, including: away mode: when the embodied perception network detects that all family members have left and the doors and windows are closed, the global deployment is automatically activated, and a multi-level linkage response (local alarm + property notification + mobile phone push) is triggered for illegal intrusion (such as breaking windows, technical unlocking); night mode: maintain selective deployment during sleeping hours, monitor unmanned bedrooms and public passages, and automatically release protection along the path when legal member activities are detected (such as going to the toilet at night), and re-deploy after the personnel return; duress handling: when a door opening command is recognized but abnormal biometric characteristics are detected (such as abnormal heart rate, duress gestures), a hidden alarm is executed and emergency recording is started.

[0098] S3. Based on the optimized security process, a security response path corresponding to the residential environment is generated; based on the security response path, the spatiotemporal distribution state of human activities in the residential environment is determined; and based on the spatiotemporal distribution state, the security efficiency value corresponding to the residential environment is calculated.

[0099] Based on the optimized security process, the present invention generates a security response path corresponding to the residential environment. It can automatically plan the shortest execution link according to the dynamic rule priority, and verify the path reliability through a rehearsal mechanism, thereby ensuring that responses are not missed in multiple scenarios and forming a precise protection system with a full-process closed loop.

[0100] Among them, the security response path refers to a full-process operation sequence from risk detection to hidden danger elimination, which is dynamically generated based on the security demand level and real-time signals. It includes the multi-device linkage sequence, parameter configuration and status verification logic to ensure the timeliness and effectiveness of the response action. For example, the response path for "kitchen fire" is "smoke sensor trigger → close the gas valve within 0.5 seconds → start the kitchen fire extinguishing device within 1 second → send an alarm to the owner + fire department within 2 seconds → cut off the kitchen power within 5 seconds". Each link is executed accurately according to priority and time sequence.

[0101] As an embodiment of the present invention, generating a security response path corresponding to the residential environment based on the optimized security process includes: parsing the security requirement level corresponding to the optimized security process; based on the security requirement level, collecting real-time security signals in the residential environment; performing channel mapping on the real-time security signals to obtain signal response channels; collecting key response points in the signal response channels; and generating a security response path corresponding to the residential environment based on the key response points.

[0102] Among them, the security demand level refers to the degree of urgency of protection divided according to behavioral feature information and risk level, which is usually divided into three levels: low (prompt), medium (alarm), and high (emergency intervention), which is used to determine the priority of the response strategy and the scale of resource investment. For example, the detection of "pet activity" corresponds to a low-level demand, which only requires recording data; "gas leakage" corresponds to a high-level demand, which requires immediate triggering of multiple operations such as valve closing, ventilation, and alarm; the real-time security signal refers to the status data fed back in real time by sensors or execution equipment during the execution of optimized security processes, including risk parameters (such as gas concentration, current value), equipment operating status (such as valve switch, camera viewing angle), etc. For example, the concentration value uploaded by the smart gas meter in real time (such as 500ppm), the locking status of the electronic lock (locked / unlocked), and the real-time picture frame sent back by the camera are all real-time security signals; the The signal response channel refers to the logical path from the acquisition end to the execution end during the transmission and processing of security signals, corresponding to the hardware device connection relationship (such as sensor → gateway → actuator) or software data flow (such as cloud → edge node → terminal). For example, the signal channel of the "door and window intrusion" scenario is "door and window sensor → Wi-Fi gateway → local edge computing node → alarm + electronic lock", forming a closed-loop path of physical connection and data transmission; the key response point refers to the node or parameter in the signal response channel that plays a decisive role in the security effect, including the key algorithm of signal processing (such as behavior recognition model), the trigger threshold of the execution device (such as current overload protection value), and the time node of multi-device collaboration (such as the synchronization of alarm and video recording). For example, in the "elderly fall" response channel, the key response points include the acceleration mutation detection threshold of the infrared radar (>3m / s 2), the communication timeout reconnection mechanism of the automatic dialing system (retry 3 times within 5 seconds), and the accuracy requirements of the positioning information (error <10 meters).

[0103] Furthermore, the analysis of the security requirement level corresponding to the optimized security process can be achieved through the hierarchical analysis method, such as: using the AHP algorithm combined with ExpertChoice software to calculate the weight index of each dimension, and finally obtaining a quantitative security requirement level; the collection of real-time security signals in the residential environment can be achieved through a multi-source sensor fusion method, such as: using WIFI The mesh networking technology is combined with the GPIO interface of the RISC-V chip to synchronously collect infrared / microwave signals, and finally obtain a time-aligned real-time security signal; the channel mapping of the real-time security signal can be achieved through a signal decomposition method, such as: based on the ICA algorithm combined with the FastICA module of Scikit-learn to separate independent components, and finally obtain the signal response channel corresponding to the physical space; the key response points in the signal response channel can be collected through a peak detection method, such as: using the FindPeaks function combined with the signal module of SciPy to locate the abnormal extreme points, and finally obtain the key response points representing the risk location; the generation of the security response path corresponding to the residential environment can be achieved through a path planning algorithm, such as: based on the A* algorithm combined with the pathfinding interface of the PyGame library to calculate the optimal patrol route, and finally obtain the security response path covering the key nodes.

[0104] Based on the security response path, the present invention determines the spatiotemporal distribution of human activities in the residential environment. With the help of multimodal sensor data fusion, it can construct a dynamic map of human movement trajectories and regional residence time in real time, accurately identify high-risk time periods and spatial blind spots, thereby improving the security system's real-time adaptability to complex scenarios.

[0105] Among them, the spatiotemporal distribution state refers to the dynamic distribution characteristics of people or objects in a residential environment in the time dimension (such as the time when the behavior occurs and the duration) and the spatial dimension (such as position coordinates, movement trajectory, and regional stay). A structured description is formed by fusing multi-source sensor data to reflect the spatiotemporal correlation and regularity of the activity. For example, "From 15:00 to 15:10 on XX day of XX month in 2025, the elderly moved slowly at a speed of 0.3 m / s in the bedroom (coordinates X=3, Y=5), stayed for 8 minutes, and then returned to the living room." The spatiotemporal trajectory of this period constructed by combining infrared radar and Wi-Fi signal data is a typical spatiotemporal distribution state description. Optionally, the spatiotemporal distribution state of human activities in the residential environment can be determined by a spatiotemporal trajectory clustering method, such as: using the DBSCAN algorithm combined with the cluster module of the Scikit-learn library to analyze UWB positioning data, and finally obtaining the spatiotemporal distribution state including activity hotspot areas.

[0106] Furthermore, the present invention calculates the security efficiency value corresponding to the residential environment based on the spatiotemporal distribution state, and can quantify the evaluation system's monitoring coverage of human activities, making the allocation of security resources more reasonable, thereby promoting the security system from passive response to active performance optimization and upgrading, and ultimately achieving quantifiable and iterative improvement of the protection effect.

[0107] Among them, the security efficiency value refers to a quantitative indicator that comprehensively measures the performance of the residential security system in terms of spatial coverage, event response and monitoring balance, reflecting the overall effectiveness level of the system in protecting residential security. The higher the value, the more efficient the security system is.

[0108] As an embodiment of the present invention, calculating the security efficiency value corresponding to the residential environment according to the spatiotemporal distribution state includes:

[0109] The security efficiency value corresponding to the residential environment is calculated using the following formula:

[0110]

[0111] Where EZ represents the security efficiency value corresponding to the residential environment, n represents the number of spatial areas divided by the residential area, i represents the number index corresponding to the spatial area, C i represents the actual coverage efficiency of the i-th spatial region, T i represents the total length of time a person stays in the i-th spatial area during the statistical period, m represents the total number of security events during the statistical period, j represents the number index corresponding to the security event, and R k represents the response efficiency index corresponding to the j-th security event, β represents the event priority weight, and D i represents the real-time monitoring density of the i-th spatial region, and Ω represents the time dimension weight corresponding to the spatiotemporal distribution state.

[0112] In detail, the spatial area refers to the different parts of a house divided according to factors such as function and layout, such as living room, bedroom, kitchen, etc., which is convenient for targeted evaluation and management of security conditions and is the basic unit for calculating security efficiency; the actual coverage efficiency refers to the proportion of the area actually effectively monitored in a single spatial area to the total area of ​​the area after deducting the overlapping coverage between sensors, reflecting the degree of effective monitoring of the area by the sensor; the statistical period refers to the specific time range set for calculating the security efficiency value, which is used to collect and count relevant data such as personnel activities and security events, and can be set to one day, one week or one month according to needs; the total stay time refers to the cumulative length of time a person stays in a specific spatial area during the statistical period, reflecting the frequency of personnel activities in the area, and is an important basis for evaluating the allocation of security resources; the security event refers to various situations that may affect safety in a residential environment, including intrusion, fire, gas, etc. Abnormal conditions such as leaks and elderly falls are objects that the security system needs to respond to and handle; the response efficiency index refers to an indicator used to measure the speed of the security system's response to a single security event from detection to response, usually calculated by comparing the response time with the benchmark time. The closer the value is to 1, the faster the response; the event priority weight refers to the weight coefficient assigned according to the severity and urgency of the security event, such as high weights for serious events such as intrusions and fires, and low weights for ordinary abnormal events, so as to highlight the impact of important events on security efficiency; the real-time monitoring density refers to the frequency of monitoring a specific spatial area at a certain moment or in a short period of time, that is, the number of times the sensor collects data per unit time, reflecting the frequency of monitoring of the area; the time dimension weight refers to the weight coefficient assigned to each time period within the statistical period, taking into account the differences in residential security needs in different time periods. For example, a higher weight can be assigned at night, reflecting the security system's adaptive adjustment to time factors.

[0113] S4. Based on the security efficiency value, determine the security optimization direction corresponding to the residential environment; based on the security optimization direction, generate a dynamic deployment unit corresponding to the residential environment; extract key execution parameters in the dynamic deployment unit; based on the key execution parameters, calculate the misjudgment attenuation rate corresponding to the residential environment.

[0114] Based on the security efficiency value, the present invention determines the security optimization direction corresponding to the residential environment, and can accurately locate the weak links of the system, such as low coverage efficiency or long response time in a certain area. Based on this, resources can be reasonably allocated, sensors can be added or equipment can be upgraded. Security strategies can be dynamically adjusted according to the weights of different time periods, thereby achieving efficient resource utilization and improving the overall security level.

[0115] Among them, the security optimization direction refers to the improvement direction and measures determined for the security optimization area to improve the security level, including: increasing the number of sensors in the optimization area, upgrading equipment performance, adjusting response strategies, etc. For example, for areas where intrusions frequently occur, it is determined to install high-definition cameras, increase infrared sensing equipment, and improve the alarm linkage mechanism. This is the security optimization direction.

[0116] As an embodiment of the present invention, determining the security optimization direction corresponding to the residential environment based on the security efficiency value includes: parsing the weak link parameters corresponding to the security efficiency value; locating the inefficient coverage area of ​​the security equipment in the residential environment according to the weak link parameters; extracting security log data corresponding to the inefficient coverage area; locating the security optimization area corresponding to the residential environment based on the security log data and historical intrusion events; and determining the security optimization direction corresponding to the security optimization area.

[0117] Among them, the weak link parameters refer to quantitative indicators that reflect the deficiencies of the security system in terms of spatial coverage, response speed, monitoring density, etc., including: low actual coverage efficiency, substandard response efficiency index, large monitoring density variance, etc. For example, the sensor overlap rate in a certain area is too high, resulting in an actual coverage efficiency of only 30%, which is far below the normal level. This is the weak link parameter; the inefficient coverage area refers to the area where the security equipment has major defects in monitoring the area, including: low actual coverage efficiency, insufficient real-time monitoring density, etc., which cannot effectively ensure the safety of the area. For example, due to signal problems in the basement, the actual coverage efficiency of the sensor is only 20%, and the monitoring density is also far lower than other areas. This is the inefficient coverage area; the security log data refers to various types of information recorded during the operation of the security system, including equipment status (switches, faults, etc.), event triggering (intrusion, abnormal behavior, etc.), response, etc. The system records the operation (alarm activation, personnel dispatch, etc.) and timestamp, for example, at 2 a.m., the kitchen smoke sensor alarms, and the system records the alarm time, sensor status, subsequent alarm response status and other information. These are security log data; the historical intrusion events refer to the records of illegal intrusion behaviors that occurred in the residential environment in the past period of time, including the time, location, method of intrusion, and consequences of the intrusion. For example, in August last year, a residential building in the community was burglarized at night. The specific time, location and method of entry of the thief are historical intrusion events; the security optimization area refers to a specific area that needs security improvement based on comprehensive factors such as weak link parameters, inefficient coverage areas and historical intrusion events. For example, there have been multiple window prying intrusions on the first floor of a residential building, and the monitoring density in this area is low. This first-floor area is the security optimization area.

[0118] Furthermore, the analysis of the weak link parameters corresponding to the security efficiency value can be achieved through a sensitivity analysis method, such as: using the Sobol index combined with the parameter sampling technology of the SALib library to calculate the contribution of each dimension, and finally obtaining the key weak link parameters that affect the system performance; the positioning of the inefficient coverage area of ​​the security equipment in the residential environment can be achieved through a spatial heat map analysis method, such as: using KDE kernel density estimation combined with the contourf function of Matplotlib to draw the distribution of device response intensity, and finally obtaining the inefficient coverage area for identifying blind spots; the extraction of the security log data corresponding to the inefficient coverage area can be achieved through a spatiotemporal query method For example, based on the spatial join operation of GeoPandas, event records in a specific area are filtered to finally obtain security log data with geographical tags; the positioning of the security optimization area corresponding to the residential environment can be achieved through a multi-objective optimization method, such as using the NSGA-II algorithm combined with the DEAP framework to solve the Pareto front of coverage and response time, and finally obtaining the security optimization area that needs priority improvement; the determination of the security optimization direction corresponding to the security optimization area can be achieved through a decision tree classification method, such as using the CART algorithm combined with the Scikit-learn tool to analyze historical optimization cases, and finally obtaining the security optimization direction based on regional characteristics.

[0119] Based on the security optimization direction, the present invention generates a dynamic deployment unit corresponding to the residential environment and extracts key execution parameters in the dynamic deployment unit. It can flexibly configure security resources according to the actual security needs of the residence, and can clearly define key execution parameters to accurately control equipment operation, improve response speed and accuracy, and facilitate system management and maintenance, ensuring efficient and stable operation of the security system.

[0120] Among them, the dynamic deployment unit refers to a module with targeted protection function formed by flexibly combining various security equipment (such as cameras, sensors, alarms, etc.) based on the direction of residential security optimization, combined with real-time personnel activities, risk conditions and other factors. It can dynamically adjust the deployment strategy as the environment changes. For example, when it is detected that there is no one in the living room at night, the motion detection sensor and infrared camera are automatically turned on to form a dynamic deployment unit for the living room, thereby strengthening the security protection of the area during this period; the key execution parameters refer to the important parameters that determine whether the dynamic deployment unit can effectively perform security tasks, covering the triggering conditions of the equipment (such as sensor sensing threshold), operating status (such as camera resolution, frame rate), response time (such as alarm delay time), etc. For example, in the kitchen smoke alarm In the deployment unit, parameters such as the alarm concentration threshold of the smoke sensor (such as 5% obs / ft) and the alarm activation delay time (0.5 seconds) directly affect the accuracy and timeliness of the fire warning. These are key execution parameters. Optionally, the generation of the dynamic deployment unit corresponding to the residential environment can be achieved through a multi-agent collaborative method, such as: using the MADDPG algorithm combined with the Ray framework to train distributed security agents, and finally obtaining an autonomous and collaborative dynamic deployment unit; the extraction of key execution parameters in the dynamic deployment unit can be achieved through a feature importance analysis method, such as: calculating parameter weights based on the feature_importance attribute of the XGBoost model, and finally obtaining key execution parameters that affect the deployment effect.

[0121] In detail, the working logic of the dynamic deployment unit should reflect the anthropomorphic decision-making of environmental perception, which includes: in the living room area, only motion detection is enabled during the day, and thermal imaging monitoring is superimposed at night; the presence perception priority principle is adopted for bedroom deployment, and door and window monitoring is activated when the infrared + pressure sensor confirms that there is no one, so as to avoid affecting the normal living of members; channel protection implements dynamic following and disarming, and tracks the movement path of members through UWB positioning, and adjusts the protection range in real time.

[0122] Furthermore, the present invention calculates the false positive attenuation rate corresponding to the residential environment based on the key execution parameters, and can quantitatively evaluate the false alarms or missed alarms caused by unreasonable parameter settings of the security system. It can optimize key parameters such as the equipment trigger threshold and response logic in a targeted manner, reduce problems such as pet activity triggering alarms and misjudgments caused by normal operation of electrical appliances, and effectively improve the accuracy of the security system.

[0123] Among them, the false positive attenuation rate refers to an indicator that measures the degree to which false alarms are reduced after the key execution parameters of the residential security system are adjusted. The closer the value is to 1, the greater the reduction in the false alarm rate due to the parameter adjustment, the higher the accuracy of the security system, and the more effectively it can avoid interference caused by false alarms.

[0124] As an embodiment of the present invention, calculating the false positive attenuation rate corresponding to the residential environment based on the key execution parameters includes:

[0125] The false positive attenuation rate corresponding to the residential environment is calculated using the following formula:

[0126]

[0127] Among them, R fad represents the false positive decay rate corresponding to the residential environment, l represents the total number of false positive types, k represents the number index corresponding to the false positive type, α k represents the adjustment coefficient of the key execution parameter corresponding to the kth false alarm type, γ k Indicates the sensitivity coefficient corresponding to the kth false alarm type, ERR k represents the false alarm rate of the kth false alarm type before the key execution parameters are adjusted, W k Indicates the weight coefficient corresponding to the k-th false alarm type.

[0128] Specifically, the false alarm type refers to the classification of different situations in which the security system generates false alarms, such as false alarms caused by pet activity, electrical interference, and environmental factors (such as strong wind and dust). Distinguishing different false alarm types helps to adjust key execution parameters in a targeted manner and reduce false alarms. The adjustment coefficient refers to the proportional or amplitude correlation coefficient of the adjustment of key execution parameters for different false alarm types, which reflects the extent to which the key execution parameters should be adjusted for a specific false alarm type to reduce the possibility of this type of false alarm. The sensitivity coefficient refers to the sensitivity of a certain false alarm type to changes in key execution parameters. The higher the sensitivity coefficient, the greater the impact of key execution parameter adjustments on this false alarm type, and the more attention should be paid to adjusting the parameters. The false alarm rate refers to the frequency of false alarms of a specific false alarm type before the key execution parameters are adjusted. It is usually expressed as the ratio of the number of false alarms to the total number of alarms within a certain period of time and is the basic data for evaluating false alarm situations. The weight coefficient refers to the importance coefficient assigned to each false alarm type based on factors such as the degree of impact of different false alarm types on the normal operation of the security system and the frequency of occurrence. False alarm types with greater impact and higher frequency have relatively higher weight coefficients.

[0129] S5. Based on the misjudgment attenuation rate, determine the security status corresponding to the residential environment, analyze the security dimension of the residential environment corresponding to the security status, and construct an intelligent security solution corresponding to the residential environment based on the security dimension.

[0130] The present invention determines the security status corresponding to the residential environment based on the false positive attenuation rate, can intuitively quantify the improvement of the security system's false alarms, accurately evaluate the reliability of the security system, and promptly discover security loopholes. For example, if a certain area has a poor security status due to frequent false positives, it is convenient to make targeted adjustments and optimizations to improve security effectiveness.

[0131] Among them, the security status refers to the overall judgment of the current safety protection status of the residence based on the comprehensive false judgment attenuation rate, the execution status of the early warning signal command, etc., which includes safety, warning, danger and other states. For example, if there is no abnormal signal and the false judgment attenuation rate is high, it is judged to be a safe state; if a warning signal command is issued but the actual danger is not confirmed, it is judged to be a warning state; if there is an actual threat such as intrusion or fire, it is judged to be a dangerous state.

[0132] As an embodiment of the present invention, determining the security status corresponding to the residential environment based on the false judgment attenuation rate includes: setting a dynamic monitoring threshold corresponding to the residential environment according to the false judgment attenuation rate; configuring a signal sampling frequency corresponding to the dynamic monitoring threshold; screening abnormal signal segments in the residential environment based on the signal sampling frequency; generating an early warning signal instruction corresponding to the abnormal signal segment; and determining the security status corresponding to the residential environment based on the early warning signal instruction.

[0133] Among them, the dynamic monitoring threshold refers to the flexibly variable trigger critical value set for various sensors (such as smoke, infrared, etc.) of the security system based on factors such as the false judgment attenuation rate. It can be adjusted as the environment or security needs change to balance false alarms and missed alarms. For example, in areas where pets are active, the motion detection threshold of the infrared sensor can be appropriately increased to reduce false alarms caused by pets; the signal sampling frequency refers to the number of times the security system collects sensor signals per unit time, which determines the frequency of obtaining environmental information. The higher the sampling frequency, the more timely the perception of environmental changes. For example, the camera collects 25 frames per second, and the 25 frames / second here is the signal sampling frequency; the abnormal signal segment is It refers to the signal part that is screened out from the signal collected by the sensor, which does not conform to the normal state characteristics and may indicate safety hazards. For example, under normal circumstances, the current of an electrical appliance is stable. If the current sensor collects a current signal that fluctuates greatly in a short period of time, this part of the fluctuating signal is an abnormal signal fragment, which may indicate an electrical failure or abnormal power consumption; the early warning signal instruction refers to the command generated according to preset rules when the system detects an abnormal signal fragment for triggering an early warning action. The instruction may include triggering an alarm to sound, sending a notification to the owner's mobile phone, turning on a video recording device, etc. For example, after the smoke sensor detects an abnormal smoke signal fragment, the system issues an instruction to let the alarm sound a high-decibel alarm and push a fire warning text message to the owner.

[0134] Furthermore, setting the dynamic monitoring threshold corresponding to the residential environment can be achieved through an adaptive threshold algorithm, such as using a sliding window standard deviation method combined with the rolling function of the Pandas library to calculate a dynamic baseline, ultimately obtaining a dynamic monitoring threshold that adjusts with environmental changes; configuring the signal sampling frequency corresponding to the dynamic monitoring threshold can be achieved through the Nyquist sampling theorem, such as using Scipy's signal.resample_poly function to perform anti-aliasing resampling based on the highest frequency component of the signal, ultimately obtaining a signal sampling frequency that meets reconstruction requirements; screening abnormal signal segments in the residential environment can be achieved through an isolation forest algorithm, such as using Scikit-learn's IsolationForest model to detect signal intervals that deviate from normal patterns, ultimately obtaining abnormal signal segments marked with abnormal timestamps; generating warning signal instructions corresponding to the abnormal signal segments can be achieved through a rule engine method, such as matching abnormal features based on the when-then rule template of the Drools framework, ultimately obtaining a warning signal instruction containing a risk level description; determining the security status corresponding to the residential environment can be achieved through a state machine modeling method, such as using a finite state automaton combined with the Transitions library to construct state transition logic, ultimately obtaining a quantitatively assessed security status.

[0135] By analyzing the security dimensions of the residential environment under the security status, the present invention can comprehensively assess the residential security situation and identify weak links. In the early warning state, the present invention can analyze each security dimension to accurately locate potential risk sources, timely detect and fill gaps, and provide a basis for subsequent security strategy adjustments, reasonably allocate resources, and improve the overall security protection level.

[0136] Among them, the security dimension refers to the perspective of measuring and evaluating residential safety from different aspects, covering physical safety (such as the strength of doors and windows, and the height of walls), equipment safety (sensor accuracy, alarm reliability), information security (monitoring data storage security, transmission encryption), environmental safety (fire, gas leakage and other risks), etc. For example, when evaluating the fire safety dimension of a residence, factors such as fire protection facilities, electrical line safety, and flammable material storage will be considered to judge the fire resistance of the residence under this dimension. Optionally, the analysis of the security dimension of the residential environment corresponding to the security state can be achieved through a multidimensional evaluation method, such as: using the hierarchical analysis method combined with Python's PyDecision library to construct a judgment matrix to calculate the weights, and finally obtaining a security dimension including multiple indicators such as intrusion risk, equipment reliability, and emergency response speed.

[0137] Furthermore, based on the security dimensions, the present invention constructs an intelligent security solution corresponding to the residential environment. It can design a layered protection system for multi-dimensional risks such as physical, equipment, information, and environment, to achieve an upgrade from passive defense to active prediction, and integrate AI algorithms to link protection modules of various dimensions to comprehensively enhance the systematic and intelligent level of residential security protection.

[0138] Among them, the smart security solution refers to a full-process intelligent protection system designed by comprehensively using technologies such as the Internet of Things, artificial intelligence, big data analysis and automated control to meet the multi-dimensional security needs of residences in terms of physical security (such as intrusion prevention), environmental safety (such as fire and leakage monitoring), equipment safety (such as sensor reliability), and information security (such as data encryption). This solution collects data in real time through various smart devices (sensors, cameras, actuators, etc.), and automatically triggers early warning, linkage disposal and other operations after algorithm analysis. It has the capabilities of dynamic self-adaptation, risk prediction, and cross-device collaboration, and can achieve an upgrade from passive response to active defense. For example, smart security The prevention solution integrates "millimeter wave radar human presence detection + smart electronic fence + multispectral fire camera + AI abnormal behavior analysis platform". When the radar detects abnormal movement trajectory in the walled area, the electronic fence immediately issues an audible and visual warning. At the same time, the camera automatically tracks the target and analyzes the behavior pattern (distinguishing between people / animals). After confirming the intrusion, the door lock is locked and the alarm system pushes the location information to the property. The whole process does not require human intervention, achieving precise protection of perimeter safety and indoor anomalies. Optionally, the construction of the intelligent security solution corresponding to the residential environment can be achieved through a multimodal fusion decision-making method, such as: using a federated learning framework combined with TensorFlow Federated to integrate multi-source sensor data such as vision, hearing, and infrared, and finally obtaining an intelligent security solution with environmental adaptability.

[0139] Specifically, the intelligent security solution should have triple cognitive capabilities, including: identity recognition: distinguishing family members from strangers through the fusion of multiple biometric features (gait, face, voiceprint); intention recognition: judging the risk level of behavior based on the length of stay, movement trajectory, and abnormal actions (such as lock picking posture); situational recognition: autonomously switching protection strategies according to time, space, and personnel distribution status to form a human-like decision-making closed loop.

[0140] Compared with the problems described in the background technology, the present invention obtains real-time perception data corresponding to the residential environment and collects behavioral feature signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves. It can realize camera-free monitoring under the premise of protecting privacy and accurately capture subtle abnormal behaviors such as climbing and rummaging; multi-dimensional data fusion can effectively distinguish scenes such as pet activities and elderly falls, and significantly reduce the false alarm rate of the security system. The present invention is based on the embodied perception network to identify behavioral feature information in the residential environment. It can use dynamic graph structure modeling to collaboratively perceive data of multiple devices and accurately capture cross-regional behavioral associations. The real-time updated network topology can adapt to environmental changes, improve the robustness and timeliness of behavior recognition in complex scenarios, and provide more reliable feature basis for security decision-making. Furthermore, based on the optimized security process, the present invention generates a security response path corresponding to the residential environment. The shortest execution link can be automatically planned based on the dynamic rule priority, and the path reliability can be verified through a rehearsal mechanism to ensure that responses are not missed in multiple scenarios, forming a precise protection system with a full-process closed loop. Furthermore, the present invention determines the security optimization direction corresponding to the residential environment based on the security efficiency value, and can accurately locate the weak links of the system, such as low coverage efficiency or long response time in a certain area. Based on this, resources can be reasonably allocated, sensors can be added or equipment can be upgraded. The security strategy can be dynamically adjusted according to the weights of different time periods to achieve efficient resource utilization and improve the overall security level. Finally, the present invention determines the security status corresponding to the residential environment based on the false positive attenuation rate, and can intuitively quantify the improvement of the false alarm of the security system, accurately evaluate the reliability of the security system, and timely discover security loopholes. For example, if the security status of a certain area is poor due to frequent false positives, it is convenient to make targeted adjustments and optimizations to improve security efficiency. Therefore, the embodiment of the present invention provides a method and system for embodied smart home security that can improve the intelligence level of the home security system.

[0141] Example 2:

[0142] like Figure 3 FIG. 1 is a functional module diagram of an embodied smart home security system according to the present invention.

[0143] The embodied smart home security system 200 described in the present invention can be installed in an electronic device. Depending on the functionality implemented, the embodied smart home security system may include a network construction module 201, a process optimization module 202, an efficiency calculation module 203, a decay rate calculation module 204, and a solution construction module 205. A module, also referred to as a unit, is a series of computer program segments that can be executed by an electronic device processor and perform a fixed function. These are stored in the electronic device's memory.

[0144] In the embodiment of the present invention, the functions of each module / unit are as follows:

[0145] The network construction module 201 is configured to obtain real-time perception data corresponding to a residential environment, collect behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, analyze dynamic activity curves corresponding to the behavioral characteristic signals, and construct an embodied perception network corresponding to the residential environment based on the dynamic activity curves;

[0146] The process optimization module 202 is configured to identify behavioral feature information in the residential environment based on the embodied perception network, analyze the decision architecture in the preset security model based on the behavioral feature information, query the behavioral matching relationship in the decision architecture, and adaptively optimize the behavioral matching relationship to obtain an optimized security process;

[0147] The efficiency value calculation module 203 is configured to generate a security response path corresponding to the residential environment based on the optimized security process, determine the spatiotemporal distribution of human activities in the residential environment based on the security response path, and calculate the security efficiency value corresponding to the residential environment based on the spatiotemporal distribution;

[0148] The attenuation rate calculation module 204 is configured to determine a security optimization direction corresponding to the residential environment based on the security efficiency value, generate a dynamic deployment unit corresponding to the residential environment based on the security optimization direction, extract key execution parameters from the dynamic deployment unit, and calculate a false positive attenuation rate corresponding to the residential environment based on the key execution parameters;

[0149] The solution construction module 205 is used to determine the security status corresponding to the residential environment based on the misjudgment attenuation rate, analyze the security dimensions of the residential environment corresponding to the security status, and construct an intelligent security solution corresponding to the residential environment based on the security dimensions.

[0150] In detail, the modules in the embodiment of the present invention based on the embodied smart home security system 200 are used in the same manner as above. Figure 1 The technical means described in the above text are similar to the embodied smart home security method and can produce the same technical effects, so I will not go into details here.

[0151] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A security method based on embodied smart home, characterized in that: The method comprises: Acquiring real-time perception data corresponding to a residential environment, collecting behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, analyzing dynamic activity curves corresponding to the behavioral characteristic signals, and constructing an embodied perception network corresponding to the residential environment based on the dynamic activity curves; Based on the embodied perception network, behavioral feature information in the residential environment is identified; based on the behavioral feature information, a decision architecture in a preset security model is analyzed; behavioral matching relationships in the decision architecture are queried; the behavioral matching relationships are adaptively optimized to obtain an optimized security process; generating a security response path corresponding to the residential environment based on the optimized security process, determining a spatiotemporal distribution of human activities in the residential environment based on the security response path, and calculating a security efficiency value corresponding to the residential environment based on the spatiotemporal distribution; Determining a security optimization direction corresponding to the residential environment based on the security efficiency value, generating a dynamic deployment unit corresponding to the residential environment based on the security optimization direction, extracting key execution parameters from the dynamic deployment unit, and calculating a false positive attenuation rate corresponding to the residential environment based on the key execution parameters; Based on the misjudgment attenuation rate, the security status corresponding to the residential environment is determined, the security dimension of the residential environment corresponding to the security status is analyzed, and based on the security dimension, an intelligent security solution corresponding to the residential environment is constructed.

2. The method for embodied smart home security according to claim 1, wherein: The step of constructing an embodied perception network corresponding to the residential environment according to the dynamic activity curve includes: Extracting a time series feature vector from the dynamic activity curve; Collecting spatial topological data in the residential environment and generating a spatial perception matrix corresponding to the spatial topological data; identifying a responsive topological link in the spatial perception matrix; Extracting a link response node in the response topology link; An embodied perception network corresponding to the residential environment is constructed based on the link response node.

3. The method for embodied smart home security according to claim 1, wherein: The identifying behavioral characteristic information in the residential environment based on the embodied perception network includes: Obtaining an original behavior signal output by the embodied perception network; Segmenting and calibrating the original behavior signal to obtain behavior time sequence segments; Performing cross-modal fusion on the behavior time sequence segments to obtain a behavior semantic vector; Performing behavior matching on the behavior semantic vector and a preset behavior template library to obtain a behavior matching label; Identify the behavior feature information in the behavior matching tag.

4. The method for embodied smart home security according to claim 1, wherein: Adaptively optimizing the behavior matching relationship to obtain an optimized security process includes: Query the historical security event set corresponding to the behavior matching relationship; Extracting abnormal behavior features corresponding to each event in the historical security event set; Analyzing the behavioral correlation mechanism between the abnormal behavioral characteristics; Analyzing the mechanism linkage data corresponding to the behavior association mechanism; Based on the mechanism linkage data, the behavior matching relationship is adaptively optimized to obtain an optimized security process.

5. The method for embodied smart home security according to claim 1, wherein: Generating a security response path corresponding to the residential environment based on the optimized security process includes: Analyze the security requirement level corresponding to the optimized security process; Based on the security requirement level, collecting real-time security signals in the residential environment; Performing channel mapping on the real-time security signal to obtain a signal response channel; collecting key response points in the signal response channel; Based on the key response points, a security response path corresponding to the residential environment is generated.

6. The method for embodied smart home security according to claim 1, wherein: Calculating the security efficiency value corresponding to the residential environment according to the spatiotemporal distribution state includes: The security efficiency value corresponding to the residential environment is calculated using the following formula: Where EZ represents the security efficiency value corresponding to the residential environment, n represents the number of spatial areas divided by the residential area, i represents the number index corresponding to the spatial area, C i represents the actual coverage efficiency of the i-th spatial region, T i represents the total length of time people stay in the i-th spatial area during the statistical period, m represents the total number of security events during the statistical period, j represents the number index corresponding to the security event, R j represents the response efficiency index corresponding to the j-th security event, β represents the event priority weight, and D i represents the real-time monitoring density of the i-th spatial region, and Ω represents the time dimension weight corresponding to the spatiotemporal distribution state.

7. The method for embodied smart home security according to claim 1, wherein: The determining, based on the security efficiency value, a security optimization direction corresponding to the residential environment includes: Analyze the weak link parameters corresponding to the security efficiency value; Locating inefficient coverage areas of security equipment in the residential environment based on the weak link parameters; Extracting security log data corresponding to the inefficient coverage area; Locating a security optimization area corresponding to the residential environment based on the security log data and historical intrusion events; Determine the security optimization direction corresponding to the security optimization area.

8. The method for embodied smart home security according to claim 1, wherein: The calculating, based on the key execution parameters, the misjudgment attenuation rate corresponding to the residential environment includes: The false positive attenuation rate corresponding to the residential environment is calculated using the following formula: Among them, R fad represents the false positive decay rate corresponding to the residential environment, l represents the total number of false positive types, k represents the number index corresponding to the false positive type, α k represents the adjustment coefficient of the key execution parameter corresponding to the kth false alarm type, γ k Indicates the sensitivity coefficient corresponding to the kth false alarm type, ERR k represents the false alarm rate of the kth false alarm type before the key execution parameters are adjusted, W k Indicates the weight coefficient corresponding to the k-th false alarm type.

9. The method for embodied smart home security according to claim 1, wherein: The determining, based on the false positive attenuation rate, a security status corresponding to the residential environment includes: Setting a dynamic monitoring threshold corresponding to the residential environment according to the false positive attenuation rate; Configuring the signal sampling frequency corresponding to the dynamic monitoring threshold; screening abnormal signal segments in the residential environment based on the signal sampling frequency; generating a warning signal instruction corresponding to the abnormal signal segment; Based on the early warning signal instruction, the security status corresponding to the residential environment is determined.

10. A smart home security system based on embodiment, characterized in that: The system comprises: a network construction module for acquiring real-time perception data corresponding to a residential environment, collecting behavioral characteristic signals corresponding to the real-time perception data through wireless Wi-Fi multipath scattered waves, analyzing dynamic activity curves corresponding to the behavioral characteristic signals, and constructing an embodied perception network corresponding to the residential environment based on the dynamic activity curves; a process optimization module for identifying behavioral characteristic information in the residential environment based on the embodied perception network, analyzing a decision architecture in a preset security model based on the behavioral characteristic information, querying behavioral matching relationships in the decision architecture, and adaptively optimizing the behavioral matching relationships to obtain an optimized security process; an efficiency value calculation module, configured to generate a security response path corresponding to the residential environment based on the optimized security process, determine a spatiotemporal distribution of human activities in the residential environment based on the security response path, and calculate a security efficiency value corresponding to the residential environment based on the spatiotemporal distribution; a decay rate calculation module, configured to determine a security optimization direction corresponding to the residential environment based on the security efficiency value, generate a dynamic deployment unit corresponding to the residential environment based on the security optimization direction, extract key execution parameters from the dynamic deployment unit, and calculate a false positive decay rate corresponding to the residential environment based on the key execution parameters; A solution construction module is used to determine the security status corresponding to the residential environment based on the misjudgment attenuation rate, analyze the security dimensions of the residential environment corresponding to the security status, and construct an intelligent security solution corresponding to the residential environment based on the security dimensions.

Images