Smart home control system based on Internet of Things

Abstract

The invention discloses a smart home control system based on Internet of Things, which belongs to the field of home control systems and comprises an Internet of Things sensor module, an edge computing gateway, a cloud analysis platform and an execution control module. The Internet of Things sensor module is deployed in a home environment, and comprises a plurality of high-definition cameras used for collecting pet dynamic video streams in real time; the microphone array is used for collecting environment sound signals; the millimeter wave radar sensor is used for detecting micro-motion characteristics of the life body; the infrared thermal imaging sensor is used for positioning the position of a living body; the environment sensor group comprises a temperature and humidity detection sensor, a door and window state detection sensor and a current detection sensor; and the edge computing gateway is connected with the Internet of Things sensor module and is used for preprocessing the video stream and the sound signal. According to the invention, life experience of pets at home can be effectively improved, and abnormity of the pets can be found in time and remedied.

Description

Technical Field

[0001] This invention relates to the field of home control systems, specifically to an Internet of Things (IoT) based smart home control system. Background Technology

[0002] Smart home technology uses the residence as a platform, integrating facilities related to home life through comprehensive wiring, network communication, security, automatic control, and audio-visual technologies to build an efficient management system for residential facilities and daily household affairs. It aims to enhance home security, convenience, comfort, and aesthetics, while achieving an environmentally friendly and energy-efficient living environment. The smart home control system is the core component of a smart home system. It coordinates and manages various smart devices in the home, enabling interconnection and collaborative operation between these devices. Through the smart home control system, users can easily control and manage home devices, achieving home intelligence and automation.

[0003] Existing patent 202210012354.2 describes an IoT-based smart home control system, comprising: a smart home module including a camera acquisition module; a storage module; a human body sensing module; a hinge switch; a main controller that is communicatively connected to multiple smart home modules, the main controller including an automatic search module; a voice acquisition module; and a storage module that compares real-time item information II with item information I. If the comparison result is "1", the match is successful and the three-dimensional coordinates and number of item information I are returned. The voice acquisition module parses the received three-dimensional coordinates and number into the location of the item and broadcasts it. If the comparison result is "0", the match fails, and the voice acquisition module receives the matching failure instruction, parses it into the voice message "no such item" and broadcasts it.

[0004] Existing technologies often focus solely on enhancing the user's home experience, neglecting the well-being of pets and failing to promptly detect and address any abnormalities in their pets. Therefore, those skilled in the art have developed an Internet of Things (IoT)-based smart home control system to address the problems described in the background. Summary of the Invention

[0005] The purpose of this invention is to provide an Internet of Things-based smart home control system that can effectively improve the living experience of pets at home and promptly detect and remedy any abnormalities in pets, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The IoT-based smart home control system includes an IoT sensor module, an edge computing gateway, a cloud analytics platform, and an execution control module.

[0008] The IoT sensor module is deployed in a home environment and includes: multiple high-definition cameras for real-time acquisition of dynamic video streams of pets; a microphone array for acquisition of environmental sound signals; a millimeter-wave radar sensor for detecting the micro-motion characteristics of living beings; an infrared thermal imaging sensor for locating the position of living beings; and an environmental sensor group, including temperature and humidity, door and window status, and current detection sensors.

[0009] The edge computing gateway is connected to the IoT sensor module for preprocessing video streams and audio signals;

[0010] The cloud-based analysis platform includes: an abnormal behavior analysis module, which analyzes abnormal pet movements, abnormal vocal spectrum, and predicts risks based on image recognition; an emotional state assessment module, which identifies the pet's positive / negative emotional state through a convolutional neural network; and a risk warning decision module, which generates a risk weight value by combining the location danger coefficient and the excitement score.

[0011] The execution control module includes: a voice playback device for playing preset owner voice messages; a smart display screen for playing preset owner videos; and a smart home central controller for linking door locks, feeders, and air conditioning equipment.

[0012] As a further aspect of the present invention: the abnormal behavior analysis module includes:

[0013] The abnormal movement recognition submodule extracts the pet's joint movement trajectory through OpenPose skeletal key point detection technology, and generates a body abnormality alarm when twitching, lameness or unbalanced movements are detected.

[0014] The abnormal vocalization identification submodule extracts the Mel frequency cepstral coefficients (MFCC) of the sound signal, identifies abnormal vocalizations such as moaning and howling through a support vector machine (SVM) classifier, and generates an abnormal vocalization alarm when the vocalization is abnormal.

[0015] The behavior prediction submodule uses the YOLOv5 object detection algorithm to predict pet behavior. When it detects behaviors such as chewing on wires or climbing to high places, it triggers the device to cut off power or close the window.

[0016] As a further aspect of the present invention, the emotional state assessment module performs the following operations:

[0017] Facial expression features of the pet are extracted using a ResNet50 network, and the probability values ​​P for positive / negative emotions are output. 情绪 ;

[0018] When P 情绪 A value <0.35 is considered a negative emotion, triggering a voice reassurance command;

[0019] After the voice reassurance is completed, recalculate P. 情绪 If P情绪 If the value is less than 0.35, a video reassurance will be initiated.

[0020] P is calculated again after the video playback ends. 情绪 If P 情绪 If the value is less than 0.35, an alarm message is sent to the master terminal.

[0021] As a further aspect of the present invention, it also includes a missing pet handling module, the workflow of which is as follows:

[0022] When the target pet does not appear in the monitoring for 30 consecutive minutes, its historical behavior database is retrieved for trajectory comparison.

[0023] If the trajectory deviation exceeds the threshold, the millimeter-wave radar and infrared thermal imaging sensor will be activated to scan the entire house.

[0024] The pet's coordinates (x, y, z) are located by fusing data from multiple sensors.

[0025] Retrieve the preset danger coefficient K for the area to which the coordinates belong, K∈[1,10], where: K=8.5 for the area within 1 meter of the power socket, K=9.0 for the balcony edge area, K=7.0 for the kitchen area, and K=1.0 for the safety mat area.

[0026] As a further aspect of the present invention: the risk warning decision module includes an excitation scoring engine, which executes:

[0027] The number of vocalizations N per unit time is collected using a sound sensor. call ;

[0028] Detecting respiratory rate F using millimeter-wave radar breath (beats / minute) and heart rate amplitude A heart (mm);

[0029] The hyperactivity score S is calculated using the following formula:

[0030] Where α, β and γ are all weighting coefficients and α+β+γ=1, and the S value is normalized to the interval [0,10].

[0031] As a further aspect of the present invention: the formula for calculating the risk weight value W is:

[0032] W = K × S, where: K is the regional risk coefficient (K≥1), and S is the hyperactivity score (0≤S≤10);

[0033] The risk warning decision module adopts a corresponding risk classification strategy based on the risk weight value W. The risk classification strategy includes:

[0034] W < 4: Log entries are recorded without alerts;

[0035] 4≤W<7: Send an app notification to the owner;

[0036] 7≤W<9: Initiate physical isolation and issue a telephone alarm (the physical isolation is performed through a dynamic barrier generation unit, an odor induction device, and an emergency power-off unit);

[0037] W≥9: Triggers the whole-house emergency mode and activates the community security system.

[0038] As a further embodiment of the present invention: the voice playback device includes:

[0039] The master voiceprint modeling unit generates personalized voice packs through the WaveNet network;

[0040] The emotion mapping unit maps comforting statements to a high-frequency, gentle tone.

[0041] The sound field positioning unit uses beamforming technology to direct the sound to the location of the pet.

[0042] As a further embodiment of the present invention: the execution control module further includes:

[0043] The dynamic barrier generation unit generates a virtual isolation wall in the danger zone through laser projection;

[0044] Scent-guided devices release pheromones to lure pets away from high-risk areas;

[0045] The emergency power-off unit cuts off the power to the corresponding circuit within 0.2 seconds when it detects a pet chewing on the wires.

[0046] As a further aspect of the present invention, the system also includes a multi-pet identification database for storing biometric templates (including iris, coat color and pattern), historical health baseline data, and behavioral preference maps for each pet; the multi-pet identification database is stored on an edge computing gateway and a cloud analysis platform using blockchain technology.

[0047] As a further improvement of the present invention: the system is equipped with a dual-redundant communication mechanism.

[0048] The main communication link uses 5G / WiFi 6 to transmit video data;

[0049] The backup communication link uses LoRa to transmit sensor alarm signals;

[0050] Automatically switch to backup link when the primary link latency is greater than 500ms.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] This invention significantly improves the safety of pets when they are alone through a multimodal perception and intelligent decision-making collaboration mechanism. The IoT sensor module integrates vision, acoustics, bio-radar, and environmental detection to achieve full-space status monitoring; the edge-cloud collaborative architecture preprocesses high-bandwidth data locally, reducing response latency while ensuring the accuracy of complex analyses. The core technological value is reflected in triple protection: at the risk prediction level, a pre-warning mechanism is built through skeletal motion tracking, vocal spectrum analysis, and YOLOv5 behavior recognition to prevent high-risk behaviors such as chewing on electrical wires from occurring; at the dynamic response level, a tiered intervention (voice → video → alarm) based on emotional probability values ​​and arousal scores, along with K-value mapping of dangerous areas, achieves precise graded responses; at the active protection level, directional sound field playback, laser virtual barriers, and 0.2-second emergency power outages form a multi-dimensional intervention matrix. Dual redundant communication and blockchain storage ensure system reliability, ultimately achieving the dual benefits of reduced pet accidental injury rates and improved owner monitoring efficiency. Attached Figure Description

[0053] Figure 1 This is a block diagram of an IoT-based smart home control system. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] As mentioned in the background section of this application, research has found that existing home control systems typically focus only on improving the user's home experience, while neglecting the living experience of pets at home. They are unable to detect and remedy any abnormalities in pets in a timely manner, thus exhibiting certain shortcomings.

[0056] To address the aforementioned shortcomings, this application discloses an Internet of Things-based smart home control system that can effectively improve the living experience of pets at home and promptly detect and remedy any abnormalities in pets.

[0057] The following will describe in detail, with reference to the accompanying drawings, how the solution of this application solves the above-mentioned technical problems.

[0058] Please see Figure 1In this embodiment of the invention, the IoT-based smart home control system includes an IoT sensor module, an edge computing gateway, a cloud analysis platform, and an execution control module. The IoT sensor module, deployed in the home environment, includes: multiple high-definition cameras for real-time acquisition of dynamic video streams of the pet; a microphone array for acquiring environmental sound signals; a millimeter-wave radar sensor for detecting micro-motion characteristics of living organisms; an infrared thermal imaging sensor for locating the position of the living organism; and an environmental sensor group including temperature and humidity sensors, door and window status sensors, and current detection sensors. The edge computing gateway connects to the IoT sensor module and is used to preprocess the video streams and sound signals. The cloud analysis platform includes: an abnormal behavior analysis module, which analyzes abnormal pet movements, abnormal vocalizations, and predicts behavioral risks based on image recognition; an emotional state assessment module, which identifies the pet's positive / negative emotional state through a convolutional neural network; and a risk warning decision module, which generates a risk weight value by combining the location hazard coefficient and the excitement score. The execution control module includes: a voice playback device for playing preset owner voice messages; a smart display screen for playing preset owner videos; and a smart home central controller that links door locks, feeders, and air conditioning equipment. This application constructs a complete smart home pet monitoring system architecture. The IoT sensor module comprehensively collects pet physiological behavior and environmental data through multimodal sensing (video, audio, millimeter wave, infrared, and environmental parameters), solving the problem of blind spots in single-sensor monitoring. The edge computing gateway performs local preprocessing of high-bandwidth video / audio data, reducing cloud load and improving real-time performance. The core value of the cloud analytics platform lies in transforming raw data into pet health and safety decisions: the abnormal behavior analysis module identifies potential risks through computer vision and acoustic models; the emotion assessment module quantifies the pet's psychological state using deep learning; and the risk warning module generates weighted values ​​by combining environmental hazard coefficients and physiological arousal levels, providing decision-making basis for the execution layer. The execution control module achieves proactive intervention through voice, vision, and device linkage, forming a closed loop of "perception-analysis-execution," significantly reducing the risk of accidents when pets are alone.

[0059] In this embodiment, the abnormal behavior analysis module includes: an abnormal movement recognition submodule, which extracts the pet's joint movement trajectory using OpenPose skeletal keypoint detection technology and generates a body abnormality alarm when twitching, lameness, or imbalance is detected; an abnormal vocalization recognition submodule, which extracts the Mel-frequency cepstral coefficients (MFCC) of the sound signal and identifies abnormal vocalizations such as whimpering and howling using a support vector machine (SVM) classifier, generating an abnormal vocalization alarm when vocalizations are abnormal; and a behavior prediction submodule, which predicts pet behavior based on the YOLOv5 object detection algorithm, triggering device power-off or window closure when behaviors such as biting electrical wires or climbing to high places are detected. This setup refines the three-layer defense mechanism for abnormal behavior analysis. The abnormal movement recognition submodule uses OpenPose skeletal tracking technology to detect pathological behaviors such as twitching / lameness through joint movement trajectory modeling, overcoming the limitation of traditional image analysis in capturing micro-movements. The abnormal vocalization recognition submodule combines Mel-frequency cepstral coefficients (MFCC) with support vector machines to classify stress vocalizations such as whimpering / howling, compensating for the auditory blind spots of pure visual monitoring. The behavior prediction submodule uses YOLOv5 to detect high-risk behaviors (such as biting wires) in real time and triggers device linkage (such as power outage) in the early stage of the behavior, overcoming the lag of post-event response and forming a dual protection of pre-event prevention and in-event blocking.

[0060] In this embodiment, the emotion state assessment module performs the following operations: extracts the pet's facial expression features using a ResNet50 network and outputs a positive / negative emotion probability value P. 情绪 When P 情绪 A value <0.35 is considered a negative emotion, triggering a voice reassurance command; after the voice reassurance is completed, P is recalculated. 情绪 If P 情绪 If P < 0.35, video reassurance will be initiated; P will be recalculated after the video playback ends. 情绪 If P 情绪 If the value is less than 0.35, an alarm message is sent to the owner's terminal. This setting establishes a tiered intervention process for emotional state assessment. The ResNet50 network extracts facial micro-expression features and outputs an emotion probability value P. 情绪 This addresses the error problem in subjective emotion interpretation. When P 情绪 When the threshold for negative emotions is less than 0.35, the system prioritizes voice reassurance (a low-intervention-cost approach); if ineffective, it escalates to video reassurance (a moderate-intervention approach); and if no improvement is achieved, it alerts the owner (a high-intervention level). This gradual strategy avoids excessive interference with the pet while ensuring effective reporting of emotional crises, making it particularly suitable for the long-term management of chronic psychological problems such as separation anxiety.

[0061] This embodiment also includes a missing pet handling module, whose workflow is as follows: when a target pet has not appeared in the monitoring for 30 consecutive minutes, its historical behavior database is retrieved for trajectory comparison; if the trajectory deviation exceeds a threshold, millimeter-wave radar and infrared thermal imaging sensors are activated to scan the entire house; the pet's coordinates (x, y, z) are located through multi-sensor data fusion; a preset danger coefficient K for the area to which these coordinates belong is retrieved, K∈[1,10], where: K=8.5 for areas within 1 meter of power outlets, K=9.0 for balcony edge areas, K=7.0 for kitchen areas, and K=1.0 for safety bedding areas. This setting enables proactive location and risk assessment of missing pets. The trajectory comparison triggered by 30 minutes of no monitoring records solves the problem of monitoring interruption caused by pets hiding. The multi-sensor fusion of millimeter-wave radar and infrared thermal imaging can accurately locate the coordinates (x, y, z) of a living pet in dark / obscured environments. By introducing a regional hazard factor K: based on the coordinate mapping and preset risk values ​​(e.g., K = 9.0 for balconies), the physical space is digitized into a hazard level map, providing spatial dimension parameters for subsequent risk decisions and avoiding the one-sidedness of traditional solutions that rely solely on time thresholds.

[0062] In this embodiment, the risk warning decision module includes an excitation scoring engine, which executes: collecting the number of calls N per unit time using a sound sensor. call ; Detecting respiratory rate F using millimeter-wave radar breath (beats / minute) and heart rate amplitude A heart (mm); The hyperactivity score S is calculated using the following formula: Where α, β, and γ are weighting coefficients and α + β + γ = 1, the S value is normalized to the [0, 10] interval. This setting defines the quantitative model for the hyperactivity score. The normalized hyperactivity score S is calculated by weighting three physiological indicators: vocalization frequency, respiratory rate, and heart rate amplitude. This model integrates discrete physiological signals into a continuous score, overcoming the susceptibility of single indicators (such as vocalization frequency) to environmental interference. The coefficients α, β, and γ can be weighted according to different pet breeds (e.g., vocalization for dogs, heart rate for cats), achieving personalized hyperactivity assessment and providing accurate physiological dimension parameters for risk warning.

[0063] In this embodiment, the risk weight value W is calculated using the formula: W = K × S, where K is the regional hazard coefficient (K≥1), and S is the arousal score (0≤S≤10). The risk warning decision module adopts corresponding risk grading strategies based on the risk weight value W. These strategies include: W < 4: Log without alarm; 4 ≤ W < 7: Send APP notification to the owner; 7 ≤ W < 9: Initiate physical isolation and issue a telephone alarm (physical isolation is executed through a dynamic barrier generation unit, odor induction device, and emergency power-off unit); W ≥ 9: Trigger the whole-house emergency mode and link the community security system. This setup constructs a risk grading response system. By multiplying the spatial hazard coefficient (K) and physiological arousal level (S) using the formula W = K × S to generate the comprehensive risk weight value W, the limitations of single-dimensional risk assessment are overcome. A four-level strategy based on the W value: W < 4, only recording (to avoid false alarms); 4 ≤ W < 7, push notification (mild warning); 7 ≤ W < 9, physical isolation + telephone alarm (moderate intervention); W ≥ 9, triggering whole-house emergency mode + linkage with community security (severe emergency). This design enables resource allocation on demand, ensuring priority response to high-crisis situations.

[0064] In this embodiment, the voice playback device includes: a voiceprint modeling unit for the owner, which generates personalized voice packages through a WaveNet network; an emotion mapping unit, which maps soothing statements into high-frequency, gentle tones; and a sound field localization unit, which uses beamforming technology to direct the playback to the pet's location. This setup optimizes the targeted delivery of soothing voice messages. WaveNet voiceprint modeling restores the owner's voice timbre, addressing the pet's aversion to mechanical voices; the emotion mapping unit converts text into high-frequency, gentle tones, aligning with animal auditory preferences; and the sound field localization unit uses beamforming technology to focus sound energy onto the pet's coordinates, overcoming the spatial diffusion limitations of traditional speakers and enabling clear soothing commands to be delivered even in noisy environments, thus improving the success rate of behavioral interventions.

[0065] In this embodiment, the execution control module further includes: a dynamic barrier generation unit that generates a virtual barrier wall in the danger zone using laser projection; an odor induction device that releases pheromones to guide the pet away from the high-risk area; and an emergency power cut-off unit that cuts off the power to the corresponding circuit within 0.2 seconds when it detects a pet chewing on an electrical wire. This setup expands active protection measures. The dynamic barrier generation unit uses laser projection to replace physical fences, generating a visible virtual wall in danger zones (such as stairwells), solving the problem of traditional fences ruining the aesthetics of the home; the odor induction device guides the pet away from the high-risk area by releasing pheromones, utilizing biological instincts to improve induction efficiency; and the emergency power cut-off unit uses a 0.2-second rapid power cut-off, significantly reducing the risk of electric shock compared to traditional circuit breakers (more than 2 seconds). The three components work together to form a three-dimensional "visual-olfactory-electrical" protection network.

[0066] In this embodiment, the system also includes a multi-pet identification database for storing each pet's biometric templates (including iris, coat color and pattern), historical health baseline data, and behavioral preference maps. The multi-pet identification database is stored on an edge computing gateway and a cloud analytics platform using blockchain technology. This setup enables personalized management of multiple pets. Biometric templates (such as cat face recognition) support accurate individual identification; historical health baseline data (such as normal heart rate range) provides a reference benchmark for anomaly detection; and behavioral preference maps (such as frequently active areas) assist in predicting missing pet trajectories. Blockchain storage ensures that data is synchronized and immutable at the edge and in the cloud, solving data consistency issues in a distributed architecture while meeting the compliance requirements for pet medical data.

[0067] In this embodiment, the system employs a dual-redundancy communication mechanism: the primary communication link uses 5G / WiFi 6 to transmit video data; the backup communication link uses LoRa to transmit sensor alarm signals; and the system automatically switches to the backup link when the primary link latency exceeds 500ms. This configuration ensures the reliability of system communication. The primary link (5G / WiFi 6) prioritizes the transmission of high-bandwidth video streams, while the backup link (LoRa) is dedicated to low-power sensor alarm signals. This dual-channel design avoids single points of failure. The 500ms latency threshold triggers automatic switching, ensuring that real-time commands (such as emergency power outages) are never lost, effectively addressing the industry pain point of monitoring interruptions during network congestion.

[0068] This invention significantly improves the safety of pets when they are alone through a multimodal perception and intelligent decision-making collaboration mechanism. The IoT sensor module integrates vision, acoustics, bio-radar, and environmental detection to achieve full-space status monitoring; the edge-cloud collaborative architecture preprocesses high-bandwidth data locally, reducing response latency while ensuring the accuracy of complex analyses. The core technological value is reflected in triple protection: at the risk prediction level, a pre-warning mechanism is built through skeletal motion tracking, vocal spectrum analysis, and YOLOv5 behavior recognition to prevent high-risk behaviors such as chewing on electrical wires from occurring; at the dynamic response level, a tiered intervention (voice → video → alarm) based on emotional probability values ​​and arousal scores, along with K-value mapping of dangerous areas, achieves precise graded responses; at the active protection level, directional sound field playback, laser virtual barriers, and 0.2-second emergency power outages form a multi-dimensional intervention matrix. Dual redundant communication and blockchain storage ensure system reliability, ultimately achieving the dual benefits of reduced pet accidental injury rates and improved owner monitoring efficiency.

[0069] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0070] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A smart home control system based on the Internet of Things, characterized in that, This includes IoT sensor modules, edge computing gateways, cloud analytics platforms, and execution control modules; The IoT sensor module is deployed in a home environment and includes: multiple high-definition cameras for real-time acquisition of dynamic video streams of pets; a microphone array for acquisition of environmental sound signals; a millimeter-wave radar sensor for detecting the micro-motion characteristics of living beings; an infrared thermal imaging sensor for locating the position of living beings; and an environmental sensor group, including temperature and humidity, door and window status, and current detection sensors. The edge computing gateway is connected to the IoT sensor module for preprocessing video streams and audio signals; The cloud-based analysis platform includes: an abnormal behavior analysis module, which analyzes abnormal pet movements, abnormal vocal spectrum, and predicts risks based on image recognition; an emotional state assessment module, which identifies the pet's positive / negative emotional state through a convolutional neural network; and a risk warning decision module, which generates a risk weight value by combining the location danger coefficient and the excitement score. The execution control module includes: a voice playback device for playing preset owner voice messages; a smart display screen for playing preset owner videos; and a smart home central controller for linking door locks, feeders, and air conditioning equipment.

2. The smart home control system based on the Internet of Things according to claim 1, characterized in that, The abnormal behavior analysis module includes: The abnormal movement recognition submodule extracts the pet's joint movement trajectory through OpenPose skeletal key point detection technology, and generates a body abnormality alarm when twitching, lameness or unbalanced movements are detected. The abnormal vocalization identification submodule extracts the Mel frequency cepstral coefficients of the sound signal and uses a support vector machine classifier to identify abnormal vocalizations such as moaning and howling, and generates an abnormal vocalization alarm when the vocalization is abnormal. The behavior prediction submodule uses the YOLOv5 object detection algorithm to predict pet behavior. When it detects behaviors such as chewing on wires or climbing to high places, it triggers the device to cut off power or close the window.

3. The smart home control system based on the Internet of Things according to claim 2, characterized in that, The emotional state assessment module performs the following operations: Facial expression features of the pet are extracted using a ResNet50 network, and the probability values ​​P for positive / negative emotions are output. 情绪 ; When P 情绪 A value <0.35 is considered a negative emotion, triggering a voice reassurance command; After the voice reassurance is completed, recalculate P. 情绪 If P 情绪 If the value is less than 0.35, a video reassurance will be initiated. P is calculated again after the video playback ends. 情绪 If P 情绪 If the value is less than 0.35, an alarm message is sent to the master terminal.

4. The smart home control system based on the Internet of Things according to claim 3, characterized in that, It also includes a missing pet handling module, whose workflow is as follows: When the target pet does not appear in the monitoring for 30 consecutive minutes, its historical behavior database is retrieved for trajectory comparison. If the trajectory deviation exceeds the threshold, the millimeter-wave radar and infrared thermal imaging sensor will be activated to scan the entire house. The pet's coordinates (x, y, z) are located by fusing data from multiple sensors. Retrieve the preset danger coefficient K for the area to which the coordinates belong, K∈[1,10], where: K=8.5 for the area within 1 meter of the power socket, K=9.0 for the balcony edge area, K=7.0 for the kitchen area, and K=1.0 for the safety mat area.

5. The smart home control system based on the Internet of Things according to claim 4, characterized in that, The risk warning decision module includes an excitation scoring engine, which executes as follows: The number of vocalizations N per unit time is collected using a sound sensor. call ; Detecting respiratory rate F using millimeter-wave radar breath and heart rate amplitude A heart ; The hyperactivity score S is calculated using the following formula: Where α, β and γ are all weighting coefficients and α+β+γ=1, and the S value is normalized to the interval [0,10].

6. The smart home control system based on the Internet of Things according to claim 5, characterized in that, The formula for calculating the risk weight value W is: W = K × S, where: K is the regional risk coefficient, and S is the hyperactivity score; The risk warning decision module adopts a corresponding risk classification strategy based on the risk weight value W. The risk classification strategy includes: W < 4: Log entries are recorded without alerts; 4≤W<7: Send an app notification to the owner; 7≤W<9: Initiate physical isolation and issue a telephone alarm; W≥9: Triggers the whole-house emergency mode and activates the community security system.

7. The smart home control system based on the Internet of Things according to claim 6, characterized in that, The voice playback device includes: The master voiceprint modeling unit generates personalized voice packs through the WaveNet network; The emotion mapping unit maps comforting statements to a high-frequency, gentle tone. The sound field positioning unit uses beamforming technology to direct the sound to the location of the pet.

8. The smart home control system based on the Internet of Things according to claim 7, characterized in that, The execution control module further includes: The dynamic barrier generation unit generates a virtual isolation wall in the danger zone through laser projection; Scent-guided devices release pheromones to lure pets away from high-risk areas; The emergency power-off unit cuts off the power to the corresponding circuit within 0.2 seconds when it detects a pet chewing on the wires.

9. The smart home control system based on the Internet of Things according to claim 8, characterized in that, The system also includes a multi-pet identification database, which stores the biometric templates, historical health baseline data, and behavioral preference maps of each pet; the multi-pet identification database is stored on an edge computing gateway and a cloud analysis platform using blockchain technology.

10. The smart home control system based on the Internet of Things according to claim 9, characterized in that, The system is equipped with a dual-redundant communication mechanism: The main communication link uses 5G / WiFi 6 to transmit video data; The backup communication link uses LoRa to transmit sensor alarm signals; Automatically switch to backup link when the primary link latency is greater than 500ms.

Images