Internet-based household intelligent alarm device

Abstract

The invention discloses a household intelligent alarm device based on the Internet, and relates to the technical field of intelligent security and protection, and the device comprises an alarm information generation module which carries out the fusion of multi-source feature data, calculates an alarm index through a weighted summation method, and generates alarm information; the control instruction module is used for encrypting the alarm information, generating an encrypted data packet, sending the encrypted data packet to a remote user side through the Internet and receiving a control instruction of the user side; and the alarm command module is used for dynamically adjusting the alarm strategy based on the control instruction of the user side and generating an alarm processing command. By dynamically distributing the weight of the multi-source sensor, the adaptability of the device is improved, and false alarms caused by environment changes in the daytime and at night are reduced; in combination with historical false alarm data analysis and a self-adaptive updating algorithm, an alarm threshold is continuously optimized, the alarm accuracy is improved, and the false alarm rate and the missing report rate are reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent security, and in particular to a home intelligent alarm device based on the Internet. Background Art

[0002] With the rapid development of Internet of Things technology, artificial intelligence, and cloud computing, home security protection devices are evolving towards intelligence and remote control. In the past, home security devices mainly used single sensors (such as infrared sensors and cameras) for alarm detection and notified users through local alarms and text messages. However, these devices have problems such as high false alarm rates and poor environmental adaptability, and cannot meet users' security needs for high precision and low false alarms. In recent years, intelligent alarm devices based on multi-sensor fusion have become a research hotspot. By using multi-source sensors such as PIR sensors, cameras, and sound sensors, and combining multi-source data processing and intelligent analysis methods, the accuracy of alarms is effectively improved. In addition, with the popularization of the Internet, the functions of remote monitoring and user interaction have become key requirements for home alarm devices, enabling users to control the home security status at any time and remotely process alarm information.

[0003] However, the existing intelligent home alarm devices still have some obvious deficiencies. First, the current multi-sensor fusion technology still has limitations in data processing. Many devices rely on simple threshold judgment and rule matching algorithms for alarm decision-making, lacking in-depth analysis of sensing data, resulting in relatively high false alarm rates and missed alarms in complex environments. Second, the existing alarm information interaction methods are relatively single. Some devices only support text messages and mobile APP notifications, lacking an encryption transmission mechanism, which poses a security risk. In addition, the device cannot dynamically adjust the alarm strategy according to user feedback, resulting in the inability of users to make effective adaptive adjustments after false alarms or false triggers, affecting the user experience. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a home intelligent alarm device based on the Internet to solve the problems of alarm accuracy and intelligent control.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides an Internet-based home intelligent alarm device, which includes a data acquisition module for acquiring multi-source sensor data and generating a sensing data packet; a feature data generation module for extracting features through low-pass filtering, short-time Fourier transform, and inter-frame pixel difference based on the sensing data packet to generate multi-source feature data; an alarm information generation module for fusing the multi-source feature data, calculating an alarm index using the weighted summation method, and generating alarm information; a control instruction module for encrypting the alarm information, generating an encrypted data packet, sending it to a remote user terminal through the Internet, and receiving the control instruction of the user terminal; and an alarm command module for dynamically adjusting the alarm strategy based on the control instruction of the user terminal and generating an alarm processing command.

[0008] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, wherein: the steps of acquiring multi-source sensor data and generating a sensing data packet are as follows:

[0009] Acquire PIR sensor data, microphone data, audio signals, spectral features, image frames, and image features through a PIR sensor, a microphone sensor, and a camera to generate multi-source sensor data;

[0010] Synchronize the data of the multi-source sensors according to the time stamp to generate a sensing data packet.

[0011] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, wherein: the steps of extracting features through low-pass filtering, short-time Fourier transform, and inter-frame pixel difference based on the sensing data packet to generate multi-source feature data are as follows:

[0012] Based on the data of the PIR sensor, use low-pass filtering to remove high-frequency noise, obtain the smoothed signal of the PIR sensor, and calculate using the first-order difference method to obtain the change rate of the PIR sensor;

[0013] Based on the microphone sensor data, segment the audio signal in a sliding window manner, and perform short-time Fourier transform on each frame of audio data to obtain the spectral features of the audio signal;

[0014] Based on the camera sensor data, use the inter-frame difference method to calculate the pixel change value between the current frame and the previous frame, and extract the moving target;

[0015] According to the moving target, use image processing for edge detection to obtain the moving contour information;

[0016] Combine the change rate of the PIR sensor, the spectral features of the audio signal, and the moving contour information detected by the camera to generate multi-source feature data.

[0017] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, the following steps are carried out: edge detection is performed using image processing based on a moving target to obtain moving contour information, and the specific steps are as follows.

[0018] Based on the moving target, Gaussian filtering is performed to remove noise and smooth the image.

[0019] The Canny operator is used to calculate the gradient of the smoothed image to obtain the target edge image.

[0020] The edge region is extracted from the target edge image, and double-threshold edge detection is used to remove weak noise edges.

[0021] The broken edges are connected through an edge tracking algorithm.

[0022] Connected component analysis is performed on the processed edge image to label all moving targets, and the convex hull algorithm is used to fit the moving target contour.

[0023] The number of pixel points within the contour is counted, the size of the moving target is estimated, and the contour area is calculated to obtain the moving contour information.

[0024] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, the following steps are carried out: the multi-source feature data are fused, and the weighted summation method is used to calculate the alarm index to generate alarm information.

[0025] The multi-source feature data are normalized using the linear normalization method.

[0026] According to the environmental conditions, weights are assigned to the multi-source sensor data, and an alarm threshold is preset.

[0027] The normalized PIR change rate, audio features, and moving target area are respectively multiplied by the assigned weights to obtain weighted feature values, and the alarm index is obtained through the weighted summation method.

[0028] Based on the alarm index, a judgment is made through the preset alarm threshold to generate alarm information.

[0029] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, the following steps are carried out: according to the environmental conditions, weights are assigned to the multi-source sensor data, and an alarm threshold is preset.

[0030] Through the environmental condition adaptive algorithm, the weight of the audio features in the night mode of the multi-source sensor is increased and the weight of the visual features is decreased, and the weight of the visual features in the day mode of the multi-source sensor is increased and the weight of the audio features is decreased.

[0031] Based on the weights of multi-source sensor data after allocation and combined with historical alarm data, statistical analysis is carried out to set the alarm threshold.

[0032] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, wherein: based on the alarm index, judgment is made through a preset alarm threshold to generate alarm information. The specific steps are as follows.

[0033] When the alarm index is greater than the alarm threshold, it indicates that there is an abnormal situation in the current environment and an alarm needs to be triggered.

[0034] When the alarm index is less than the alarm threshold, it indicates that the current environment is normal and there is no need to trigger an alarm.

[0035] Based on the triggered alarm status, record the alarm index, alarm time and event location of the alarm event to generate alarm information.

[0036] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, wherein: encrypt the alarm information to generate an encrypted data packet, send it to the remote user terminal through the Internet, and receive the control instruction of the user terminal. The specific steps are as follows.

[0037] Based on the alarm information, perform multi-source data encapsulation, and use a symmetric encryption algorithm to encrypt the multi-source data to generate an encrypted data packet.

[0038] Send the encrypted data packet to the remote user terminal through the Internet, and receive the control instruction of the user terminal through the MQTT protocol.

[0039] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, wherein: based on the control instruction of the user terminal, dynamically adjust the alarm strategy to generate an alarm processing command. The specific steps are as follows.

[0040] According to the deviation between the alarm index and the actual security event, obtain the historical alarm error data set.

[0041] According to the historical alarm error data set and the alarm adjustment coefficient, calculate the adaptive alarm threshold to obtain the optimized adaptive alarm threshold.

[0042] According to the optimized adaptive alarm threshold, evaluate the alarm trigger status to generate an alarm processing command.

[0043] As a preferred solution of the Internet-based home intelligent alarm device described in the present invention, wherein: according to the historical alarm error data set and the alarm adjustment coefficient, calculate the adaptive alarm threshold to obtain the optimized adaptive alarm threshold. The specific steps are as follows.

[0044] Dynamically generate an alarm adjustment coefficient by integrating historical false alarm analysis and user manual setting to adjust parameters;

[0045] Optimize using the adaptive dynamic update method based on historical alarm errors and the alarm adjustment coefficient to obtain an optimized adaptive alarm threshold.

[0046] The beneficial effects of the present invention are as follows: By dynamically allocating the weights of multi-source sensors, the adaptability of the device is improved, and false alarms caused by environmental changes between day and night are reduced; Combining historical false alarm data analysis and the adaptive update algorithm, the alarm threshold is continuously optimized, the alarm accuracy is improved, and the false alarm and missed alarm rates are reduced; At the same time, it supports users to customize the weight adjustment strategy to achieve personalized settings, making the device more in line with the needs of different users and enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0048] Figure 1 It is a schematic diagram of a home intelligent alarm device based on the Internet in Embodiment 1.

[0049] Figure 2 It is a schematic diagram of the multi-source feature data processing flow in Embodiment 1.

[0050] Figure 3 It is a schematic diagram of the image feature extraction flow in Embodiment 1.

[0051] Figure 4 It is a schematic diagram of the alarm threshold adaptive adjustment flow in Embodiment 1. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention will be described in detail below with reference to the drawings in the specification.

[0053] Embodiment 1, referring to Figure 1 and Figure 2 , is the first embodiment of the present invention. This embodiment provides a home intelligent alarm device based on the Internet, including the following steps:

[0054] A data acquisition module that acquires multi-source sensor data and generates a sensing data packet.

[0055] Collect PIR sensor data, microphone data, audio signals, spectral features, image frames, and image features through PIR sensors, microphone sensors, and cameras to generate multi-source sensor data;

[0056] It should be noted that the PIR sensor determines whether there is an object moving within the monitoring range by sensing changes in infrared radiation and can provide basic motion detection data. Audio data can reflect abnormal sounds (such as glass breaking sounds, impact sounds, etc.) and, combined with data from other sensors, improve the reliability of alarms. The camera can record real-time images of the monitored area and extract feature information such as moving targets and edge contours through subsequent image processing techniques to improve the ability to judge abnormal situations.

[0057] Synchronize the data of multi-source sensors according to timestamps to generate sensing data packets;

[0058] It should be noted that a unified time reference is used to add timestamps to the data of all sensors to ensure the time consistency of the data. For cases where the data acquisition times are inconsistent, interpolation or alignment algorithms are used to adjust the time points of different sensor data to align them to the nearest common time window.

[0059] The feature data generation module extracts features based on the sensing data packets through low-pass filtering, short-time Fourier transform, and inter-frame pixel difference to generate multi-source feature data.

[0060] Based on the data of the PIR sensor, use low-pass filtering to remove high-frequency noise, obtain the smoothed signal of the PIR sensor, and calculate using the first-order difference method to obtain the change rate of the PIR sensor;

[0061] It should be noted that the low-pass filter is used to retain the low-frequency moving signal of the PIR sensor while suppressing high-frequency noise. Smooth the PIR signal to make the detected moving curve more stable and reduce false alarms caused by high-frequency jitter. Quickly identify significant changes in the PIR signal to determine whether there is a moving target. If the change rate is large, it means that the sensor has detected a relatively fast movement; if the change rate is close to zero, it indicates that the environment is stationary.

[0062] Based on the microphone sensor data, segment the audio signal in a sliding window manner, and perform short-time Fourier transform on each frame of audio data to obtain the spectral features of the audio signal;

[0063] It should be noted that by maintaining the timing information, it is possible to analyze the changes of the audio on the time axis, avoiding information loss caused by processing the entire audio at once. This improves the spectral stability, with the audio signal changing less within a short period, making feature extraction more stable. Through the STFT transformation, the energy distribution of the audio signal at different frequencies can be obtained, that is, the spectral features. Low frequencies may correspond to footsteps, wind sounds, etc., while high frequencies may correspond to glass breaking sounds, alarm sounds, etc., which helps to distinguish different alarm events.

[0064] Based on the camera sensor data, the frame difference method is used to calculate the pixel change value between the current frame and the previous frame, and the moving target is extracted;

[0065] It should be noted that the video stream obtained by the camera sensor consists of a series of consecutive image frames, and the changes between adjacent frames can reflect the dynamic information in the environment. Through the frame difference method, the pixel values of the current frame and the previous frame are calculated to obtain the pixel change area, thereby extracting the possible moving target. This method can effectively detect moving objects, such as intruders or abnormal activities, while reducing the impact of the static background on the detection result, improving the recognition accuracy of the moving target, and providing basic data for subsequent edge detection and target tracking.

[0066] Based on the moving target, edge detection is performed using image processing to obtain the motion contour information.

[0067] Based on the moving target, through Gaussian filtering, noise is removed and the image is smoothed;

[0068] It should be noted that Gaussian filtering is used to smooth the image, removing small-scale noise and making edge detection more stable. This improves the recognition effect of the moving target, ensuring that real intruders or abnormal situations are accurately detected, rather than misidentifying background noise or shadows.

[0069] The Canny operator is used to calculate the gradient of the smoothed image to obtain the target edge image;

[0070] It should be noted that by calculating the gradient of the image, the Canny operator can clearly identify the contour of the moving target. Even in complex backgrounds or low-light conditions, it can effectively distinguish the target. The non-maximum suppression and double-threshold detection of the Canny operator can effectively eliminate noise points and false edges, thereby reducing false alarms and missed detections.

[0071] The edge region is extracted from the target edge image, and double-threshold edge detection is used to remove weak noise edges;

[0072] It should be noted that the dual-threshold method can effectively remove noise in the image, especially weak noise edges. By setting reasonable low and high thresholds, it can ensure that only real edge information is retained. In an intelligent alarm device, the accuracy of edge detection is crucial. By removing weak noise edges, the device can avoid misidentifying noise as a moving target, reducing false alarms, and at the same time, it can also avoid missing real edge information.

[0073] Connect the broken edges through an edge tracking algorithm;

[0074] It should be noted that first, select an unprocessed edge point from the detected edge map as the starting point. Starting from the starting point, gradually expand to adjacent pixel points in a certain search direction to find adjacent edge points. If it is found that the adjacent pixel points also belong to the edge, connect these points to the current edge. If there is a break in the current edge and there are other edge points connecting the two broken parts, connect the broken parts through these points. Search until no more edge points are found, or all edge points have been tracked and connected.

[0075] Perform connected component analysis on the processed edge image, label all moving targets, and use the convex hull algorithm to fit the contour of the moving target;

[0076] It should be noted that the convex hull algorithm simplifies the geometric calculation of the target by fitting the target into a simple convex shape, facilitating subsequent target analysis and tracking. For complex target contours, the convex hull can effectively remove the influence of irregular shapes, making the shape of the target more standardized and easier to process.

[0077] Count the number of pixel points inside the contour, estimate the size of the moving target, calculate the contour area, and obtain the moving contour information;

[0078] It should be noted that according to the resolution of the image, the number of pixel points of the area can be converted into the actual physical area. Assuming the actual size of each pixel, the number of pixels in the image can be converted into the physical size in the real world through the relationship between the pixel area and the unit pixel. After obtaining the target contour through connected component analysis, image processing tools can be used to directly calculate the area of the contour. This area refers to the total number of all pixel points inside the contour and can be converted into the physical area through the unit length of the pixel. The larger the area, the larger the space occupied by the target in the image. The device can further evaluate whether the target is a threat based on this information.

[0079] Combine the change rate of the PIR sensor, the spectral characteristics of the audio signal, and the moving contour information detected by the camera to generate multi-source feature data;

[0080] It should be noted that by combining data from multiple sensors, the motion characteristics, sound characteristics, and heat changes of the target can be comprehensively analyzed, thereby reducing false alarms and missed alarms caused by misjudgment of a single sensor.

[0081] An alarm information generation module that fuses multi-source feature data, calculates an alarm index using the weighted summation method, and generates alarm information.

[0082] The multi-source feature data is normalized using the linear normalization method.

[0083] It should be noted that different types of sensor data may have different dimensions (such as temperature, audio spectrum, pixel values, etc.). Through linear normalization, the differences brought about by these dimensions can be eliminated, enabling each feature to have the same scale during processing. The normalized data can be more conveniently used for weighted summation or other fusion methods, ensuring more reasonable fusion between different data sources and enhancing the accuracy of overall decision-making.

[0084] According to the environmental conditions, weights are assigned to the multi-source sensor data, and an alarm threshold is preset.

[0085] Through the environmental condition adaptive algorithm, the audio feature weight of the multi-source sensor in night mode is increased and the visual feature weight is reduced, while the visual feature weight in day mode is enhanced and the audio feature weight is reduced.

[0086] It should be noted that through the environmental condition adaptive algorithm, the reference weights of different features are set. The weight of the visual feature is w v , the weight of the audio feature is w a , and the weight of the infrared thermal imaging feature is w t ;

[0087] Calculate the weight adjustment coefficient of the visual feature. The expression is:

[0088] λ v =1 - L ′ ;

[0089] Among them, λ v represents the influence of light on the visual feature. At night, the visual signal is weak, and the weight is reduced. L ′ represents the normalized light intensity.

[0090] Calculate the weight adjustment coefficient of the audio feature. The expression is:

[0091] λ a =1 - N ′ ;

[0092] Among them, λ a represents the influence of environmental noise on the audio feature. When the noise is large, the audio analysis error is large, and the weight is reduced. N′ Indicates the normalized noise intensity.

[0093] Calculate the weight adjustment coefficient of the infrared thermal imaging feature, and the expression is:

[0094] λ t = 1 - T ′ ;

[0095] Among them, λ t Indicates the influence of temperature on the thermal imaging feature. When the temperature is high, the temperature difference between the human body and the environment decreases, reducing the weight. T ′ Indicates the normalized temperature.

[0096] Calculate the environment adaptive comprehensive adjustment factor, and the expression is:

[0097]

[0098] Among them, λ indicates the overall environment adaptation factor.

[0099] Calculate the final visual feature weight according to the environment adaptation factor λ, and the expression is:

[0100]

[0101] Among them, w v Indicates the final weight of the visual feature, the signal importance of this feature in the current environment, Indicates the default weight of the visual feature is 50%, k v Indicates the adjustment factor of the visual feature weight, used to control the influence degree of environmental changes on the visual feature weight. (1 - λ) indicates that when the environment is harsh, the amplitude of weight adjustment is larger.

[0102] Calculate the final audio feature weight, and the expression is:

[0103]

[0104] Among them, w a Indicates the final weight of the audio feature, the contribution degree of the audio signal, Indicates that the default weight of the audio feature is 30%, k a Indicates the adjustment factor of the audio feature weight, used to control the influence degree of environmental noise changes on the audio feature weight.

[0105] Calculate the final infrared thermal imaging feature weight, and the expression is:

[0106]

[0107] Among them, w t Indicates the final weight of the infrared thermal imaging feature, the importance of this feature in the current environment, Indicates that the default weight of the infrared thermal imaging feature is 20%, k t Represents the adjustment factor for the weight of the infrared thermal imaging feature, which is used to control the influence degree of environmental temperature change on the weight of the thermal imaging feature.

[0108] At night, the ambient light is low, the effectiveness of visual features decreases, and audio features (such as sound detection) are more prominent. The night mode will increase the importance of audio data in the alarm decision by increasing the weight of audio features, while reducing the weight of visual features to reduce the interference caused by insufficient light in visual sensors (such as cameras). In the daytime environment, with sufficient light, visual sensors (such as cameras) can usually provide clearer images and motion information. The daytime mode will enhance the weight of visual features, making visual data have a greater impact on the alarm decision. And suppressing audio interference (such as daily noise) helps to reduce false alarms caused by environmental noise (such as traffic noise, human voices, etc.).

[0109] Based on the weights of the multi-source sensor data after allocation and combined with historical alarm data, perform statistical analysis and set the alarm threshold;

[0110] It should be noted that by statistically analyzing the data of historical alarm events, the device can identify in which situations the alarm is more accurate, and in which situations false alarms or missed alarms occur. By analyzing the false alarm and missed alarm patterns in historical data, the device can identify the sources of alarm errors and adjust the alarm threshold accordingly. According to the statistical results and false alarm rate of historical data, the device can adaptively update the alarm threshold. For example, when the device identifies that the audio interference situation in a high-noise environment may cause false alarms, the device will adjust the alarm threshold to avoid overly sensitive alarm responses; conversely, in some more urgent situations, the threshold will be appropriately lowered to improve the alarm response.

[0111] Multiply the normalized PIR change rate, audio feature, and moving target area by the allocated weights respectively to obtain weighted feature values, and obtain the alarm index through the weighted summation method;

[0112] It should be noted that adding the weighted feature values of all sensors gives a comprehensive alarm index. This alarm index reflects the urgency of the alarm in the current environment. If the alarm index exceeds the preset threshold, the device will trigger an alarm; otherwise, the device considers the environment normal and no alarm is required. The alarm index is a comprehensive indicator that synthesizes the feedback information of PIR sensors, audio sensors, and visual sensors. The higher the alarm index, the greater the possibility of abnormal events and the higher the probability of alarm triggering.

[0113] Based on the alarm index, make a judgment through the preset alarm threshold to generate alarm information.

[0114] It should be noted that the alarm threshold is a boundary value set by the alarm device.

[0115] When the alarm index reaches or exceeds this threshold, the device considers that there is a potential anomaly or safety threat in the environment, thus triggering an alarm. The setting of the alarm threshold is usually based on environmental conditions, sensor characteristics, and user requirements.

[0116] When the alarm index is greater than the alarm threshold, it indicates that there is an abnormal situation in the current environment and an alarm should be triggered.

[0117] When the alarm index is less than the alarm threshold, it indicates that the current environment is normal and there is no need to trigger an alarm.

[0118] Based on the triggered alarm status, record the alarm index, alarm time, and event location of the alarm event to generate alarm information;

[0119] It should be noted that the value calculated by the alarm device after comprehensively evaluating the environment represents the safety status of the current environment. The higher the alarm index, the greater the possibility of anomaly. Recording the alarm index helps to understand the degree of anomaly in the environment when the alarm occurs, facilitating post-event analysis. Record the specific time of each alarm trigger. The timestamp is an important attribute of the alarm event because it can help users track the specific moment when the event occurred. Users can trace back the event based on the alarm time to check whether there are other abnormal behaviors or potential safety risks within a specific time period. This refers to the specific location where the alarm occurred or the location where the sensor captured the anomaly. Multi-source sensors are responsible for monitoring different areas of the home. The device will record the sensor location where the alarm occurred, or determine the exact location of the event by combining information from multiple sensors.

[0120] The control instruction module encrypts the alarm information to generate an encrypted data packet, sends it to the remote user terminal via the Internet, and receives the control instructions from the user terminal.

[0121] Based on the alarm information, perform multi-source data encapsulation and encrypt the multi-source data using a symmetric encryption algorithm to generate an encrypted data packet;

[0122] It should be noted that the process of data encapsulation involves formatting and organizing data from different sensors into a structured format, usually using the JSON format. The characteristic of symmetric encryption is that the same key is used for data encryption and decryption. Both the sender and the receiver use the same key for encryption and decryption. By using a symmetric encryption algorithm, it can be ensured that the alarm information and multi-source sensor data are not illegally accessed or tampered with during transmission. This is crucial for protecting the privacy of home users and ensuring the security of the device.

[0123] Send the encrypted data packet to the remote user terminal via the Internet and receive the control instructions from the user terminal through the MQTT protocol;

[0124] It should be noted that through the Internet and the MQTT protocol, users can remotely control the alarm device, adjust the alarm strategy at any time, or view alarm information. Users can flexibly adjust the alarm device according to real-time environmental changes, improving the adaptability of the device and personalized services. By encrypting data packets and the security of the MQTT protocol, the security of device data transmission and privacy protection are ensured.

[0125] An alarm command module dynamically adjusts the alarm strategy based on control instructions from the user side to generate an alarm processing command.

[0126] Obtain a historical alarm error data set based on the deviation between the alarm index and the actual security event;

[0127] It should be noted that the detailed information of each alarm event is recorded, including the alarm index value calculated during the alarm, whether an actual security event has occurred, such as intrusion, fire, etc. Each alarm event is marked, and the time, location, and type of occurrence are recorded in detail. According to the comparison between the alarm index and the actual event, the deviation of each alarm, that is, false alarm, the alarm index is relatively high, but there is no actual security event; missed alarm, the alarm index is relatively low, but an actual security event has occurred. The deviation (false alarm and missed alarm) of each alarm event is recorded in the historical alarm error data set. This data set not only contains the results of the alarm but also includes information such as relevant environmental conditions, sensor data, and alarm strategies.

[0128] Calculate the adaptive alarm threshold based on the historical alarm error data set and the alarm adjustment coefficient to obtain the optimized adaptive alarm threshold.

[0129] Dynamically generate an alarm adjustment coefficient by integrating historical false alarm analysis and user manual setting of adjustment parameters;

[0130] It should be noted that historical false alarm data is collected and analyzed, including false alarm rate, missed alarm rate, and environmental characteristics of false alarms, to extract false alarm patterns and missed alarm trends. Combining user-defined adjustment parameters, such as sensor weight preferences or alarm sensitivity settings within a specific period, forms a personalized alarm adjustment strategy. Based on these data, an alarm adjustment coefficient is calculated, which is used to dynamically adjust the alarm threshold and the calculation method of the alarm index, enabling the alarm device to adaptively optimize under different environments and user requirements, improving alarm accuracy, and reducing false alarm and missed alarm rates.

[0131] Optimize according to the historical alarm error and the alarm adjustment coefficient using the adaptive dynamic update method to obtain the optimized adaptive alarm threshold;

[0132] It should be noted that by analyzing the false alarm rate and the missed alarm rate, calculating the error correction value, and combining the false alarm correction factor, the user adjustment factor, and the environmental adaptation factor, the alarm threshold is dynamically adjusted to automatically optimize it in different environments, improve the alarm accuracy, reduce false alarms and missed alarms, and ensure the intelligent adaptability and user experience of the device.

[0133] According to the optimized adaptive alarm threshold, evaluate the alarm trigger status and generate an alarm processing command;

[0134] It should be noted that after receiving new sensor data, calculate the alarm index and compare it with the optimized alarm threshold. If the alarm index exceeds the threshold, the device determines that an abnormal situation exists, triggers an alarm, and generates a corresponding alarm processing command, such as activating an alarm, sending a notification to the user, or linking other security devices; if the alarm index is lower than the threshold, maintain the normal state to avoid false alarms. Through this mechanism, the device can automatically adjust the alarm strategy according to the actual situation, improving the reliability and intelligence of the alarm.

[0135] In summary, the present invention achieves the following: dynamically allocating the weights of multi-source sensors to improve the adaptability of the device and reduce false alarms caused by environmental changes between day and night; combining historical false alarm data analysis and an adaptive update algorithm to continuously optimize the alarm threshold, improve the alarm accuracy, and reduce the false alarm and missed alarm rates; at the same time, supporting users to customize the weight adjustment strategy to achieve personalized settings, making the device more in line with the needs of different users and enhancing the user experience.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. An Internet-based home intelligent alarm device, characterized in that: include, Data acquisition module, collects multi-source sensor data and generates sensor data packets; The feature data generation module generates multi-source feature data based on the sensor data packet by extracting features through low-pass filtering, short-time Fourier transform and inter-frame pixel difference; The alarm information generation module fuses multi-source feature data, calculates the alarm index using the weighted sum method, and generates alarm information; The control instruction module encrypts the alarm information, generates an encrypted data packet, sends it to the remote user end through the Internet, and receives the control instruction from the user end; The alarm command module dynamically adjusts the alarm strategy based on the control instructions from the user side and generates alarm processing commands.

2. The Internet-based home intelligent alarm device according to claim 1, characterized in that: The specific steps of collecting multi-source sensor data and generating sensor data packets are as follows: Through the PIR sensor, microphone sensor and camera, PIR sensor data, microphone data, audio signal, spectrum characteristics, image frame and image characteristics are collected to generate multi-source sensor data; The data from multi-source sensors are synchronized according to timestamps to generate sensor data packets.

3. The Internet-based home intelligent alarm device according to claim 2, characterized in that: The method generates multi-source feature data based on the sensing data packet by extracting features through low-pass filtering, short-time Fourier transform and inter-frame pixel difference. The specific steps are as follows: Based on the data of the PIR sensor, the high-frequency noise is removed by low-pass filtering to obtain the smoothed signal of the PIR sensor, and the change rate of the PIR sensor is obtained by first-order difference method. Based on the microphone sensor data, the audio signal is segmented using a sliding window method, and a short-time Fourier transform is performed on each frame of audio data to obtain the spectral characteristics of the audio signal; Based on the camera sensor data, the inter-frame difference method is used to calculate the pixel change value between the current frame and the previous frame to extract the moving target; According to the moving target, edge detection is performed using image processing to obtain the moving contour information; Multi-source feature data is generated based on the combination of the PIR sensor change rate, the spectral characteristics of the audio signal and the motion profile information detected by the camera.

4. The Internet-based home intelligent alarm device according to claim 3, characterized in that: According to the moving target, edge detection is performed using image processing to obtain motion contour information. The specific steps are as follows: According to the moving target, the noise is removed and the image is smoothed by Gaussian filtering; Use the Canny operator to calculate the gradient of the smoothed image and obtain the target edge image; Extract edge regions from target edge images and use double threshold edge detection to remove weak noise edges. Through edge tracking algorithm, the broken edges are connected; Perform connected region analysis on the processed edge image, mark all moving targets, and use the convex hull algorithm to fit the contour of the moving target; Count the number of pixels within the contour, estimate the size of the moving target, calculate the contour area, and obtain the motion contour information.

5. The Internet-based home intelligent alarm device according to claim 3, characterized in that: The multi-source feature data is fused, the alarm index is calculated using the weighted summation method, and the alarm information is generated. The specific steps are as follows: The multi-source feature data is normalized using a linear normalization method; According to environmental conditions, weights are assigned to multi-source sensor data and alarm thresholds are preset; The normalized PIR change rate, audio features, and moving target area are multiplied by the assigned weights to obtain weighted feature values, and the alarm index is obtained by weighted summation method; Based on the alarm index, a judgment is made through the preset alarm threshold to generate alarm information.

6. The Internet-based home intelligent alarm device according to claim 5, characterized in that: According to environmental conditions, the weights of multi-source sensor data are assigned and the alarm thresholds are preset. The specific steps are as follows: Through the environmental condition adaptive algorithm, the audio feature weight of the multi-source sensor night mode is increased and the visual feature weight is reduced, and the visual feature weight of the multi-source sensor day mode is increased and the audio feature weight is reduced; Based on the assigned multi-source sensor data weights combined with historical alarm data, statistical analysis is performed to set alarm thresholds.

7. The Internet-based home intelligent alarm device according to claim 5, characterized in that: Based on the alarm index, the preset alarm threshold is used to make a judgment and generate alarm information. The specific steps are as follows: When the alarm index is greater than the alarm threshold, it means that there is an abnormality in the current environment and an alarm is triggered; When the alarm index is less than the alarm threshold, it means that the current environment is normal and there is no need to trigger an alarm; Based on the triggered alarm status, the alarm event alarm index, alarm time and event location are recorded to generate alarm information.

8. The Internet-based home intelligent alarm device according to claim 7, characterized in that: The alarm information is encrypted to generate an encrypted data packet, which is sent to a remote user terminal via the Internet, and a control instruction from the user terminal is received. The specific steps are as follows: Based on the alarm information, multi-source data is encapsulated and encrypted using a symmetric encryption algorithm to generate an encrypted data packet; The encrypted data packets are sent to the remote client via the Internet, and the control instructions from the client are received via the MQTT protocol.

9. The Internet-based home intelligent alarm device according to claim 8, characterized in that: The control instructions based on the user end are used to dynamically adjust the alarm strategy and generate alarm processing commands. The specific steps are as follows: According to the deviation between the alarm index and the actual security incident, a historical alarm error data set is obtained; According to the historical alarm error data set and the alarm adjustment coefficient, the adaptive alarm threshold is calculated to obtain the optimized adaptive alarm threshold; According to the optimized adaptive alarm threshold, the alarm trigger status is evaluated and the alarm processing command is generated.

10. The Internet-based home intelligent alarm device according to claim 9, characterized in that: According to the historical alarm error data set and the alarm adjustment coefficient, the adaptive alarm threshold is calculated to obtain the optimized adaptive alarm threshold. The specific steps are as follows: Dynamically generate alarm adjustment coefficients by integrating historical false alarm analysis with user-manually set adjustment parameters; According to the historical alarm error and the alarm adjustment coefficient, the adaptive dynamic update method is used to optimize and obtain the optimized adaptive alarm threshold.

Images