Human body induction automatic switch of intelligent lighting system

Abstract

The invention discloses an intelligent lighting system human body induction automatic switch, and relates to the lighting system control technology field, the intelligent lighting system human body induction automatic switch comprises a learning module, an adjusting module and a database, the learning module is connected with the adjusting module and the database through signal transmission, and the learning module comprises a visual sensor, a radar sensor and a controller; the radar sensor is used for positioning the position of a user, the visual sensor is used for collecting user information, and the controller transmits the information to the database through signal transmission after receiving the information transmitted by the visual sensor and the radar sensor. By designing the learning module, the function of predicting user behaviors is achieved, the problems of poor environmental adaptability, poor user experience and energy waste are solved, illumination parameters can be dynamically adjusted, energy waste caused by global constant illumination is avoided, the application scene is widened, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of lighting system control, and particularly to a human body sensing automatic switch for an intelligent lighting system. Background Art

[0002] Under the global energy crisis and the "dual carbon" goal, the problem of energy waste caused by frequent manual operation of traditional lighting systems has become prominent. It is urgent to reduce ineffective energy consumption through automatic control. By using pyroelectric or radar technology to achieve precise control of "lights on when people are present, lights off when people leave", the lighting power consumption can be reduced by 30% - 50%. Most existing systems for intelligent lighting systems only rely on infrared sensors or radar sensors to obtain the human presence status, lack the collection of multi-modal data along the user's activities, and it is difficult to build an accurate behavior prediction model, lacking the prediction of user behavior, with relatively large limitations.

[0003] Patent CN104837239B discloses an induction switch and its control method. The above patent realizes the effect of turning off the operation of the human body sensing module to avoid the lighting of the lighting lamp when the user accidentally triggers the operation of the human body sensing module and affects the user's sleep, and re-opening the human body sensing module to trigger the lighting of the lighting lamp by sensing a specific change in the external environment brightness, which is convenient for users to use.

[0004] The above patent senses a specific change in the external environment brightness through a light sensing module. When the user is resting at night, it sends a control command to turn off the operation of the human body sensing module to avoid accidentally triggering the operation of the human body sensing module due to the user's body movements during the user's sleep, and re-opening the human body sensing module to trigger the lighting of the lighting lamp after the rest ends. There is room for optimization in predicting user behavior.

[0005] Therefore, this application proposes a human body sensing automatic switch for an intelligent lighting system that predicts user behavior. Summary of the Invention

[0006] The purpose of the present invention is to provide a human body sensing automatic switch for an intelligent lighting system to solve the technical problem of user behavior prediction proposed in the above background art.

[0007] To achieve the above purpose, the present invention provides the following technical solution: A human body sensing automatic switch for an intelligent lighting system, including a learning module, an adjustment module, and a database. The learning module is connected to the adjustment module and the database through signal transmission. The learning module includes a vision sensor, a radar sensor, and a controller. The radar sensor is used to locate the user's position, the vision sensor is used to collect user information, and after receiving the information transmitted by the vision sensor and the radar sensor, the controller transmits the information to the database through signal transmission. When the radar sensor detects the approach of a user, it collects user information through the vision sensor and transmits the information to the controller. The controller retrieves information from the database to confirm the user's identity, and collects the user's behavior information through the vision sensor and transmits the information to the controller. The controller predicts the subsequent user behavior based on the received user behavior information and controls the lighting system to provide lighting for the user.

[0008] Preferably, an LSTM neural network algorithm and an XGBoost integrated learning algorithm are set in the controller. The time series features extracted by the LSTM neural network algorithm are spliced with the statistical features constructed by the XGBoost integrated learning algorithm to form a high-dimensional feature space. The LSTM neural network algorithm is used to learn the time dependence of user operations and capture the periodic fluctuations of environmental parameters. The XGBoost integrated learning algorithm is used to construct a non-linear mapping between the user profile and environmental parameters and screen high-value feature combinations.

[0009] Preferably, the learning module is connected to the sensing module through signal transmission. The sensing module includes: an infrared sensor, a photosensitive sensor, a temperature sensor, and a humidity sensor. The infrared sensor is used to detect the user. When the user appears, it transmits the information to the controller, and the controller controls the lighting system to give preliminary lighting. The photosensitive sensor is used to detect the ambient light intensity and transmits the information to the controller. The controller compares the light information with the set parameters in the database. When the ambient light intensity is not within the set parameter range, the controller controls the lighting system not to give lighting. The temperature sensor is used to detect the ambient temperature and transmits the information to the controller. The controller compares the temperature data with the set parameters in the database. When the actual temperature is not within the set parameters, it transmits the temperature information to the compensation module. The humidity sensor is used to detect the ambient humidity and transmits the information to the controller. The controller compares the humidity data with the set parameters in the database. When the actual humidity is not within the set parameters, it transmits the humidity information to the compensation module.

[0010] Preferably, the sensing module is connected to the compensation module through signal transmission. The compensation module includes: a temperature compensation unit, a humidity compensation unit, and a light compensation unit. The temperature compensation unit is connected to the temperature sensor through signal transmission. After the controller transmits the compensation information to the temperature compensation unit, the temperature compensation unit compensates the ambient temperature according to the temperature information. The humidity compensation unit is connected to the humidity sensor through signal transmission. After the controller transmits the compensation information to the humidity compensation unit, the humidity compensation unit compensates the ambient humidity according to the humidity information. The light compensation unit is connected to the photosensitive sensor through signal transmission. After receiving the compensation information, it compensates the threshold of the photosensitive sensor.

[0011] Preferably, the adjustment module is connected to the database through signal transmission. The adjustment module includes a touch screen, a microphone, and a Bluetooth unit. The touch screen is used for users to adjust lighting parameters. The microphone is used to receive user instructions to adjust lighting parameters. The Bluetooth unit is used for remote control to adjust lighting parameters.

[0012] Preferably, after the visual sensor and the radar sensor collect user behavior information, they extract user behavior characteristics and transmit them to the controller. The input layer of the LSTM neural network algorithm receives the information transmitted in the past 30 minutes and outputs the predicted probability of user behavior in the next 5 minutes. The XGBoost ensemble learning algorithm dynamically generates an adjustment decision tree based on feature importance. When the confidence level of the behavior prediction is greater than 85%, the lighting parameter pre-adjustment is triggered 300 ms in advance.

[0013] Preferably, after the photosensitive sensor detects that the ambient light intensity is lower than the preset threshold, it detects the human body through the infrared sensor to trigger the adjustment of lighting parameters.

[0014] Preferably, when the infrared sensor is mis-triggered and the radar sensor does not detect user movement, the visual sensor collects information about the user, and the controller then adjusts the lighting parameters.

[0015] Preferably, after the database receives the adjustment of lighting parameters by the user through the adjustment module, it transmits the information to the controller, converts the user adjustment operation sequence into a time series feature vector, represents the user's habits through the hidden state h_t of the LSTM neural network algorithm, and extracts personalized parameters in the XGBoost features to construct a user profile.

[0016] Preferably, the microphone extracts the user's voice feature by collecting the user's voice command information and transmits the information to the controller. The controller extracts the voice embedding vector through the LSTM neural network algorithm and inputs it into the XGBoost ensemble learning algorithm for user identity selection. The controller adjusts the personalized lighting parameters according to the user's identity.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. By designing a learning module, the present invention realizes the function of predicting user behavior, solves the problems of poor environmental adaptability, poor user experience, and energy waste, can dynamically adjust lighting parameters, avoids energy waste caused by full-domain constant illuminance, broadens the application scenarios, and improves the user experience; 2. The present invention is equipped with a sensing module and a learning module, achieving the function of obtaining environmental information, solving the problems of inaccurate information collection and high energy consumption, being able to reduce the false trigger rate, reduce energy consumption, improve the accuracy of information acquisition, and enhance the user experience; 3. The present invention is equipped with a compensation module, achieving the function of eliminating environmental interference, solving the problems of induction failure, device false trigger, abnormal energy consumption, and poor user experience, being able to maintain stable environmental parameters, reduce the false trigger rate, improve the reliability of the switch, and optimize the user experience; 4. The present invention is designed with an adjustment module, achieving the function of users adjusting lighting parameters, solving the problems of conflict between personalized needs and automation, energy waste, and low reliability, being able to balance automation efficiency and user autonomy, reduce false trigger interference, and adapt to complex environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a schematic diagram of the composition of the human body induction automatic switch of the present invention; Figure 2 is a schematic diagram of the composition of the learning module of the present invention; Figure 3 is a schematic diagram of the composition of the sensing module of the present invention; Figure 4 is a schematic diagram of the composition of the compensation module of the present invention; Figure 5 is a schematic diagram of the composition of the adjustment module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] Embodiment 1: Please refer to Figure 1 and Figure 2 , an intelligent lighting system human body induction automatic switch, including a learning module, an adjustment module, and a database. The learning module is connected to the adjustment module and the database through signal transmission; The learning module includes a visual sensor, a radar sensor, and a controller. The radar sensor is used to locate the user's position, the visual sensor is used to collect user information, and after receiving the information transmitted by the visual sensor and the radar sensor, the controller transmits the information to the database through signal transmission; When the radar sensor detects the approach of a user, the visual sensor collects user information and transmits it to the controller. The controller retrieves database information to confirm the user's identity. Based on the user behavior information collected by the visual sensor, it transmits the information to the controller. The controller predicts the subsequent user behavior according to the received user behavior information and controls the lighting system to provide lighting for the user. The LSTM neural network algorithm and the XGBoost integrated learning algorithm are set in the controller. The temporal features extracted by the LSTM neural network algorithm are concatenated with the statistical features constructed by the XGBoost integrated learning algorithm to form a high-dimensional feature space. The LSTM neural network algorithm is used to learn the time dependence of user operations and capture the periodic fluctuations of environmental parameters. The XGBoost integrated learning algorithm is used to construct the non-linear mapping between the user profile and environmental parameters and screen high-value feature combinations. After the visual sensor and the radar sensor collect the user behavior information, they extract the user behavior features and transmit them to the controller. The input layer of the LSTM neural network algorithm receives the information transmitted in the past 30 minutes and outputs the prediction probability of the user behavior in the next 5 minutes. The XGBoost integrated learning algorithm dynamically generates adjusted decision trees based on feature importance. When the confidence level of the behavior prediction is greater than 85%, the lighting parameter pre-adjustment is triggered 300 ms in advance. Further, the vision sensor is used to capture the user's gestures and micro-movements, and the radar sensor obtains the user's spatial coordinates and moving speed in real time. The input layer of the LSTM neural network algorithm inputs the data transmitted by the vision sensor and the radar sensor in the past 30 minutes. The hidden layer is a bidirectional LSTM and Dropout to capture long-term dependencies. The output layer is the probability distribution of the user's behavior in the next 5 minutes, including moving target areas or staying intentions, etc. An LSTM time series model is established. The XGBoost decision engine generates a final adjustment instruction by inputting the user behavior probability vector output by the LSTM, the environmental light intensity change detected by the photosensitive sensor, and the time period feature encoding, and adjusts the lighting parameters through the controller. When the behavior prediction confidence is greater than 85%, the lighting parameter pre-adjustment is triggered 300 ms in advance. Among them, the short-term dynamic prediction is dominated by the LSTM neural network algorithm, and the XGBoost ensemble learning algorithm synthesizes environmental parameters to generate the final adjustment instruction. At the same time, the vision sensor can identify different users, so different user models are constructed in the user model constructed by the controller. Before behavior prediction, the vision sensor collects the user's characteristic information and transmits the information to the controller. The controller compares the received user characteristic information with the user characteristic information stored in the database to confirm the user's identity, matches the LSTM time series storage unit corresponding to the user identity, isolates the influence of different user individual data, and the XGBoost reconstructs the user-specific subtree every day, eliminating outdated decision rules such as branches that have not been triggered for 7 consecutive days. When the amount of new user data is insufficient, through similarity matching, the parameters of a similar user model are borrowed for initialization and adjusted according to the behavior characteristics of the new user, quickly constructing the personalized model of the new user, realizing the function of predicting the user's behavior, solving the problems of poor environmental adaptability, poor user experience and energy waste, being able to dynamically adjust the lighting parameters, avoiding energy waste caused by constant illuminance in the whole area, broadening the application scenarios, and improving the user experience.

[0021] Embodiment 2: Please refer to Figure 1 and Figure 3 , an intelligent lighting system human body induction automatic switch, the learning module is connected to the sensing module through signal transmission, and the sensing module includes: an infrared sensor, a photosensitive sensor, a temperature sensor and a humidity sensor. The infrared sensor is used to detect the user. When the user appears, the information is transmitted to the controller, and the controller controls the lighting system to initially provide lighting; The photosensitive sensor is used to detect the environmental light intensity, transmit the information to the controller, and the controller compares the light information with the set parameters in the database. When the environmental light intensity is not within the set parameter range, the controller controls the lighting system not to provide lighting; The temperature sensor is used to detect the ambient temperature and transmit the information to the controller. The controller compares the temperature data with the parameters set in the database. When the actual temperature is not within the set parameters, the temperature information is transmitted to the compensation module; The humidity sensor is used to detect the ambient humidity and transmit the information to the controller. The controller compares the humidity data with the parameters set in the database. When the actual humidity is not within the set parameters, the humidity information is transmitted to the compensation module; After the photosensitive sensor detects that the ambient light intensity is lower than the preset threshold, the infrared sensor is used to detect the human body to trigger the adjustment of lighting parameters; When the infrared sensor is mis-triggered and the radar sensor does not detect the movement of the user, the visual sensor is used to collect information about the user, and the controller then adjusts the lighting parameters; Furthermore, the photosensitive sensor perceives the ambient light intensity in real time and transmits it to the controller. The controller determines whether to activate the lighting system according to the preset brightness threshold. When indoors, the threshold of the photosensitive sensor is such that when the photosensitive sensor detects that the ambient light is higher than the set threshold and the infrared sensor detects the presence of a user, the controller does not trigger the lighting system to provide lighting. When the ambient light intensity is lower than the preset threshold and the infrared sensor detects the activity of a user, the controller activates the lighting system to provide lighting for the user. However, using only the infrared sensor to assist the photosensitive sensor for lighting condition triggering is too restrictive. When the ambient temperature is close to the human body temperature, it is difficult for the infrared sensor to distinguish the difference between human body heat radiation and ambient heat sources, which easily leads to mis-triggering or missed detection. Moreover, the infrared sensor can only detect moving humans and cannot continuously detect the presence when the user is stationary, such as reading or sleeping, resulting in abnormal lighting. When the conditions for providing lighting are met under the detection of the photosensitive sensor and the infrared sensor, the visual sensor is used to collect the user's behavior information, the radar sensor is used to confirm the user's position information. After providing lighting, the radar sensor scans the space in real time to make up for the problem that the infrared sensor cannot recognize stationary humans and avoid mis-turning off the lights when the user is stationary. The visual sensor can collect the user's behavior information, can obtain the user's behavior information when the light is sufficient, and the radar sensor obtains the user's behavior information in a low light intensity scene, realizing the function of obtaining environmental information, solving the problems of inaccurate information collection and high energy consumption, being able to reduce the mis-triggering rate, reducing energy consumption, improving the accuracy of information acquisition, and enhancing the user experience.

[0022] Example 3: Please refer to Figure 1 、 Figure 3 and Figure 4, an intelligent lighting system body sensing automatic switch, the learning module is connected to the sensing module through signal transmission. The sensing module includes: an infrared sensor, a photosensitive sensor, a temperature sensor, and a humidity sensor. The infrared sensor is used to detect users. When a user is detected, the information is transmitted to the controller, and the controller controls the lighting system to initially provide lighting; The photosensitive sensor is used to detect the ambient light intensity and transmit the information to the controller. The controller compares the light information with the set parameters in the database. When the ambient light intensity is not within the set parameter range, the controller controls the lighting system not to provide lighting; The temperature sensor is used to detect the ambient temperature and transmit the information to the controller. The controller compares the temperature data with the set parameters in the database. When the actual temperature is not within the set parameters, the temperature information is transmitted to the compensation module; The humidity sensor is used to detect the ambient humidity and transmit the information to the controller. The controller compares the humidity data with the set parameters in the database. When the actual humidity is not within the set parameters, the humidity information is transmitted to the compensation module; The sensing module is connected to the compensation module through signal transmission. The compensation module includes: a temperature compensation unit, a humidity compensation unit, and a light compensation unit; The temperature compensation unit is connected to the temperature sensor through signal transmission. After the controller transmits the compensation information to the temperature compensation unit, the temperature compensation unit compensates the ambient temperature according to the temperature information; The humidity compensation unit is connected to the humidity sensor through signal transmission. After the controller transmits the compensation information to the humidity compensation unit, the humidity compensation unit compensates the ambient humidity according to the humidity information; The light compensation unit is connected to the photosensitive sensor through signal transmission. After receiving the compensation information, it compensates the threshold of the photosensitive sensor; Further, temperature sensors and humidity sensors are used to detect the ambient temperature and humidity. After transmitting the information to the controller, the controller determines whether compensation is required based on the preset temperature range of 0°C to 40°C and humidity range of 20% to 85%. When the temperature exceeds the set range, the detection performance of the photosensitive sensor, vision sensor, and radar sensor decreases. More seriously, under high-temperature interference, the infrared sensor has a significantly increased missed detection rate for human movement, and there will be a situation where no one passes by the infrared sensor but a signal indicating someone passing by is transmitted, resulting in wasted energy. When the temperature and humidity are outside the set range, the detection accuracy of the sensors decreases. Therefore, temperature and humidity compensation are required. When the temperature or humidity is abnormal, the temperature compensation unit and humidity compensation unit compensate for the ambient temperature and ambient humidity to ensure that it is within the normal operating range of the sensors, rather than adjusting the detection parameters of the sensors from within the sensors, avoiding the impact of the environment on the service life of the sensors, reducing the cost of maintaining the sensors. In different weather or seasons, the natural light intensity is different, which affects the detection results of the photosensitive sensor. The controller obtains the current weather or season situation through the database, obtains the weather data and combines the temperature and humidity data transmitted by the temperature sensor and humidity sensor, and adjusts the gain parameters of the photosensitive sensor through the light compensation unit to reduce the detection error, realizing the function of eliminating environmental interference, solving the problems of induction failure, device mis-triggering, abnormal energy consumption, and poor user experience, being able to maintain stable environmental parameters, reducing the mis-triggering rate, improving the reliability of the switch, and optimizing the user experience.

[0023] Embodiment 4: Please refer to Figure 1 、 Figure 2 and Figure 5 , an intelligent lighting system human body induction automatic switch, including a learning module, an adjustment module, and a database. The learning module is connected to the adjustment module and the database through signal transmission; The learning module includes a vision sensor, a radar sensor, and a controller. The radar sensor is used to locate the user's position, the vision sensor is used to collect user information, and after receiving the information transmitted by the vision sensor and radar sensor, the controller transmits the information to the database through signal transmission; When the radar sensor detects the approach of a user, it collects user information through the vision sensor and transmits the information to the controller. The controller retrieves the database information to confirm the user's identity, and based on the user behavior information collected by the vision sensor, transmits the information to the controller. The controller predicts the subsequent user behavior based on the received user behavior information and controls the lighting system to provide lighting for the user according to the prediction result; The adjustment module is connected to the database through signal transmission. The adjustment module includes a touch screen, a microphone, and a Bluetooth unit. The touch screen is used for users to adjust lighting parameters. The microphone is used to receive user instructions to adjust lighting parameters. The Bluetooth unit is used for remote control to adjust lighting parameters; After the database receives the adjustment of lighting parameters by the user through the adjustment module, it transmits the information to the controller, converts the user adjustment operation sequence into a time series feature vector, represents the user's habits through the hidden state h_t of the LSTM neural network algorithm, and extracts personalized parameters in the XGBoost features to construct a user profile; The microphone collects user voice command information, extracts the user's voice features, and transmits the information to the controller. The controller extracts the voice embedding vector through the LSTM neural network algorithm and inputs it into the XGBoost ensemble learning algorithm for user identity selection. The controller adjusts the personalized lighting parameters according to the user's identity; Furthermore, after the controller adjusts the lighting parameters provided by the lighting system according to the perception module, the learning module, and the adjustment module, the user can still adjust the lighting parameters through the touch screen. After the user completes the adjustment through the touch screen, the visual sensor and the radar sensor record the user's subsequent behavior information, that is, what the user is doing after adjusting the lighting parameters, convert the user adjustment operation sequence into a time series feature vector, represent the user's habits through the hidden state h_t of the LSTM neural network algorithm, extract personalized parameters in the XGBoost features to construct a user profile. When the user adjusts the lighting parameters through the microphone, the controller collects the user's voice information, matches it with the user voice model established in the database, and dynamically updates the voice model to adapt to environmental changes. After extracting the user's voice features, the information is transmitted to the controller. The controller extracts the voice embedding vector through the LSTM neural network algorithm and inputs it into the XGBoost ensemble learning algorithm for user identity selection. The controller adjusts the personalized lighting parameters according to the user's identity. When the user remotely controls through the Bluetooth unit, if in a large space, a relay module can be arranged or switched to Zigbee / Wi-Fi hybrid networking. The controller adjusts the lighting parameters according to the user through the mobile terminal, classifies the user's mobile terminal signal, captures the time series dependence relationship through the LSTM neural network algorithm, dynamically updates the user preference model, constructs a user personality model according to the adjustment of lighting parameters by the user through the adjustment module, and periodically updates the user personality model, realizing the function of the user to adjust the lighting parameters, solving the problems of conflict between personalized needs and automation, energy waste, and low reliability, being able to balance the automation efficiency and user autonomy, reducing mis-trigger interference, and adapting to complex environmental conditions.

[0024] Example 5: Please refer to Figure 1 、 Figure 2 、 Figure 3 andFigure 5 , an intelligent lighting system human body induction automatic switch, comprising a learning module, an adjustment module and a database, wherein the learning module is connected to the adjustment module and the database through signal transmission; The learning module includes a vision sensor, a radar sensor and a controller. The radar sensor is used to locate the user's position, the vision sensor is used to collect user information, and after receiving the information transmitted by the vision sensor and the radar sensor, the controller transmits the information to the database through signal transmission; When the radar sensor detects the approach of a user, it collects user information through the vision sensor and transmits the information to the controller. The controller retrieves the database information to confirm the user's identity, and through the user behavior information collected by the vision sensor, it transmits the information to the controller. The controller predicts the subsequent user behavior based on the received user behavior information and controls the lighting system to provide lighting for the user according to the prediction result; The learning module is connected to the sensing module through signal transmission. The sensing module includes: an infrared sensor, a photosensitive sensor, a temperature sensor and a humidity sensor. The infrared sensor is used to detect the user. When the user appears, it transmits the information to the controller, and the controller controls the lighting system to give preliminary lighting; The photosensitive sensor is used to detect the ambient light intensity and transmits the information to the controller. The controller compares the light information with the set parameters in the database. When the ambient light intensity is not within the set parameter range, the controller controls the lighting system not to give lighting; The temperature sensor is used to detect the ambient temperature and transmits the information to the controller. The controller compares the temperature data with the set parameters in the database. When the actual temperature is not within the set parameters, it transmits the temperature information to the compensation module; The humidity sensor is used to detect the ambient humidity and transmits the information to the controller. The controller compares the humidity data with the set parameters in the database. When the actual humidity is not within the set parameters, it transmits the humidity information to the compensation module; The adjustment module is connected to the database through signal transmission. The adjustment module includes a touch screen, a microphone and a Bluetooth unit. The touch screen is used for the user to adjust the lighting parameters, the microphone is used to receive user instructions to adjust the lighting parameters, and the Bluetooth unit is used for remote control to adjust the lighting parameters; Furthermore, there are three ways to adjust the lighting parameters. The perception module senses the environment and automatically makes adjustments. The learning module collects user behavior information for prediction and adjustment. The user makes adjustments through the adjustment module. Among them, the first to be triggered is the perception module to adjust the lighting conditions by sensing the environment. Subsequently, the learning module is triggered to collect user behavior information for prediction and adjustment. Finally, the user makes adjustments through the adjustment module. As time goes by, after the learning module constructs a user personalized model, the detection of environmental conditions by the perception module becomes a prerequisite for triggering the learning module to make prediction adjustments. That is, when the perception module detects that the environmental light intensity meets the condition for enabling lighting, the infrared sensor is used to assist the vision sensor and radar sensor in the learning module to detect the presence of the user. After detecting the presence of the user, the vision sensor and radar sensor are used to obtain user behavior information, and user behavior prediction is performed. The data in the user personalized model is called to provide lighting for the user. After completing the lighting parameter adjustment, if there is a user who adjusts the lighting parameters through the adjustment module, the learning module updates the parameters in the user personalized model, realizing the function of continuously improving the user personalized model, solving the problems of prediction failure, performance degradation, increased maintenance costs, and poor user experience, being able to optimize the user personalized lighting parameters, reducing energy consumption, and improving the user experience.

[0025] Working principle: The photosensitive sensor perceives the ambient light intensity in real time and transmits it to the controller. The controller determines whether to activate the lighting system according to the preset brightness threshold. When indoors, for the threshold of the photosensitive sensor, when the photosensitive sensor detects that the ambient light is higher than the set threshold and the infrared sensor detects the presence of a user, the controller does not trigger the lighting system to provide lighting. When the ambient light intensity is lower than the preset threshold and the infrared sensor detects user activity, the controller activates the lighting system to provide lighting for the user. However, using only the infrared sensor to assist the photosensitive sensor in triggering lighting conditions is too restrictive. When the ambient temperature is close to the human body temperature, it is difficult for the infrared sensor to distinguish the difference between human body heat radiation and ambient heat sources, easily leading to false triggering or missed detection. Moreover, the infrared sensor can only detect moving humans and cannot continuously detect the presence when the user is stationary, such as reading or sleeping, resulting in abnormal lighting. When the conditions for providing lighting are met under the detection of the photosensitive sensor and the infrared sensor, the visual sensor is used to collect user behavior information, and the radar sensor is used to confirm the user's position information. After providing lighting, the radar sensor scans the space in real time to make up for the problem that the infrared sensor cannot recognize stationary humans and avoid mis-turning off the lights when the user is stationary. The visual sensor can collect user behavior information, obtain user behavior information when the light is sufficient, and obtain user behavior information by the radar sensor in a low light intensity scenario. The temperature sensor and the humidity sensor are used to detect the ambient temperature and humidity. After transmitting the information to the controller, the controller judges whether compensation is needed according to the preset temperature range of 0°C to 40°C and humidity range of 20% to 85%. When the temperature and humidity are not within the set range, the detection accuracy of the sensor decreases. Therefore, temperature and humidity compensation are required. When the temperature or humidity is abnormal, the temperature compensation unit and the humidity compensation unit compensate the ambient temperature and humidity to ensure that the sensor is within the normal working range, rather than adjusting the detection parameters of the sensor from within the sensor, avoiding the impact of the environment on the service life of the sensor and reducing the cost of maintaining the sensor. In different weather or seasons, the natural light intensity is different, which affects the detection results of the photosensitive sensor. The controller obtains the current weather or season situation through the database, obtains the weather data combined with the temperature and humidity data transmitted by the temperature sensor and the humidity sensor, and adjusts the gain parameters of the photosensitive sensor through the light compensation unit to reduce the detection error; The vision sensor is used to capture the user's posture and micro-actions, and the radar sensor obtains the user's spatial coordinates and moving speed in real time. The input layer of the LSTM neural network algorithm inputs the data transmitted by the vision sensor and the radar sensor in the past 30 minutes. The hidden layer is a bidirectional LSTM and Dropout to capture long-term dependencies. The output layer is the probability distribution of the user's behavior in the next 5 minutes, including the moving target area or the intention to stay, etc., for LSTM time series modeling. The XGBoost decision engine generates the final adjustment instruction by inputting the user behavior probability vector output by the LSTM, the change in ambient light intensity detected by the photosensitive sensor, and the time period feature encoding, and adjusts the lighting parameters through the controller. When the behavior prediction confidence is greater than 85%, the lighting parameter pre-adjustment is triggered 300 ms in advance. Among them, the LSTM neural network algorithm dominates the short-term dynamic prediction, and the XGBoost ensemble learning algorithm synthesizes the environmental parameters to generate the final adjustment instruction. At the same time, the vision sensor can identify different users, so in the user model constructed by the controller, different user models are constructed. Before behavior prediction, the vision sensor collects the user's feature information and transmits the information to the controller. The controller compares the received user feature information with the user feature information stored in the database to confirm the user identity, matches the LSTM time series storage unit corresponding to the user identity, isolates the influence of different user individual data, and the XGBoost reconstructs the user-specific sub-tree daily, eliminating the outdated decision rules such as the branches that have not been triggered for 7 consecutive days. When the new user data volume is insufficient, through similarity matching, the parameters of the similar user model are borrowed for initialization and adjusted according to the behavior characteristics of the new user to quickly construct the personalized model of the new user; After the controller adjusts the lighting parameters provided by the lighting system according to the perception module, learning module, and adjustment module, the user can still adjust the lighting parameters through the touch screen. After the user completes the adjustment through the touch screen, the visual sensor and radar sensor record the user's subsequent behavior information, that is, what the user is doing after adjusting the lighting parameters. The user adjustment operation sequence is converted into a time series feature vector, and the hidden state h_t of the LSTM neural network algorithm represents the user's habits. The personalized parameters in the XGBoost features are extracted to construct a user profile. When the user adjusts the lighting parameters through the microphone, the controller collects the user's voice information, matches it with the user voice model established in the database, and dynamically updates the voice model to adapt to environmental changes. After extracting the user voice features, the information is transmitted to the controller. The controller extracts the voice embedding vector through the LSTM neural network algorithm and inputs it into the XGBoost ensemble learning algorithm for user identity decision-making. The controller adjusts the personalized lighting parameters according to the user identity. When the user remotely controls through the Bluetooth unit, if in a large space, a relay module can be arranged or switched to Zigbee / Wi-Fi hybrid networking. The controller classifies the user's mobile signal according to the lighting parameter adjustment by the user through the mobile terminal, captures the time series dependence relationship through the LSTM neural network algorithm, dynamically updates the user preference model, constructs a user personality model according to the adjustment of the lighting parameters by the user through the adjustment module, and periodically updates the user personality model; There are three ways to adjust the lighting parameters. The perception module senses the environment and automatically adjusts. The learning module collects the user's behavior information for prediction and adjustment. The user adjusts through the adjustment module. Among them, the first to be triggered is the perception module to adjust the lighting conditions by sensing the environment. Subsequently, the learning module is triggered to collect the user's behavior information for prediction and adjustment. Finally, the user adjusts through the adjustment module. As time goes by, after the learning module constructs the user's personalized model, the detection of the environmental conditions by the perception module becomes a prerequisite for triggering the learning module to perform predictive adjustment. That is, when the perception module detects that the environmental light intensity meets the lighting activation condition, the infrared sensor assists the visual sensor and radar sensor in the learning module to detect the appearance of the user. After detecting the user's appearance, the visual sensor and radar sensor obtain the user's behavior information and perform user behavior prediction. The data in the user's personalized model is called to provide lighting for the user. After completing the lighting parameter adjustment, if there is a user who adjusts the lighting parameters through the adjustment module, the learning module updates the parameters in the user's personalized model.

[0026] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in all respects, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Accordingly, all changes that fall within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. An intelligent lighting system with a human body induction automatic switch, characterized in that: It includes a learning module, an adjustment module and a database. The learning module is connected to the adjustment module and the database through signal transmission; The learning module includes a vision sensor, a radar sensor and a controller. The radar sensor is used to locate the user's position. The vision sensor is used to collect user information. After receiving the information transmitted by the vision sensor and the radar sensor, the controller transmits the information to the database through signal transmission; When the radar sensor detects the approach of the user, it collects user information through the vision sensor and transmits the information to the controller. The controller retrieves the database information to confirm the user's identity. Through the user behavior information collected by the vision sensor, the information is transmitted to the controller. The controller predicts the subsequent user behavior based on the received user behavior information and controls the lighting system to provide lighting for the user.

2. The automatic human body induction switch of an intelligent lighting system according to claim 1, characterized in that: The LSTM neural network algorithm and the XGBoost integrated learning algorithm are set in the controller. The time series features extracted by the LSTM neural network algorithm are spliced with the statistical features constructed by the XGBoost integrated learning algorithm to form a high-dimensional feature space; The LSTM neural network algorithm is used to learn the time dependence of user operations and capture the periodic fluctuations of environmental parameters; The XGBoost integrated learning algorithm is used to construct the non-linear mapping between the user portrait and environmental parameters and screen high-value feature combinations.

3. The automatic human body induction switch of an intelligent lighting system according to claim 1, characterized in that: The learning module is connected to the sensing module through signal transmission. The sensing module includes: an infrared sensor, a photosensitive sensor, a temperature sensor and a humidity sensor. The infrared sensor is used to detect the user. When the user appears, it transmits the information to the controller, and the controller controls the lighting system to give preliminary lighting; The photosensitive sensor is used to detect the environmental light intensity and transmit the information to the controller. The controller compares the light information with the set parameters in the database. When the environmental light intensity is not within the set parameter range, the controller controls the lighting system not to give lighting; The temperature sensor is used to detect the environmental temperature and transmit the information to the controller. The controller compares the temperature data with the set parameters in the database. When the actual temperature is not within the set parameters, it transmits the temperature information to the compensation module; The humidity sensor is used to detect the environmental humidity and transmit the information to the controller. The controller compares the humidity data with the set parameters in the database. When the actual humidity is not within the set parameters, it transmits the humidity information to the compensation module.

4. An automatic human body induction switch for an intelligent lighting system according to claim 3, characterized in that: The sensing module is connected to the compensation module through signal transmission. The compensation module includes: a temperature compensation unit, a humidity compensation unit and a light compensation unit; The temperature compensation unit is connected to the temperature sensor through signal transmission. After the controller transmits the compensation information to the temperature compensation unit, the temperature compensation unit compensates the environmental temperature according to the temperature information; The humidity compensation unit is connected to the humidity sensor through signal transmission. After the controller transmits the compensation information to the humidity compensation unit, the humidity compensation unit compensates the environmental humidity according to the humidity information; The light compensation unit is connected to the photosensitive sensor through signal transmission. After receiving the compensation information, it compensates the threshold of the photosensitive sensor.

5. An automatic human body induction switch for an intelligent lighting system according to claim 1, characterized in that: The adjustment module is connected to the database through signal transmission. The adjustment module includes a touch screen, a microphone, and a Bluetooth unit. The touch screen is used for users to adjust lighting parameters. The microphone is used to receive user instructions to adjust lighting parameters. The Bluetooth unit is used for remote control to adjust lighting parameters.

6. The automatic human body induction switch of an intelligent lighting system according to claim 1, wherein: After the visual sensor and the radar sensor collect user behavior information, user behavior characteristics are extracted and transmitted to the controller. The input layer of the LSTM neural network algorithm receives the information transmitted in the past 30 minutes and outputs the predicted probability of user behavior in the next 5 minutes. The XGBoost integrated learning algorithm dynamically generates an adjustment decision tree based on feature importance. When the confidence level of behavior prediction is greater than 85%, the lighting parameter pre-adjustment is triggered 300 ms in advance.

7. An automatic human body sensing switch for an intelligent lighting system according to claim 3, characterized in that: After the photosensitive sensor detects that the ambient light intensity is lower than the preset threshold, the infrared sensor is used to detect the human body to trigger the adjustment of lighting parameters.

8. An automatic human body induction switch for an intelligent lighting system according to claim 7, characterized in that: When the infrared sensor is mis-triggered and the radar sensor does not detect user movement, the visual sensor is used to collect user information, and the controller then adjusts the lighting parameters.

9. An automatic human body induction switch for an intelligent lighting system according to claim 5, characterized in that: After the database receives the adjustment of lighting parameters by the user through the adjustment module, the information is transmitted to the controller. The user adjustment operation sequence is converted into a time series feature vector. The hidden state h_t of the LSTM neural network algorithm is used to represent user habits, and personalized parameters in the XGBoost features are extracted to construct a user profile.

10. The automatic human body induction switch of an intelligent lighting system according to claim 5, characterized in that: The microphone extracts user voice feature information by collecting user voice command information and transmits the information to the controller. The controller extracts the voice embedding vector through the LSTM neural network algorithm and inputs it into the XGBoost integrated learning algorithm for user identity selection. The controller performs personalized lighting parameter adjustment according to the user identity.

Images