Intelligent lamp control method and system based on radar induction

Abstract

The invention relates to the technical field of intelligent lamp control, and discloses an intelligent lamp control method and system based on radar induction. The method comprises the following steps: dividing a radar detection range and setting regional parameters according to user position information; analyzing the noise of each region, identifying an interference source and correcting the position; establishing a behavior sequence, and predicting user behaviors in combination with a Markov model; and matching a lighting scene, and sending a control instruction to the lamp through the serial interface. According to the invention, static human bodies can be accurately identified, environmental interference can be effectively suppressed, user behaviors can be intelligently predicted, illumination parameters can be automatically adjusted, and the intelligent degree, user experience and energy utilization efficiency of the illumination system are improved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent control of lamps, and in particular to a method and system for intelligent control of lamps based on radar sensing. Background Art

[0002] With the rapid development of smart home technology, lighting control systems have gradually evolved from traditional manual switch mode to intelligent sensing control mode. At present, the intelligent lighting control systems on the market mainly use infrared sensing, sound sensing or light sensing technologies to achieve automatic control. Among them, infrared sensing technology is widely used in the field of lighting control because of its simple implementation and low cost. This type of system usually triggers the switch of lamps by detecting the change of heat radiated by human body in the area, and combines light sensors to judge the light conditions. Another common type of intelligent lighting solution uses cameras for image recognition and controls lamps by analyzing the movement of human body in video images. In addition, with the popularization of Internet of Things technology, remote control solutions based on smartphone APP or voice assistants have gradually become an important part of intelligent lighting. Users can turn on and off lighting equipment, adjust brightness and switch scenes through mobile phones or voice commands.

[0003] However, the existing intelligent lighting control technology still has some obvious shortcomings. First, the infrared sensing-based system can only detect the movement changes of the human body. It is difficult to effectively identify the human body that is stationary or moving slightly, which can easily lead to the problem of the light being mistakenly turned off when the person is still in the scene. Secondly, although the camera's image-based recognition method can improve the accuracy of personnel detection, there are serious privacy issues. At the same time, the recognition effect is greatly reduced at night or in low-light environments. In addition, the existing intelligent lighting systems often lack the ability to learn and predict user behavior patterns, and cannot automatically adjust lighting parameters according to user habits and preferences. They can only execute pre-set simple logic and cannot truly realize intelligent and personalized lighting control. In complex environments, existing technologies are also difficult to effectively distinguish between human targets and environmental interference sources. They are easily disturbed by factors such as electrical appliances and curtain swings, causing false triggering, affecting user experience and energy efficiency. Summary of the invention

[0004] The present application provides a method and system for intelligent control of lamps based on radar sensing, which can accurately identify stationary human bodies, effectively suppress environmental interference, intelligently predict user behavior and automatically adjust lighting parameters, thereby improving the intelligence level of the lighting system, user experience and energy utilization efficiency.

[0005] In the first aspect, the present application provides a method for intelligent control of lamps based on radar sensing, and the method for intelligent control of lamps based on radar sensing includes: identifying the position information when the radio frequency radar signal of the lamp is generated, and the position information includes the user's first position; dividing the radar detection range into multiple areas according to the user's first position, and setting parameters for each distance partition after the division to obtain differentiated area control parameters; based on the differentiated area control parameters, performing noise analysis on each distance partition to obtain an interference source identification parameter set, and correcting the user's first position based on the interference source identification parameter set to obtain the user's second position; by constructing a human behavior sequence and combining a Markov prediction model, performing a behavior pattern analysis on the user's second position to obtain a user behavior pattern library and a behavior prediction result; according to the user behavior pattern library and the behavior prediction result, matching the scene pattern from the lighting scene model library, and sending a control instruction to the lamp control unit through a serial communication interface.

[0006] In a second aspect, the present application provides a radar-sensing-based intelligent control system for lamps, the radar-sensing-based intelligent control system for lamps comprising: An extraction module, used for identifying the position information of the radio frequency radar signal of the lamp, wherein the position information includes the first position of the user; A division module, used to divide the radar detection range into multiple areas according to the first position of the user, and set parameters for each distance partition after the division to obtain differentiated area control parameters; An analysis module, configured to perform noise analysis on each distance partition based on the differentiated area control parameter to obtain an interference source identification parameter set, and to correct the user's first position based on the interference source identification parameter set to obtain the user's second position; A construction module is used to analyze the behavior pattern of the user's second position by constructing a human behavior sequence and combining it with a Markov prediction model to obtain a user behavior pattern library and a behavior prediction result; The matching module is used to match the scene mode from the lighting scene model library according to the user behavior pattern library and the behavior prediction result, and send a control instruction to the lamp control unit through the serial communication interface.

[0007] In a third aspect, a radar sensing-based intelligent control device for lamps is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the radar sensing-based intelligent control device for lamps executes the above-mentioned radar sensing-based intelligent control method for lamps.

[0008] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, and when the computer-readable storage medium is run on a computer, the computer executes the above-mentioned intelligent control method of lamps based on radar sensing.

[0009] In the technical solution provided by the present application, the position information when identifying the RF radar signal of the lamp, the position information includes the user's first position, the radar detection range is divided into multiple areas according to the user's first position and differentiated area control parameters are set, the noise analysis is performed on each distance partition to obtain the interference source identification parameter set and correct the user's first position, by constructing a human behavior sequence and combining the Markov prediction model to perform behavior pattern analysis, and finally matching the appropriate lighting scene mode according to the user behavior pattern library and the behavior prediction results, effectively solving the technical problems existing in the existing lamp control technology, and having significant technical effects: using 24GHz Compared with traditional infrared sensors, FMCW frequency-modulated continuous wave millimeter wave radar can effectively detect human bodies in a stationary or slightly moving state, solving the problem that traditional sensing technology is difficult to identify stationary human bodies and improving the detection reliability of the system; through multi-area division and differentiated parameter configuration, precise control of areas at different distances is achieved, making the lighting effect more in line with the actual needs of the human body and avoiding the extensive control mode of the overall switch; through the establishment and application of interference source identification parameter sets, the influence of environmental interference sources such as air-conditioning fans and swinging curtains is effectively filtered out, significantly reducing the false trigger rate and improving system stability; through the learning of state transition probabilities, accurate prediction of human behavior is achieved, and a suitable lighting environment can be prepared before the user needs it; through scenario-based lighting control strategies, abstract behavior prediction results are converted into specific lighting parameters, achieving precise matching of lighting effects with human activity needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 A schematic diagram of an embodiment of a method for intelligently controlling a lamp based on radar sensing in an embodiment of the present application; Figure 2 This is a schematic diagram of an embodiment of a lamp intelligent control system based on radar sensing in an embodiment of the present application; Figure 3 It is a schematic block diagram of the structure of a lamp intelligent control device based on radar sensing in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a method and system for intelligent control of lamps based on radar sensing. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiment of the present application, an embodiment of the intelligent control method of lamps based on radar sensing includes: Step S101, identifying the location information of the radio frequency radar signal of the lamp, wherein the location information includes a first location of the user; Step S102: dividing the radar detection range into multiple areas according to the user's first position, and setting parameters for each distance partition after the division to obtain differentiated area control parameters; Step S103: Based on the differentiated area control parameters, noise analysis is performed on each distance partition to obtain an interference source identification parameter set, and the user's first position is corrected based on the interference source identification parameter set to obtain the user's second position; Step S104, by constructing a human behavior sequence and combining it with a Markov prediction model, a behavior pattern analysis is performed on the user's second position to obtain a user behavior pattern library and a behavior prediction result; Step S105: Match the scene mode from the lighting scene model library according to the user behavior pattern library and the behavior prediction result, and send a control instruction to the lamp control unit through the serial communication interface.

[0014] It is understandable that the execution subject of the present application can be a radar-sensing-based intelligent lighting control system, or a terminal or a server, which is not limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0015] Specifically, the 24GHz FMCW frequency modulated continuous wave millimeter wave radar sensor collects radar signals in the environment. This type of radar sensor has the characteristics of strong penetration and sensitivity to stationary targets. Compared with traditional infrared sensors, it can effectively detect human targets in a stationary or micro-moving state. After the signal emitted by the radar is reflected by the target, a difference frequency signal is obtained through mixing processing. The difference frequency signal contains distance information and relative speed information. The difference frequency signal is converted into a digital form and a fast Fourier transform is performed to convert the time domain signal into a frequency domain signal. The CFAR constant false alarm rate detection algorithm is used to extract target features from the frequency domain signal. The CFAR algorithm uses an adaptive threshold to determine whether the signal is a target echo, effectively suppressing background noise interference. After obtaining the target detection result, a two-dimensional distance-angle spectrum analysis is performed to extract the distance value and angle value of the target. The former represents the straight-line distance between the detection target and the radar sensor, and the latter represents the offset angle of the detection target relative to the central axis of the radar sensor. The spatial position information is compared and analyzed with the reflection characteristics of the radar signal. By identifying the features in the stationary or micro-moving state, the human body and non-human targets are distinguished, thereby obtaining the user's first position.

[0016] The detection range is divided into multiple zones based on the user's first position. According to the radar sensor's 0.75-meter distance resolution, the total detection distance of 9 meters is divided into 12 distance gates. These distance gates are then grouped, and the 1st to 3rd distance gates are divided into the close-range area (0-2.25 meters), the 4th to 8th distance gates are divided into the medium-range area (2.25-6 meters), and the 9th to 12th distance gates are divided into the long-range area (6-9 meters). Based on the historical distribution characteristics of the user's first position, different sensitivity thresholds are set for the three-level distance partitions: 30 for the close-range area, 20 for the medium-range area, and 15 for the long-range area. The lower the value, the higher the sensitivity. When the detected target energy value exceeds the threshold, it is determined that there is a person. Then set the trigger conditions according to the partition sensitivity configuration: set the human presence trigger condition for the close-range and medium-range areas, and set the human motion trigger condition for the long-range area. Finally, set the response mode: the close-range area adopts immediate response, the medium-range area sets a 10-second delayed response, and the long-range area sets a 30-second delayed response. These configuration combinations form differentiated regional control parameters. When the system is initialized, the background noise detection program is started, and the background signal of each distance partition is collected for 30 seconds in an unmanned state to obtain the original background noise data. The noise energy mean and standard deviation of each distance partition are calculated for these data to establish a background noise baseline model. Subsequently, the sensitivity compensation coefficient is calculated based on the model, and the ratio of the standard deviation to the mean is multiplied by the preset factor to obtain the sensitivity adjustment value, which is combined with the partition sensitivity configuration to form an interference source identification parameter set. The target reflection energy in the user's first position is pattern matched with the interference source identification parameter set. When the matching degree exceeds the preset threshold of 85%, it is determined to be an interference signal and filtered out. The spatial consistency of the filtered position information is verified to eliminate abnormal data points that do not conform to the physical characteristics of human motion. The target position and motion state are optimally estimated through the Kalman filter algorithm to obtain the user's second position.

[0017] The Doppler shift characteristics of the user's second position are analyzed, and the signals with different Doppler shift sizes and fluctuation ranges are divided into four states: still, walking, standing, and sitting. Then the human motion state information is combined with the user's second position to construct a human behavior sequence containing state, position, and time triplets. The sliding time window statistical analysis is used on these time series behavior data to calculate the occurrence frequency and conversion probability of behavior patterns in different time periods and different regions, and establish a user behavior pattern library. Based on this pattern library, a first-order Markov model is constructed, which includes a state transfer matrix and an initial state vector, where the elements in the state transfer matrix represent the conditional probability of transferring from one behavior state to another. The currently observed behavior state sequence is input into the model, and the most likely state sequence is solved by the Viterbi algorithm. The Viterbi algorithm first calculates the initial probability of each possible state, then recursively calculates the maximum probability of each state at subsequent moments, records the state transfer path, and finally backtracks to obtain the most likely state sequence. The probability values ​​in the state transfer matrix are continuously updated through Bayesian inference to obtain continuously optimized behavior prediction results.

[0018] A lighting scene model library is established, including work scenes, leisure scenes, sleep scenes, and transition scenes. Each scene defines a specific combination of lighting parameters, including brightness level, color temperature value, and lighting area range. The matching score is calculated based on the matching degree between the user behavior pattern library and the behavior prediction results and each scene, and the scene mode with the highest score is selected as the current control scene. The target lighting control parameters are packaged into a control instruction data packet, and the checksum and timestamp information are added. It is sent to the lighting control unit through the serial communication interface in PWM modulation. Feedback detection is performed on the execution results. By comparing the difference between the actual lighting state and the target lighting state, the scene matching rules in the user behavior pattern library are updated.

[0019] For example, when a user enters a bedroom, the radar sensor detects a human target. After mixing, FFT transformation and CFAR processing, the 24GHz millimeter-wave radar signal determines that the target distance is 3 meters and the angle is 30 degrees. Through spatial position information and reflection feature analysis, it is confirmed to be a human target. The system divides the detection range into three levels of distance partitions. The target is located in the middle distance area, and the sensitivity threshold is set to 20. The background noise detection program analyzes the background noise in the area and calculates the noise energy mean to be 5 and the standard deviation to be 1, forming an interference source identification parameter set. Through behavioral sequence analysis, the typical behavior of the user after entering the bedroom is identified as "walking in → standing → sitting down → staying for a long time", and the Markov prediction model predicts that the user will stay in this position for a long time. The system matches the "leisure scene" from the lighting scene model library, sets the brightness to level 5 (about 500lux), and the color temperature to 4000K, illuminating only the main activity area. The control command is sent to the lighting control unit through the serial communication interface to realize intelligent lighting control.

[0020] In the embodiment of the present application, the user's first position is obtained by extracting the position information of the collected radar signal, the radar detection range is divided into multiple areas according to the user's first position and differentiated area control parameters are set, noise analysis is performed on each distance partition to obtain the interference source identification parameter set and correct the user's first position, and the human behavior sequence is constructed and combined with the Markov prediction model to perform behavior pattern analysis, and finally the appropriate lighting scene mode is matched according to the user behavior pattern library and the behavior prediction results, which effectively solves the technical problems existing in the existing lighting control technology and has significant technical effects: using 24GHz Compared with traditional infrared sensors, FMCW frequency-modulated continuous wave millimeter wave radar can effectively detect human bodies in a stationary or slightly moving state, solving the problem that traditional sensing technology is difficult to identify stationary human bodies and improving the detection reliability of the system; through multi-area division and differentiated parameter configuration, precise control of areas at different distances is achieved, making the lighting effect more in line with the actual needs of the human body and avoiding the extensive control mode of the overall switch; through the establishment and application of interference source identification parameter sets, the influence of environmental interference sources such as air-conditioning fans and swinging curtains is effectively filtered out, significantly reducing the false trigger rate and improving system stability; through the learning of state transition probabilities, accurate prediction of human behavior is achieved, and a suitable lighting environment can be prepared before the user needs it; through scenario-based lighting control strategies, abstract behavior prediction results are converted into specific lighting parameters, achieving precise matching of lighting effects with human activity needs.

[0021] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Performing mixing processing on the collected radio frequency radar signal to obtain a difference frequency signal, wherein the difference frequency signal includes distance information and relative speed information; Perform analog-to-digital conversion and fast Fourier transform on the difference frequency signal to obtain a frequency domain signal; The target features are extracted from the frequency domain signal through the CFAR constant false alarm rate detection algorithm to obtain the target detection result; Based on the target detection results, a two-dimensional distance-angle spectrum analysis is performed to obtain spatial position information, which includes a target distance value and an angle value. The target distance value is used to represent the straight-line distance between the detection target and the radar sensor, and the angle value is used to represent the offset angle of the detection target relative to the central axis of the radar sensor. Compare and analyze the spatial position information with the radar signal reflection characteristics, identify the features in the static or micro-movement state, and obtain the results of distinguishing human and non-human targets; The human target is screened out according to the differentiation result and converted to the user's first position.

[0022] Specifically, a 24GHz FMCW frequency modulated continuous wave millimeter wave radar sensor is used for signal acquisition. The working principle of FMCW radar is to transmit a radio frequency signal whose frequency changes linearly with time. When the signal encounters a target and is reflected back, the received echo signal is mixed with the transmitted signal. The mixing process refers to multiplying the received signal with the transmitted signal, and then filtering the high-frequency component through a low-pass filter to obtain a difference frequency signal. The difference frequency signal carries the distance and speed information of the target, where the difference frequency is proportional to the target distance, and the rate of change of the frequency is proportional to the relative speed of the target.

[0023] The difference frequency signal is converted into a digital signal by the analog-to-digital conversion circuit. The analog-to-digital conversion sampling frequency is usually set at 250MHz to ensure that the slight changes of the target can be captured. Subsequently, the digitized difference frequency signal is subjected to a fast Fourier transform (FFT) to convert the time domain signal into a frequency domain signal. FFT processing can separate the different frequency components mixed together, thereby identifying multiple targets at different distances. The frequency domain signal after FFT processing contains an amplitude spectrum and a phase spectrum. The amplitude spectrum reflects the reflection intensity of the target at different distances, and the phase spectrum contains the angle information of the target. The target feature is extracted from the frequency domain signal by the CFAR constant false alarm rate detection algorithm. The working principle of the CFAR algorithm is to dynamically estimate the background noise level, and then set an adaptive threshold. When the signal amplitude exceeds the threshold, it is determined to be a valid target. The specific operation is to select a group of reference cells around the unit to be tested, calculate the average power of these reference cells, and then multiply it by a false alarm rate factor as the threshold. For lighting control applications, radar usually adopts the CA-CFAR (unit average CFAR) method, that is, assuming that the background noise is uniformly distributed in the reference cell, and directly taking the average value as the noise estimate.

[0024] Based on the effective target points detected by CFAR, a two-dimensional distance-angle spectrum analysis is performed to determine the precise spatial position of each target. The distance-angle spectrum is obtained by processing the phase difference of the signals received by different antennas. The signal of each receiving antenna is processed by FFT to obtain the distance spectrum; then, the signals of different antennas on each distance unit are processed by a second FFT to obtain the angle spectrum. Through this two-dimensional FFT processing, a two-dimensional distance-angle heat map is formed, and the peak point in the map is the position of the potential target. The target distance value is directly calculated from the peak position of the distance spectrum, representing the straight-line distance between the detected target and the radar sensor; the angle value is calculated from the peak position of the angle spectrum, representing the offset angle of the detected target relative to the central axis of the radar sensor.

[0025] After obtaining the spatial position information of the target, it is necessary to further determine whether the target is a human body. The spatial position information is compared and analyzed with the reflection characteristics of the radar signal. Human targets have specific radar cross-section (RCS) and micro-Doppler characteristics, especially the characteristics in a static or micro-movement state are different from other fixed targets. By analyzing the intensity change pattern of the target's reflected signal, the frequency offset characteristics, and the periodic changes in the time domain waveform, the micro-displacement characteristics caused by breathing and heartbeat that are unique to human targets are identified. For a static human body, the chest and abdominal displacement caused by breathing is in the millimeter level and can be detected by phase changes; for a human body in a micro-movement state, such as a slight adjustment of the sitting posture or a small movement of the hand, a change in the low-frequency component will occur on the Doppler spectrum. These features are matched with the pre-established human feature model to distinguish between human and non-human targets.

[0026] According to the results of distinguishing human and non-human targets, all human targets are screened out, and their precise spatial coordinates, motion status, target size and other information are extracted and converted into a structured user-first position as input data for subsequent lighting control decisions.

[0027] For example, when a person enters a conference room where a smart lighting control system is installed, the 24GHz millimeter-wave radar first collects the reflected signal. Through frequency mixing, a difference frequency signal containing distance and speed information is obtained. After analog-to-digital conversion and FFT processing, a clear peak is presented in the frequency domain signal. The CFAR algorithm compares this peak with the surrounding background noise level and confirms that it is a valid target. Subsequent two-dimensional distance-angle spectrum analysis determines that the target is located 3.5 meters away from the radar and 22 degrees off-angle. By analyzing the reflection characteristics of the target, a weak periodic phase change is detected with a frequency of about 0.3Hz, which is consistent with the characteristics of human breathing. At the same time, there is a weak low-frequency component in the Doppler spectrum, indicating that the target has a slight movement. Based on these characteristics, the system successfully identifies the target as a human body and records its location information. As the human body moves in the conference room, the system updates the location information in real time, and the smart lamps automatically adjust the lighting mode and brightness according to the human body's position and behavior pattern to provide a comfortable lighting environment.

[0028] In a specific embodiment, the process of executing step S102 may specifically include the following steps: The detection range is divided into 12 range gates at intervals of 0.75 meters according to the range resolution of the radar sensor, and a multi-area division result covering a total detection range of 9 meters is obtained; The multi-region division results are grouped by distance, and the 1st to 3rd range gates are divided into the short-range area, the 4th to 8th range gates are divided into the medium-range area, and the 9th to 12th range gates are divided into the long-range area, thus obtaining a three-level range partition; Based on the historical distribution characteristics of the user's first position, sensitivity thresholds are set for the three-level distance zones, with a sensitivity threshold of 30 set for the short-distance zone, a sensitivity threshold of 20 set for the medium-distance zone, and a sensitivity threshold of 15 set for the long-distance zone, to obtain a zone sensitivity configuration; According to the partition sensitivity configuration, trigger conditions are set for each distance partition, a human presence trigger condition is set for the close distance area, a human presence trigger condition is set for the medium distance area, and a human motion trigger condition is set for the long distance area, to obtain a partition trigger condition configuration; The response mode is set for the partition trigger condition configuration, an immediate response mode is set for the short-distance area, a 10-second delayed response mode is set for the medium-distance area, and a 30-second delayed response mode is set for the long-distance area, to obtain the partition response mode configuration; The partition sensitivity configuration, partition trigger condition configuration and partition response mode configuration are combined to obtain differentiated area control parameters.

[0029] Specifically, the detection range is divided into multiple range gates according to the range resolution of the radar sensor. The range gate refers to the smallest distance unit that the radar system can distinguish, and its size is determined by the system bandwidth of the radar. For the 24GHz FMCW millimeter wave radar, the distance resolution calculation formula is: distance resolution = speed of light / (2×bandwidth). In this solution, the radar bandwidth is 250MHz, and the distance resolution is about 0.75 meters. Based on this resolution, the total detection distance of 0 to 9 meters is evenly divided into 12 range gates, each of which represents a distance interval of 0.75 meters. This division method enables the radar to accurately locate the specific distance interval where the human target is located, providing a basis for subsequent differentiated control.

[0030] The 12 distance gates obtained by division are reasonably grouped to form three-level distance partitions. The specific division method is: the 1st to 3rd distance gates (0-2.25 meters) are divided into short-distance areas, the 4th to 8th distance gates (2.25-6 meters) are divided into medium-distance areas, and the 9th to 12th distance gates (6-9 meters) are divided into long-distance areas. This grouping method takes into account the interaction characteristics between the human body and lamps in actual application scenarios: the short-distance area is usually the core area where the human body is active frequently, the medium-distance area is the auxiliary area for daily activities, and the long-distance area is the edge area where the human body occasionally passes or moves. The division of three-level distance partitions lays the foundation for the subsequent differentiated parameter configuration.

[0031] Based on the historical distribution characteristics of the user's first position, sensitivity thresholds are set for the three-level distance partitions. The sensitivity threshold refers to the minimum value that the target energy value needs to exceed before it is judged as a valid target. The lower the threshold value, the higher the sensitivity, and the easier it is to detect weak target signals. According to actual application requirements, a sensitivity threshold of 30 is set for the close-range area, a sensitivity threshold of 20 is set for the medium-range area, and a sensitivity threshold of 15 is set for the long-range area. This configuration takes the following factors into consideration: on the one hand, the intensity of the target reflection signal decays with increasing distance, and the reflection signal of the distant target is weaker, requiring higher sensitivity (lower threshold) to detect; on the other hand, the sensitivity of the closer area needs to be reduced to avoid erroneous responses to small targets or non-target objects. Through this differentiated configuration, a partition sensitivity configuration that adapts to the characteristics of different distance areas is formed.

[0032] According to the partition sensitivity configuration, different trigger conditions are set for each distance partition. The trigger condition defines what kind of target state will trigger the lighting control. This solution defines two basic trigger conditions: human presence trigger and human motion trigger. Human presence trigger means that as long as the presence of a human target is detected, whether it is stationary or in motion, the control will be triggered; human motion trigger requires that the target must be in motion to be triggered. According to actual application requirements, human presence trigger conditions are set for close-range and medium-range areas, and human motion trigger conditions are set for long-range areas. This configuration conforms to the natural law of human-computer interaction: in the core activity area, stationary human bodies also need lighting; in the edge area, lighting needs to be turned on in advance only when human movement approaches.

[0033] The response mode is set for the partition trigger condition configuration, and the response timing characteristics of the lamp after the trigger condition is met are defined. This solution defines two response modes: instant response mode and delayed response mode. The instant response mode means that once the trigger condition is met, the control instruction is executed immediately; the delayed response mode waits for the preset delay time after the trigger condition is met before executing the control instruction. According to the actual application scenario, the instant response mode is set for the close-range area, the 10-second delayed response mode is set for the medium-range area, and the 30-second delayed response mode is set for the long-range area. This configuration makes the control of lamps more humane: in the core activity area, the lamps need to respond immediately to meet the lighting needs; in the medium-distance area, a short delay is used to avoid frequent switching of the lights due to the temporary stay of the human body; in the long-distance area, a longer delay is used to confirm that the human body is indeed approaching rather than just passing by briefly.

[0034] By combining the partition sensitivity configuration, partition trigger condition configuration and partition response mode configuration, we can get complete differentiated zone control parameters. These parameters are stored in the form of structured data, including the sensitivity threshold, trigger condition type and response delay time of each distance partition. The design of this differentiated control parameter fully considers the human-computer interaction characteristics of different distance zones, and realizes the refinement and intelligence of lamp control.

[0035] For example, a large office area is equipped with a radar-sensing-based intelligent lighting control system. When the radar detects employee activity, intelligent lighting control needs to be implemented based on the employee's location. The entire office area is divided into 12 distance doors, covering a detection range of 9 meters. The employee desk concentration area is in the short-distance area (0-2.25 meters), the meeting discussion area is in the medium-distance area (2.25-6 meters), and the corridor and entrance are in the long-distance area (6-9 meters). According to the historical distribution of personnel in the office area, a sensitivity threshold of 30 is set for the short-distance area so that it can effectively detect employees who sit still; a sensitivity threshold of 20 is set for the medium-distance area so that it is sensitive to the activities of people in the discussion area; and a sensitivity threshold of 15 is set for the long-distance area so that it can sense the approach of people in the corridor area in advance. In terms of trigger conditions, human presence triggers are set for the desk area and the discussion area to ensure that lighting is maintained when there are people; human motion triggers are set for the corridor area, and the lights are turned on only when someone is walking. In terms of response mode, an immediate response is set for the desk area, a 10-second delayed response is set for the discussion area, and a 30-second delayed response is set for the corridor area. In this way, when employees move around the office area, the lighting system can intelligently adjust the lighting status of each area according to their location and behavior, ensuring lighting needs while avoiding unnecessary energy waste.

[0036] In a specific embodiment, the process of executing step S103 may specifically include the following steps: During the system initialization phase, the background noise detection program is started to collect background signals in each distance zone under unmanned conditions for 30 seconds to obtain the original background noise data of each distance zone. The background noise original data of each distance partition are respectively calculated to obtain the background noise baseline model of each distance partition, and the background noise baseline model includes the noise energy statistical characteristics of each distance partition; Calculate the sensitivity compensation coefficient for each distance partition based on the background noise baseline model, obtain the sensitivity adjustment value of each distance partition by multiplying the ratio of the standard deviation to the mean by a preset factor, and obtain the interference source identification parameter set by combining the sensitivity adjustment value with the partition sensitivity configuration; The target reflected energy in the user's first position is subjected to pattern matching analysis with the interference source identification parameter set. When the matching degree exceeds a preset threshold of 85%, the corresponding signal is determined as an interference signal and filtered out to obtain the filtered position information; The filtered position information is verified for spatial consistency. By comparing the rationality of the target position changes in multiple consecutive frames of data, abnormal data points that do not conform to the physical characteristics of human motion are eliminated to obtain the position information after spatial consistency verification. The position information after spatial consistency verification is time-series fused with the historical trajectory data, and the Kalman filter algorithm is used to optimally estimate the target position and motion state to obtain the user's second position.

[0037] Specifically, the background noise detection program is started during the system initialization phase to collect background signals for each distance partition. Background noise detection refers to collecting and analyzing background noise signals in the environment while ensuring that there are no human targets in the environment. This process lasts for 30 seconds, during which the radar continuously collects background signals from each distance partition, obtains one frame of data in each sampling cycle, and obtains a total of about 7,500 frames of data (calculated at a sampling rate of 250 MHz). These raw data are classified and stored by distance partition to form the raw background noise data of each distance partition.

[0038] Perform statistical analysis on the acquired raw background noise data, and calculate the mean and standard deviation of the background noise energy for each distance partition. The specific calculation process is to first calculate the signal energy value of each frame of data in each distance partition (square the signal amplitude and sum it), and then calculate the arithmetic mean of these energy values ​​as the mean background noise energy of the distance partition, and at the same time calculate its standard deviation to measure the degree of fluctuation of the background noise. For example, for 7500 frames of data collected in 30 seconds in a certain distance partition, the signal energy value of each frame is calculated separately, and the average energy value is 5 units and the standard deviation is 1 unit. These statistical values ​​constitute the background noise baseline model, which contains the statistical characteristics of the noise energy of each distance partition, and provides a basis for subsequent interference source identification and signal filtering.

[0039] Based on the established background noise baseline model, the sensitivity compensation coefficient is calculated for each distance partition. The sensitivity compensation coefficient is calculated by multiplying the ratio of the standard deviation to the mean by a preset factor. The ratio of the standard deviation to the mean reflects the relative fluctuation of the background noise. The larger the ratio, the greater the noise fluctuation and the greater the sensitivity compensation required. The preset factor is usually set according to the system requirements, for example, set to 2. For the distance partition in the previous example, the ratio of the standard deviation to the mean is 1 / 5=0.2, and after multiplying by the preset factor 2, the sensitivity compensation coefficient is 0.4. This compensation coefficient is multiplied by the previously set partition sensitivity threshold (for example, 20) to obtain a sensitivity adjustment value of 20×0.4=8. This adjustment value is added to the original sensitivity threshold to form an adjusted threshold of 28. The same calculation process is performed on all distance partitions to obtain a set of adjusted sensitivity thresholds, which together with other interference identification parameters constitute the interference source identification parameter set. The target reflection energy in the user's first position is subjected to pattern matching analysis with the interference source identification parameter set. Pattern matching analysis refers to the calculation of the similarity between the currently detected target signal characteristics and the signal characteristics of the known interference source. Similarity calculation methods include correlation coefficient method, Euclidean distance method, etc. In this scheme, the cosine similarity calculation method of signal energy feature vector is adopted. When the calculated matching degree exceeds the preset threshold (for example, 85%), it means that the current signal is highly similar to the known interference source characteristics, and it is judged as an interference signal and filtered out. The filtering method is to mark the signal as an invalid target in subsequent processing and not participate in the calculation of human body position. In this way, false targets generated by environmental interference sources such as fans, air conditioners, and swinging curtains are effectively removed, and filtered location information is obtained.

[0040] The spatial consistency of the filtered position information is verified, mainly to check whether the change in the target position conforms to the physical characteristics of human motion. The specific method is to compare the change in the target position in multiple consecutive frames of data, calculate the target's moving speed and acceleration, and determine whether it is within the reasonable range of human motion. For example, if the target position changes more than 3 meters between two frames, and the frame interval is only 0.1 seconds, the calculated speed is 30 meters per second, which is far beyond the normal human movement speed and should be judged as an abnormal data point. Similarly, if the calculated acceleration exceeds the physiological limit of the human body, it should also be judged as abnormal. In this way, the position jump points caused by factors such as multipath effects and reflection aliasing are eliminated, and the position information after spatial consistency verification is obtained.

[0041] The position information after spatial consistency verification is time-series fused with historical trajectory data, and the Kalman filter algorithm is used to optimally estimate the target position and motion state. Kalman filtering is a recursive data processing algorithm that performs weighted averaging of measured values ​​and predicted values ​​to obtain the optimal estimate of the true state. In this scheme, the state vector of the Kalman filter contains information such as the position and speed of the target, and iterates continuously through the two stages of prediction-update. In the prediction stage, the state at the current moment is predicted based on the state estimate value at the previous moment and the system dynamic model; in the update stage, the prediction result is corrected in combination with the current measurement value. Through the Kalman filter algorithm, the target trajectory is effectively smoothed, the measurement noise is filtered out, and a more accurate and stable second position of the user is obtained.

[0042] For example, a living room is equipped with a smart lighting control system based on radar sensing. When the system is initialized, make sure that no one is in the room and start the background noise detection program. During the 30-second acquisition period, the air conditioner, TV and other electrical appliances in the living room are in operation, generating a certain amount of background noise. The system collects and analyzes the background signals of each distance partition, and calculates that the mean energy of the background noise in the middle distance area (about 3 meters) is 8, and the standard deviation is 2, forming a background noise baseline model. Based on this model, the sensitivity compensation coefficient of the middle distance area is calculated to be (2 / 8)×2=0.5, the original sensitivity threshold is 20, and the adjusted threshold is 20+20×0.5=30. When a family member enters the living room, the radar detects a target in the middle distance area with an energy value of 45, which is significantly higher than the adjusted threshold of 30, so it is determined to be a valid target. At the same time, the system also detects a signal in the long distance area with an energy value of 25, which matches the interference characteristics of the air conditioner fan by 90%, so it is determined to be an interference source and filtered out. The system continuously tracks the position of the valid target and finds that a certain frame of data shows that the target suddenly moves near the wall, which is quite different from the position of the previous and next frames. After spatial consistency verification, it is determined to be an abnormal data point and removed. Through the Kalman filter algorithm, the system obtains a smooth and accurate human motion trajectory.

[0043] In a specific embodiment, the process of executing step S104 may specifically include the following steps: The Doppler frequency shift characteristics of the user's second position are analyzed, and the user's state is divided into four states: still, walking, standing, and sitting according to the Doppler frequency shift size and fluctuation range, so as to obtain the human body motion state information; Combining the human motion state information with the user's second position, constructing a human behavior sequence including a state, position and time triple, and obtaining time series behavior data; Use sliding time window statistical analysis on time series behavior data to calculate the occurrence frequency and conversion probability of behavior patterns in different time periods and different regions, and obtain a user behavior pattern library; According to the user behavior pattern library, a first-order Markov model consisting of a state transfer matrix and an initial state vector is established to obtain a Markov behavior prediction model; The currently observed behavior state sequence is input into the Markov behavior prediction model, and the most likely state sequence is solved by the Viterbi algorithm to obtain the behavior prediction result of the human target; The behavior prediction results are compared with the actual observed behaviors, and the probability values ​​of the state transfer matrix are updated through Bayesian inference to obtain the behavior prediction results.

[0044] Specifically, the Doppler frequency shift characteristics of the user's second position are analyzed to extract the motion state information of the human target. Doppler frequency shift refers to the change in the frequency of the reflected signal after the signal emitted by the radar encounters a moving target, and its magnitude is proportional to the radial velocity of the target relative to the radar. The radar signal processing unit extracts the amplitude and spectrum characteristics of the Doppler frequency shift by analyzing the continuous multi-frame signal, and judges the motion state of the human body based on this. According to the magnitude and fluctuation range characteristics of the Doppler frequency shift, the state of the human target is divided into four categories: when the Doppler frequency shift is close to zero and the amplitude fluctuation is less than 0.05Hz, it is judged as a stationary state; when the Doppler frequency shift fluctuation range is ±0.5Hz to ±2Hz, it is judged as a walking state; when the Doppler frequency shift is close to zero and the amplitude fluctuation is between 0.05Hz and 0.1Hz, it is judged as a standing state; when the Doppler frequency shift is close to zero and the amplitude fluctuation is between 0.01Hz and 0.05Hz, it is judged as a sitting or lying state. Through this classification method, the basic motion state information of the human body is extracted from the original radar signal.

[0045] The extracted human motion state information is combined with the user's second position to construct a human behavior sequence. The human behavior sequence is a time series data structure, and each element is a triple containing three attributes: state, position, and time, in the form of {state i, position i, time i}. Among them, state i represents the human motion state (still, walking, standing, or sitting or lying) at the i-th moment; position i represents the spatial position coordinates of the human body at that moment; and time i records the specific timestamp when the state occurs. For example, a typical human behavior sequence may contain the following data: {walking, position (2 meters, 30 degrees), timestamp 1}, {standing, position (3 meters, 25 degrees), timestamp 2}, {sitting or lying, position (3.2 meters, 22 degrees), timestamp 3}, etc.

[0046] The constructed time series behavior data is analyzed by sliding time window statistics to extract user behavior patterns. The sliding time window sets a fixed-length window (e.g., 30 minutes) on the time axis and slides the window along the time axis at a certain step length (e.g., 5 minutes), and statistically analyzes the data at each window position. For the behavior sequence data in each window, the number of occurrences of different state transitions is counted, the conditional probability of state transitions is calculated, and the frequency of occurrence of each state in different time periods and different regions is counted. For example, statistics show that within a certain time window, the number of times the user changes from the "walking" state to the "standing" state is 8 times, and the "walking" state appears a total of 10 times, so the conversion probability of "walking→standing" is 0.8. Similarly, all possible state transition combinations are counted to form a complete state transition probability matrix. These statistical results constitute a user behavior pattern library, which records the typical behavior patterns and probability distribution of users at different times and in different regions.

[0047] Based on the user behavior pattern library, a first-order Markov model is established for behavior prediction. The Markov model is a random process model, which is characterized by the fact that the state of the system at the next moment is only related to the current state and has nothing to do with the earlier historical state. The first-order Markov model consists of two core components: the initial state vector and the state transfer matrix. The initial state vector π describes the probability distribution of the initial state of the system, where πi represents the probability that the initial state of the system is i; the state transfer matrix P describes the conditional probability of transitioning from one state to another, where Pij represents the probability of transitioning to state j at the next moment when the current state is i. In this solution, the initial state vector is obtained from the state frequency statistics in the user behavior pattern library, and the state transfer matrix is ​​composed of the aforementioned state transition probabilities. These two parts are combined to form a Markov behavior prediction model, which is used to predict the possible behavior state of the human body in the next step.

[0048] The currently observed behavior state sequence is input into the Markov behavior prediction model, and the most likely state sequence is solved by the Viterbi algorithm. The Viterbi algorithm is a dynamic programming algorithm used to find the hidden state sequence that is most likely to produce the observation sequence. The algorithm is first initialized to calculate the probability of each state under the condition of the first observation value; then recursively calculate the maximum probability value of each state at each subsequent moment and the corresponding previous state; finally, the entire optimal path is found by backtracking. In actual operation, the Viterbi algorithm processes discrete probability values, and gradually constructs the most likely state transfer path through table lookup and multiplication operations. The comprehensive application of the Markov model and the Viterbi algorithm can effectively predict the next possible behavior state and position of the human target, and provide a forward-looking decision-making basis for the intelligent control of lamps. The behavior prediction results are compared with the actual observed behavior, and the probability value of the state transfer matrix is ​​updated through Bayesian inference. Bayesian inference is a statistical inference method based on conditional probability. The core is Bayes' theorem, that is, the posterior probability is proportional to the product of the prior probability and the likelihood function. In this solution, the difference between the predicted result and the actual result is used as the new observation data, and the Bayesian inference formula is applied to update the probability value in the state transfer matrix. The specific operation is that when a new state transition is observed (from state i to state j), the count of the corresponding position of the state transfer matrix Pij is increased, and the probability value of the row is recalculated (ensuring that the sum of the probabilities of each row is 1). For example, if Pij=0.3 in the original state transfer matrix, after observing a new transition from i to j, the recalculated Pij may become 0.35. In this way, the system can continuously learn the user's behavior patterns, continuously optimize the prediction model, and improve the prediction accuracy.

[0049] For example, a certain residence is equipped with a radar-based intelligent lighting control system. When family members return home, the radar first collects the position and motion information of the human target. Through Doppler frequency shift analysis, it is found that the Doppler frequency shift of the human target after entering the door fluctuates around ±1Hz, and it is determined to be in a walking state. As the target moves to the central area of ​​the living room, the Doppler frequency shift gradually decreases to near zero, and the fluctuation amplitude drops to around 0.08Hz, and it is determined to be in a standing state. After that, the Doppler frequency shift further decreases, and the fluctuation drops to 0.03Hz, and it is determined to be in a sitting or lying state. The system combines these state changes with the corresponding position and time information to construct a behavior sequence {walking, (door coordinates), time 1}→{standing, (living room center coordinates), time 2}→{sitting or lying, (sofa area coordinates), time 3}. By analyzing historical data through a sliding time window, it is found that the user has an 80% probability of following this behavior pattern after returning home at night. Based on this statistical result, the system establishes a Markov model. When the user is observed to return home again and appear in the "walking, (door coordinates)" state, the Viterbi algorithm predicts the most likely subsequent behavior sequence. Therefore, the system adjusts the living room lighting to the comfort mode in advance, and then intelligently switches to the leisure reading mode without the user having to manually operate the switch. As the user's behavior habits change slightly, the system continuously updates the state transition probability through Bayesian inference, making the prediction results more and more in line with the user's actual needs, thus realizing truly intelligent lighting control.

[0050] In a specific embodiment, the process of inputting the currently observed behavior state sequence into the Markov behavior prediction model may specifically include the following steps: The human behavior observation values ​​at the current moment are constructed into an observation sequence set, in which each element represents the behavior feature observed at a specific moment, and a behavior observation sequence is obtained; Based on the behavior observation sequence, the probability matrix of the Viterbi algorithm is initialized, including the initial state probability vector, the state transition probability matrix and the observation probability matrix, to obtain the initial parameters of the Viterbi calculation; Starting from the first observation, calculate the initial probability of each possible state, and obtain the initial state probability distribution by calculating the product of the initial probability of each possible state and the probability of observing the current value in the corresponding state; For each subsequent moment, recursively calculate the maximum probability of each state, that is, the product of the maximum probability of all possible states at the previous moment and the corresponding state transition probability multiplied by the probability of observing the current value in the current state, and record the previous state corresponding to the maximum probability to obtain the state transition path; When all observations are processed, the state with the highest probability is selected as the end point from the state at the last moment, and then the target state sequence is obtained by backtracking the recorded state transition path; Based on the matching of the target state sequence with the historical patterns in the user behavior pattern library, the most likely behavior state and position at the next moment are predicted to obtain the behavior prediction result of the human target.

[0051] Specifically, the human behavior observations at the current moment are constructed into an observation sequence set, which is the data basis for behavior prediction. The human behavior observations are derived from the Doppler frequency shift characteristics, location information, and historical behavior patterns of the human body detected by the radar. Each observation contains a timestamp, location coordinates, and a corresponding behavior feature identifier. The observation sequence set is a time series data structure that records the human behavior observations at multiple consecutive moments in chronological order to form a behavior observation sequence. For example, for a 10-second observation window, data is collected once per second, forming an observation sequence of length 10, in which each element contains the behavior characteristics of the moment, such as "walking", "standing", etc. This data structure enables the algorithm to capture the temporal evolution of human behavior.

[0052] Based on the constructed behavior observation sequence, the probability matrix required by the Viterbi algorithm is initialized, including the initial state probability vector, the state transition probability matrix and the observation probability matrix, to obtain the initial parameters of the Viterbi calculation. The initial state probability vector describes the possibility that the system is initially in each state, and its value comes from the statistical frequency of each state in the user behavior pattern library. The state transition probability matrix describes the probability of transitioning from one state to another, and its element values ​​are also extracted from the user behavior pattern library. The observation probability matrix describes the probability of generating various observation values ​​in various states, and its construction method is to statistically analyze the distribution of observation features corresponding to each state in the historical data. For example, from the analysis of historical data, it is found that in the "walking" state, the probability of the Doppler frequency shift being within the range of ±1Hz is 0.8, and the probability within the range of ±1~2Hz is 0.2; while in the "standing" state, the probability of the Doppler frequency shift being close to 0 and the fluctuation amplitude being within the range of 0.05~0.1Hz is 0.9. The observation probability matrix constructed in this way can effectively describe the correspondence between the observation value and the state, providing a basis for subsequent state estimation.

[0053] Starting from the first observation, the initial probability of each possible state is calculated. This step is the initialization stage of the Viterbi algorithm. The specific calculation method is to multiply the initial probability of each state by the probability of observing the current value in that state to obtain the initial state probability distribution. For example, suppose the state space contains three states: "walking", "standing", and "sitting and lying", and their initial probabilities are 0.3, 0.4, and 0.3 respectively. The first observation is "Doppler shift ±0.8Hz". According to the observation probability matrix, the probability of this observation in the "walking" state is 0.8, the probability in the "standing" state is 0.2, and the probability in the "sitting and lying" state is 0.1. Then the initial state probability distribution is calculated as follows: the initial probability of the "walking" state is 0.3×0.8=0.24, the initial probability of the "standing" state is 0.4×0.2=0.08, and the initial probability of the "sitting and lying" state is 0.3×0.1=0.03. In this way, the initial state probability distribution considering the observation value is obtained, which lays the foundation for subsequent recursive calculations.

[0054] For each subsequent moment, the maximum probability of each state is recursively calculated. This is the core iterative process of the Viterbi algorithm. The calculation method is: for each possible state at the current moment, consider the paths transferred from all possible states at the previous moment, calculate the probability of each path, and then select the path with the highest probability. The calculation formula for path probability is: the maximum probability of the state at the previous moment multiplied by the state transition probability, and then multiplied by the probability of observing the current value in the current state. At the same time, record which state at the previous moment the maximum probability of each state comes from, that is, the previous state corresponding to the maximum probability. These records form the state transition path. For example, for the "standing" state at the second moment, the probabilities of transferring from the three states of "walking", "standing", and "sitting" at the first moment to the current "standing" state are calculated as follows: 0.24×0.6×0.7=0.1008, 0.08×0.7×0.7=0.0392, 0.03×0.3×0.7=0.0063 (where 0.6, 0.7, and 0.3 are the probabilities of transferring from "walking", "standing", and "sitting" to "standing", respectively, and 0.7 is the probability of observing the current value in the "standing" state). Comparing these three probabilities, 0.1008 is the maximum value, so the maximum probability of recording the "standing" state at the second moment is 0.1008, and the previous state is "walking". In this way, all states at all moments are recursively calculated to gradually construct the most likely state transition path.

[0055] When all observations are processed, the state with the highest probability is selected from the state at the last moment as the end point. This is the termination stage of the Viterbi algorithm. For example, for the last moment, the probabilities of the three states of "walking", "standing", and "sitting and lying" are calculated to be 0.05, 0.12, and 0.08 respectively, so the state of "standing" with the highest probability is selected as the end point. Then, the most likely complete state sequence is found by backtracking the recorded state transition path. The specific method is: starting from the end state, according to the previous state corresponding to each state recorded previously, the entire path is traced back until the initial moment is reached. For example, if the state with the highest probability at the last moment is "standing", its previous state is recorded as "walking", and the previous state is recorded as "walking", then the state sequence obtained by backtracking is "walking→walking→standing". This is the target state sequence estimated by the Viterbi algorithm, which represents the state change process that is most likely to produce the current observation sequence.

[0056] Based on the obtained target state sequence and the historical patterns in the user behavior pattern library, the most likely behavior state and position at the next moment are predicted. This step is to apply the output of the Viterbi algorithm to the actual prediction. The matching method is: find the historical pattern that is most similar to the current target state sequence in the user behavior pattern library, and extract the state and position of the next step of the pattern as the prediction result. Similarity calculation can use methods such as sequence edit distance and pattern matching algorithm. For example, if the current estimated state sequence is "walking → standing → sitting and lying down", a highly similar historical record "walking → standing → sitting and lying down → sleeping" is found in the user behavior pattern library, then the next state can be predicted to be "sleeping". At the same time, based on the position distribution corresponding to this state in the historical data, the possible position of the human body at the next moment is predicted. In this way, the behavior prediction result of the human target is obtained, including the predicted behavior state and possible position range.

[0057] For example, a user installed a radar-based intelligent lighting control system, which continuously monitors the user's behavior patterns. One night, when the radar detected that the user entered the bedroom, it first constructed the current behavior observation sequence, including the observation data of the last 10 seconds, such as [Doppler shift ±1.2Hz, position in the door area], [Doppler shift ±0.9Hz, position moving toward the bed], etc. Based on the historical user behavior pattern library, the system initialized the probability matrix of the Viterbi algorithm, including the initial probability of each state, the transition probability between states, and the probability of observing different features in various states. For the first observation value, the system calculated the initial probability of each possible state ("walking", "standing", "sitting and lying", etc.), and found that the probability of the "walking" state was the highest. Subsequently, for each subsequent moment of observation, the maximum probability of each state and the corresponding previous state are recursively calculated, and the path with the highest probability is recorded. After processing all observation values, the system finds the state with the highest probability "sitting and lying" from the last moment, and then traces back the entire state path to obtain the complete state sequence "walking→standing→sitting and lying". By matching this state sequence with the historical patterns in the user behavior pattern library, it is found that in most cases, after the user enters the bedroom and sits on the bed, the next step will be the "sleep" state, and the position will stay in the bed area for a long time. Based on this prediction, the system adjusts the bedroom lighting to the bedtime mode in advance and sets the timed lights-off function, creating a smart lighting experience for users without manual operation.

[0058] In a specific embodiment, the process of executing step S105 may specifically include the following steps: Establish a lighting scene model library including work scenes, leisure scenes, sleep scenes and transition scenes, define a specific lighting parameter combination for each scene mode, and obtain a scene lighting parameter table, which includes the brightness level, color temperature value and lighting area range of each scene; Perform feature matching based on the behavior patterns in the user behavior pattern library and the behavior prediction results, and obtain the matching score between the current behavior and each scene pattern by setting the scene correlation degree corresponding to different behavior patterns; Sort the matching scores, select the scene mode with the highest matching score as the current control scene, and extract the lighting parameters corresponding to the current control scene to obtain the target lighting control parameters; Packing the target lighting control parameters into a control instruction data packet including a scene identifier, a brightness level, a color temperature value, and a lighting area range to obtain a lighting control instruction; Add checksum and timestamp information to the lamp control command, and send it to the lamp control unit through the serial communication interface at a frequency of 20kHz according to the PWM modulation method with a duty cycle accuracy of 0.1% to obtain the lamp control execution signal.

[0059] Specifically, a lighting scene model library containing a variety of typical usage scenarios is established. The lighting scene model library is a structured data set that stores the definitions of different lighting scenes and their corresponding lighting parameter configurations. In this solution, lighting scenes are mainly divided into four categories: work scenes, leisure scenes, sleep scenes, and transition scenes. Work scenes are suitable for activities that require high concentration, such as reading, writing, housework, etc.; leisure scenes are suitable for relaxation, social activities, etc.; sleep scenes are suitable for bedtime preparation and night rest; transition scenes are temporary lighting modes for people to pass by or move briefly. Define a specific lighting parameter combination for each scene, including brightness level, color temperature value, and lighting area range, so as to construct a scene lighting parameter table. The brightness level defines the brightness of the lamp, which is usually divided into 1 to 10 levels, and the larger the value, the higher the brightness; the color temperature value defines the color characteristics of the light, which is usually expressed in absolute temperature units K (Kelvin), ranging from 2700K for warm tones to 6500K for cold tones; the lighting area range defines the range of space that needs to be illuminated, which can be accurate to each independently controlled lamp unit. For example, the working scene sets the brightness to level 9 (about 900 lumens), color temperature 6000K, and full-area lighting; the leisure scene sets the brightness to level 5 (about 500 lumens), color temperature 4000K, and main activity area lighting; the sleeping scene sets the brightness to level 2 (about 100 lumens), color temperature 2700K, and local area lighting; the transition scene sets the brightness to level 3 (about 200 lumens), color temperature 3500K, and area lighting within 3 meters around the human body. These parameter combinations form the scene lighting parameter table, which provides basic data for subsequent intelligent control.

[0060] The key to intelligent scene selection is to match the behavior patterns in the user behavior pattern library with the behavior prediction results. The feature matching process is to compare the currently observed human behavior features with the predefined behavior patterns and calculate the similarity between the two. The specific method is to set the scene association corresponding to different behavior patterns, that is, to establish a mapping relationship between the behavior pattern and the lighting scene. The association design takes into account multiple factors such as the type, duration, and location area of ​​the behavior. For example, for the behavior pattern of "enter → sit down → hold for a long time", when it occurs in the living room area, the association with the leisure scene is set to 0.8, the association with the work scene is set to 0.4, the association with the sleep scene is set to 0.2, and the association with the transition scene is set to 0.1. When the current behavior pattern is detected and combined with the behavior prediction results, the matching score between the current behavior and each scene pattern is obtained by similarity calculation. Similarity calculation methods include cosine similarity, Euclidean distance, etc. This scheme adopts the cosine similarity calculation method of weighted feature vectors. The calculation process is to represent the behavior pattern and the scenario model as feature vectors respectively, and then calculate the cosine value of the angle between the two vectors. The closer the value is to 1, the higher the similarity. This calculation method takes into account the importance of different features and reflects their influence on the final similarity by assigning different weights. In this way, the matching score between the current behavior and each predefined scenario is finally obtained, providing a quantitative basis for scenario selection.

[0061] The calculated matching scores are sorted, and the scene mode with the highest matching score is selected as the current control scene. The sorting process is to arrange the matching scores of each scene in order from high to low, and take the scene at the front as the lighting scene of the current application. If the matching scores of multiple scenes are very close (the difference is less than the preset threshold, such as 0.05), consider introducing time factors or user historical preferences as auxiliary decision conditions. After determining the current control scene, the lighting parameters corresponding to the scene are extracted from the scene lighting parameter table, including brightness level, color temperature value and lighting area range, which together constitute the target lighting control parameters. For example, if the currently selected scene is "leisure scene", the extracted parameters are: brightness level 5, color temperature value 4000K, and lighting range is the main activity area. These target lighting control parameters are the direct basis for the subsequent generation of specific control instructions. The target lighting control parameters are packaged into a control instruction data packet. This step is to convert the abstract control parameters into executable specific instructions. The control instruction data packet is a structured data format that contains key information fields such as scene identifier, brightness level, color temperature value and lighting area range. The scene identifier is a unique code used to identify the selected scene type; the brightness level, color temperature value and lighting area range are directly derived from the target lighting control parameters extracted in the previous step. The data packet packaging process is carried out according to the predefined data protocol, including the data encoding format, field arrangement order and byte alignment. For example, using binary encoding, the scene identifier occupies 1 byte (0-255), the brightness level occupies 1 byte (0-10), the color temperature value occupies 2 bytes (2700-6500), and the lighting area range description occupies 4 bytes (region coordinate encoding), forming a total of 8 bytes of basic control instruction data packets. This compact data structure ensures transmission efficiency and contains complete control information, thereby obtaining structured lighting control instructions.

[0062] Adding checksum and timestamp information to the lighting control instructions is a step to ensure the reliability and timing of instruction transmission. Checksum is an error detection mechanism used to verify whether the data has been damaged or tampered with during transmission. The calculation method is to process all bytes in the data packet with a specific algorithm (such as CRC-16 cyclic redundancy check) to obtain a check value, which is added to the data packet as an additional field. The timestamp information records the time when the instruction is generated, which is used to ensure the execution order of the instructions and remove expired instructions. The format is the number of milliseconds since the Unix epoch. After adding these two fields, the total length of the control instruction data packet increases to 12 bytes (8 bytes for basic instructions, 2 bytes for checksum, and 2 bytes for timestamp). Subsequently, the complete control instruction is sent to the lighting control unit through the serial communication interface. The serial communication interface is a communication method that transmits data in sequence. Common implementations include UART, SPI, I2C, etc. This solution uses PWM (pulse width modulation) for actual control. PWM is a technology that converts digital signals into analog control quantities. It controls the average voltage by adjusting the duty cycle of the pulse (the ratio of the on time to the cycle), thereby achieving fine adjustment of brightness and color temperature. The transmission frequency is set to 20kHz, which means that 20,000 pulses are generated per second, and the duty cycle accuracy is 0.1%, which means that 1000 levels of fine control can be achieved. In this way, the abstract scene control parameters are converted into specific lamp execution signals, realizing intelligent lighting control.

[0063] For example, when a family member enters the living room at night, the radar sensing system first detects a human target. Through the aforementioned steps of signal processing, multi-region division, noise suppression, and behavior prediction, the current user behavior pattern is identified as "enter → sit down → watch". Based on this behavior pattern, the system starts scene matching calculation. The parameters of four predefined scenes are extracted from the lighting scene model library: work scene (brightness level 9, color temperature 6000K, full area), leisure scene (brightness level 5, color temperature 4000K, main activity area), sleep scene (brightness level 2, color temperature 2700K, local area) and transition scene (brightness level 3, color temperature 3500K, area around the human body). Then, the correlation score between the current behavior pattern "enter → sit down → watch" and each scene is calculated, and the results are: work scene 0.3 points, leisure scene 0.9 points, sleep scene 0.2 points, and transition scene 0.5 points. After sorting these scores, it is found that the matching score of the leisure scene is the highest, so the leisure scene is selected as the current control scene. From the parameter settings of the leisure scene, the brightness level 5, color temperature 4000K, and main activity area (sofa area) are extracted as the target lighting control parameters. These parameters are packaged into a control instruction data packet in the format of: scene identifier (leisure = 2), brightness level (5), color temperature value (4000), lighting area range (coordinate encoding of the sofa area). After adding the CRC-16 checksum and the current timestamp, it is sent to the lighting control unit through the serial communication interface. After receiving the instruction, the control unit converts the control parameters into actual driving signals through PWM modulation: the brightness level 5 corresponds to a PWM duty cycle of 50%, the color temperature 4000K corresponds to a specific mixing ratio of cold and warm light sources, and the lighting area is realized by controlling the switch state of the lamps in the corresponding area. Finally, the living room lighting is automatically adjusted to a comfortable leisure lighting mode with moderate brightness and mild color tone, and only the lamps in the sofa area are lit, creating an ideal leisure environment for users, demonstrating the practical value of the intelligent control method of lamps based on radar sensing.

[0064] The above describes the intelligent control method of lamps based on radar sensing in the embodiment of the present application. The following describes the intelligent control system of lamps based on radar sensing in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the intelligent control system of lamps based on radar sensing includes: An extraction module 201 is used to identify the location information of the radio frequency radar signal of the lamp, wherein the location information includes a first location of the user; A division module 202 is used to divide the radar detection range into multiple areas according to the user's first position, and set parameters for each distance partition after the division to obtain differentiated area control parameters; The analysis module 203 is used to perform noise analysis on each distance partition based on the differentiated area control parameters to obtain an interference source identification parameter set, and correct the user's first position based on the interference source identification parameter set to obtain the user's second position; A construction module 204 is used to analyze the behavior pattern of the user's second position by constructing a human behavior sequence and combining it with a Markov prediction model to obtain a user behavior pattern library and a behavior prediction result; The matching module 205 is used to match the scene mode from the lighting scene model library according to the user behavior pattern library and the behavior prediction result, and send a control instruction to the lamp control unit through the serial communication interface.

[0065] Through the cooperation of the above components, the user's first position is obtained by extracting the position information of the collected radar signal, and the radar detection range is divided into multiple areas according to the user's first position and differentiated area control parameters are set. The noise analysis of each distance partition is performed to obtain the interference source identification parameter set and correct the user's first position. By constructing a human behavior sequence and combining the Markov prediction model to analyze the behavior pattern, the appropriate lighting scene mode is finally matched according to the user behavior pattern library and the behavior prediction results, which effectively solves the technical problems existing in the existing lighting control technology and has significant technical effects: using 24GHz Compared with traditional infrared sensors, FMCW frequency-modulated continuous wave millimeter wave radar can effectively detect human bodies in a stationary or slightly moving state, solving the problem that traditional sensing technology is difficult to identify stationary human bodies and improving the detection reliability of the system; through multi-area division and differentiated parameter configuration, precise control of areas at different distances is achieved, making the lighting effect more in line with the actual needs of the human body and avoiding the extensive control mode of the overall switch; through the establishment and application of interference source identification parameter sets, the influence of environmental interference sources such as air-conditioning fans and swinging curtains is effectively filtered out, significantly reducing the false trigger rate and improving system stability; through the learning of state transition probabilities, accurate prediction of human behavior is achieved, and a suitable lighting environment can be prepared before the user needs it; through scenario-based lighting control strategies, abstract behavior prediction results are converted into specific lighting parameters, achieving precise matching of lighting effects with human activity needs.

[0066] above Figure 2 The intelligent control system for lamps based on radar sensing in the embodiment of the present invention is described in detail from the perspective of modular functional entities. The intelligent control device for lamps based on radar sensing in the embodiment of the present invention is described in detail from the perspective of hardware processing.

[0067] Figure 33 is a schematic diagram of a structure of a radar-sensing-based intelligent control device for lamps provided in an embodiment of the present invention. The radar-sensing-based intelligent control device for lamps 300 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 (for example, one or more mass storage device terminals) storing application programs 333 or data 332. Among them, the memory 320 and the storage medium 330 may be short-term storage or permanent storage. The program stored in the storage medium 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the radar-sensing-based intelligent control device for lamps 300. Furthermore, the processor 310 may be configured to communicate with the storage medium 330, and execute a series of instruction operations in the storage medium 330 on the radar-sensing-based intelligent control device for lamps 300 to implement the steps of the above-mentioned radar-sensing-based intelligent control method for lamps.

[0068] The radar-sensing-based intelligent lamp control device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input and output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Figure 3 The structure of the radar-sensing-based intelligent lighting control device shown does not constitute a limitation on the radar-sensing-based intelligent lighting control device provided by the present invention, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0069] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the radar-sensing-based intelligent control method for lamps.

[0070] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0071] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a radar-sensing-based lighting intelligent control device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0072] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent control of lamps based on radar sensing, characterized in that: The method comprises: Position information when identifying the radio frequency radar signal of the lamp, the position information including the first position of the user; Dividing the radar detection range into multiple regions according to the first position of the user, and setting parameters for each distance partition after the division to obtain differentiated regional control parameters; Based on the differentiated area control parameters, noise analysis is performed on each distance partition to obtain an interference source identification parameter set, and based on the interference source identification parameter set, the first position of the user is corrected to obtain a second position of the user; By constructing a human behavior sequence and combining it with a Markov prediction model, the behavior pattern of the user's second position is analyzed to obtain a user behavior pattern library and behavior prediction results; According to the user behavior pattern library and the behavior prediction result, the scene pattern is matched from the lighting scene model library, and a control instruction is sent to the lamp control unit through the serial communication interface.

2. The method for intelligent control of lamps based on radar sensing according to claim 1, characterized in that: The location information when identifying the radio frequency radar signal of the lamp, wherein the location information includes the first location of the user, includes: Performing mixing processing on the collected radio frequency radar signal to obtain a difference frequency signal, wherein the difference frequency signal includes distance information and relative speed information; Performing analog-to-digital conversion and fast Fourier transform on the difference frequency signal to obtain a frequency domain signal; Extracting target features from the frequency domain signal using a CFAR constant false alarm rate detection algorithm to obtain a target detection result; Performing a two-dimensional distance-angle spectrum analysis based on the target detection result to obtain spatial position information, wherein the spatial position information includes a target distance value and an angle value, wherein the target distance value is used to represent the straight-line distance between the detection target and the radar sensor, and the angle value is used to represent the offset angle of the detection target relative to the central axis of the radar sensor; Compare and analyze the spatial position information with the radar signal reflection characteristics, identify the features in a stationary or slightly moving state, and obtain the results of distinguishing human and non-human targets; A human target is screened out according to the differentiation result and converted into the user's first position.

3. The method for intelligent control of lamps based on radar sensing according to claim 2, characterized in that: The radar detection range is divided into multiple regions according to the first position of the user, and parameters are set for each distance partition after the division to obtain differentiated area control parameters, including: The detection range is divided into 12 range gates at intervals of 0.75 meters according to the range resolution of the radar sensor, and a multi-area division result covering a total detection range of 9 meters is obtained; The multi-region division results are grouped by distance, the first to third range gates are divided into a short-range region, the fourth to eighth range gates are divided into a medium-range region, and the ninth to twelfth range gates are divided into a long-range region, to obtain three-level range partitions; Based on the historical distribution characteristics of the user's first position, setting sensitivity thresholds for the three-level distance zones respectively, setting a sensitivity threshold of 30 for the short-distance zone, a sensitivity threshold of 20 for the medium-distance zone, and a sensitivity threshold of 15 for the long-distance zone, to obtain a zone sensitivity configuration; According to the partition sensitivity configuration, trigger conditions are set for each distance partition, a human presence trigger condition is set for the close distance area, a human presence trigger condition is set for the medium distance area, and a human motion trigger condition is set for the long distance area, to obtain a partition trigger condition configuration; The response mode is set for the partition trigger condition configuration, an immediate response mode is set for the short-distance area, a 10-second delayed response mode is set for the medium-distance area, and a 30-second delayed response mode is set for the long-distance area, to obtain a partition response mode configuration; The partition sensitivity configuration, the partition trigger condition configuration and the partition response mode configuration are combined to obtain differentiated area control parameters.

4. The intelligent control method of lamps based on radar sensing according to claim 3 is characterized in that: The performing noise analysis on each distance partition based on the differentiated area control parameter to obtain an interference source identification parameter set, and correcting the user's first position based on the interference source identification parameter set to obtain the user's second position includes: During the system initialization phase, the background noise detection program is started to collect background signals in each distance zone under unmanned conditions for 30 seconds to obtain the original background noise data of each distance zone. The background noise raw data of each distance partition are respectively calculated to obtain a background noise baseline model, wherein the background noise baseline model includes noise energy statistical characteristics of each distance partition; Calculating a sensitivity compensation coefficient for each distance partition based on the background noise baseline model, obtaining a sensitivity adjustment value for each distance partition by multiplying a ratio of a standard deviation to a mean by a preset factor, and combining the sensitivity adjustment value with the partition sensitivity configuration to obtain an interference source identification parameter set; Performing pattern matching analysis on the target reflected energy in the first position of the user and the interference source identification parameter set, and when the matching degree exceeds a preset threshold of 85%, determining the corresponding signal as an interference signal and filtering it out to obtain filtered position information; Performing spatial consistency verification on the filtered position information, by comparing the rationality of the target position change in multiple consecutive frames of data, eliminating abnormal data points that do not conform to the physical characteristics of human movement, and obtaining the position information after spatial consistency verification; The position information after the spatial consistency check is time-series fused with the historical trajectory data, and the Kalman filter algorithm is used to optimally estimate the target position and motion state to obtain the user's second position.

5. The method for intelligent control of lamps based on radar sensing according to claim 1, characterized in that: The method constructs a human behavior sequence and combines it with a Markov prediction model to analyze the behavior pattern of the user's second position, thereby obtaining a user behavior pattern library and a behavior prediction result, including: Performing Doppler frequency shift characteristic analysis on the second position of the user, dividing the user into four states of stillness, walking, standing and sitting according to the Doppler frequency shift size and fluctuation range, and obtaining human body motion state information; Combining the human motion state information with the second position of the user to construct a human behavior sequence including a state, a position and a time triplet to obtain time series behavior data; The time series behavior data is statistically analyzed using a sliding time window to calculate the occurrence frequency and conversion probability of behavior patterns in different time periods and different regions to obtain a user behavior pattern library; According to the user behavior pattern library, a first-order Markov model consisting of a state transfer matrix and an initial state vector is established to obtain a Markov behavior prediction model; Inputting the currently observed behavior state sequence into the Markov behavior prediction model, solving the most likely state sequence through the Viterbi algorithm, and obtaining the behavior prediction result of the human target; The behavior prediction result is compared with the actual observed behavior, and the state transfer matrix probability value is updated through Bayesian inference to obtain the behavior prediction result.

6. The radar-sensing-based intelligent control method for lamps according to claim 5, characterized in that: The currently observed behavior state sequence is input into the Markov behavior prediction model, and the most likely state sequence is solved by the Viterbi algorithm to obtain the behavior prediction result of the human target, including: The human behavior observation values ​​at the current moment are constructed into an observation sequence set, in which each element represents the behavior feature observed at a specific moment, and a behavior observation sequence is obtained; Based on the behavior observation sequence, the probability matrix of the Viterbi algorithm is initialized, including the initial state probability vector, the state transition probability matrix and the observation probability matrix, to obtain the initial parameters of the Viterbi calculation; Starting from the first observation, calculate the initial probability of each possible state, and obtain the initial state probability distribution by calculating the product of the initial probability of each possible state and the probability of observing the current value in the corresponding state; For each subsequent moment, recursively calculate the maximum probability of each state, that is, the product of the maximum probability of all possible states at the previous moment and the corresponding state transition probability multiplied by the probability of observing the current value in the current state, and record the previous state corresponding to the maximum probability to obtain the state transition path; When all observations are processed, the state with the highest probability is selected as the end point from the state at the last moment, and then the target state sequence is obtained by backtracking the recorded state transition path; Based on matching the target state sequence with the historical patterns in the user behavior pattern library, the most likely behavior state and position at the next moment are predicted to obtain the behavior prediction result of the human target.

7. The intelligent control method of lamps based on radar sensing according to claim 1, characterized in that: The matching of scene modes from a lighting scene model library according to the user behavior pattern library and the behavior prediction result, and sending a control instruction to a lamp control unit through a serial communication interface, comprises: Establish a lighting scene model library including work scenes, leisure scenes, sleep scenes and transition scenes, define a specific lighting parameter combination for each scene mode, and obtain a scene lighting parameter table, wherein the lighting parameter table includes the brightness level, color temperature value and lighting area range of each scene; Perform feature matching between the behavior patterns in the user behavior pattern library and the behavior prediction results, and obtain matching scores between the current behavior and each scene pattern by setting scene associations corresponding to different behavior patterns; Sorting the matching scores, selecting the scene mode with the highest matching score as the current control scene, and extracting the lighting parameters corresponding to the current control scene to obtain the target lighting control parameters; Packing the target lighting control parameters into a control instruction data packet including a scene identifier, a brightness level, a color temperature value, and a lighting area range to obtain a lighting control instruction; Add checksum and timestamp information to the lamp control instruction, and send it to the lamp control unit through the serial communication interface at a frequency of 20kHz in a PWM modulation mode with a duty cycle accuracy of 0.1% to obtain a lamp control execution signal.

8. A radar-based intelligent lighting control system, characterized in that: Used to implement the intelligent control method of lamps based on radar sensing as described in any one of claims 1 to 7, the intelligent control system of lamps based on radar sensing comprises: An extraction module, used for identifying the position information of the lamp radio frequency radar signal, wherein the position information includes the first position of the user; A division module, used to divide the radar detection range into multiple areas according to the first position of the user, and set parameters for each distance partition after the division to obtain differentiated area control parameters; An analysis module, configured to perform noise analysis on each distance partition based on the differentiated area control parameter to obtain an interference source identification parameter set, and to correct the user's first position based on the interference source identification parameter set to obtain the user's second position; A construction module is used to analyze the behavior pattern of the user's second position by constructing a human behavior sequence and combining it with a Markov prediction model to obtain a user behavior pattern library and a behavior prediction result; The matching module is used to match the scene mode from the lighting scene model library according to the user behavior pattern library and the behavior prediction result, and send a control instruction to the lamp control unit through the serial communication interface.

9. A radar-sensing-based intelligent control device for lamps, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and the processor implements the radar sensing-based intelligent control method for lamps described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor is enabled to execute the intelligent control method for lamps based on radar sensing as claimed in any one of claims 1 to 7.

Images