Intelligent lighting control method and system for intelligent building
By using multi-channel sound signal processing and mirror sound source cancellation technology, combined with ambient light signals, the system accurately identifies human activity, solving the problem of false triggering in intelligent building lighting control and achieving a natural and energy-saving lighting wake-up effect.
Patent Information
- Application Number
- CN202610933904.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-26
- Publication Date
- 2026-11-03
AI Technical Summary
In existing intelligent building lighting control systems, sound-based solutions are easily affected by environmental noise, making it difficult to accurately identify human activities, resulting in frequent false triggers and poor user experience. Furthermore, existing methods cannot effectively distinguish between direct sound sources and mirrored sound sources.
The system employs multi-channel ambient sound signal acquisition, acoustic feature extraction and mirror sound source elimination processing, combined with a dual-stream convolutional recurrent neural network to identify valid human activity sounds, and combines ambient light signals to generate lighting switch control commands.
It achieves high signal-to-noise ratio extraction of effective human activity sounds in complex environments, reduces the trigger threshold, and allows users to naturally trigger lighting through quiet movements, providing a seamless and energy-saving lighting wake-up experience.
Smart Images
Figure CN123035581A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of lighting control technology, and in particular relates to a method and system for intelligent lighting control in smart buildings. Background Technology
[0002] In the field of intelligent building lighting control, accurately determining the presence of human activity in a space is a core issue. Existing technologies typically employ passive infrared sensors, ultrasonic sensors, or ordinary sound sensors. However, infrared sensors are susceptible to interference from environmental heat sources and cannot detect stationary individuals; ultrasonic sensors have limited coverage; and ordinary sound sensors struggle to distinguish between human activity sounds and environmental interference such as air conditioning noise and outdoor traffic noise, leading to frequent false triggering of lighting.
[0003] Currently, sound perception-based solutions typically calculate the total ambient sound pressure level and compare it to a fixed threshold. However, steady-state noises such as air conditioning, outdoor traffic, and thunderstorms easily mask the actual, quiet sounds of activity. To ensure reliable triggering, the system has to set the sound intensity threshold quite high, often requiring users to consciously perform actions such as stomping, clapping, or coughing to wake the lights, resulting in a poor user experience.
[0004] Furthermore, even with the introduction of microphone arrays in existing technologies, beamforming algorithms can be used to obtain the spatial location of sound sources and combine this with sound characteristics for human activity identification. However, in real-world environments, sound is reflected multiple times by walls and the ground, creating mirrored sound sources, causing the microphone array to locate multiple false sound source coordinates. Existing methods cannot effectively distinguish between direct sound sources and mirrored sound sources, resulting in significant errors in human activity judgments based on sound source location. Summary of the Invention
[0005] The purpose of this invention is to provide a smart lighting control method for intelligent buildings, aiming to solve the above-mentioned technical problems.
[0006] This invention is implemented as follows: a smart lighting control method for intelligent buildings, comprising the following steps: Collect multi-channel ambient sound and ambient light signals from the target area; Acoustic feature extraction and sound source localization are performed on the multi-channel environmental sound signal to obtain an acoustic feature matrix and at least one set of initial sound source spatial coordinates; The initial sound source spatial coordinates are subjected to mirror sound source elimination processing to obtain the direct sound source spatial coordinates; Based on a preset sound recognition model, the confidence level of containing valid human activity sound signals within the target area is determined according to the acoustic feature matrix and the spatial coordinates of the direct sound source. Based on the confidence level, the effective personnel activity sound signal is enhanced, and the effective personnel activity sound intensity in the target area is calculated. The ambient light intensity of the target area is determined based on the ambient light signal. Based on the sound intensity of effective human activity and the ambient light intensity in the target area, generate lighting switch control commands.
[0007] Furthermore, the method for acoustic feature extraction is as follows: adaptive filtering and noise reduction are performed on the multi-channel ambient sound signal to obtain a noise-reduced multi-channel ambient sound signal; Mel frequency cepstral coefficients and constant Q transform features are extracted from the noise-reduced multi-channel ambient sound signal to form an acoustic feature matrix.
[0008] Furthermore, the method for sound source localization is as follows: a controllable beamforming algorithm based on phase transformation weighting is used to locate the sound source in the noise-reduced multi-channel ambient sound signal to obtain the initial spatial coordinates of the sound source.
[0009] Furthermore, the multi-channel ambient sound signal is acquired through a microphone array installed in the target area; the step of performing mirror sound source cancellation processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates specifically includes: Obtain the acoustic geometry parameters of the target area and the microphone array installation location, and construct an acoustic mirror model of the target area; Based on the acoustic mirror model of the target area, calculate the mirror sound source coordinates of each order corresponding to the initial sound source spatial coordinates, and form a candidate set of mirror sound sources. Calculate the measured time difference vector from the initial sound source spatial coordinates to each element of the microphone array; The measured arrival time difference vector is matched with the theoretical arrival time difference vector of each order of mirror sound source coordinates in the mirror sound source candidate set to obtain the sound source similarity. Initial sound source spatial coordinates with similarity exceeding a preset rejection threshold are marked as mirror sound sources and rejected. Unmarked initial sound source spatial coordinates are used as direct sound source spatial coordinates.
[0010] Furthermore, the sound recognition model is trained based on a two-stream convolutional recurrent neural network. Its first input is the acoustic feature matrix, the second input is the spatial coordinates of the direct sound source, and the output is the confidence level that the target area contains sound signals of valid human activity.
[0011] Furthermore, the step of enhancing the effective personnel activity sound signal based on the confidence level and calculating the effective personnel activity sound intensity in the target area specifically includes: Determine whether the confidence level is higher than a preset confidence threshold; If so, then mark the spatial coordinates of the direct sound source corresponding to the effective personnel activity sound signal as the spatial coordinates of the target direct sound source; Using the spatial coordinates of the target sound source as the target direction, beamforming enhancement is performed on the multi-channel environmental sound signal to obtain an enhanced effective human activity sound signal. Calculate the sound pressure level of the enhanced effective human activity sound signal, which is taken as the effective human activity sound intensity.
[0012] Furthermore, the ambient light signal is collected by a light sensor installed in the target area; the step of determining the ambient light intensity of the target area based on the ambient light signal specifically includes: Determine the initial light intensity based on the currently collected ambient light signal; Collect the occlusion status signal of the light sensor to determine the occlusion coefficient; Determine the corresponding self-illuminating light contribution value based on the current lighting switch status; The initial light intensity is corrected based on the occlusion coefficient and its own illumination contribution value to obtain a corrected light intensity value, which is used as the ambient light intensity of the target area.
[0013] Furthermore, the step of generating lighting switch control commands based on the effective human activity sound intensity and ambient light intensity in the target area specifically includes: The sound intensity of effective human activity in the target area is compared with a preset sound intensity threshold to obtain the sound intensity comparison result; The ambient light intensity of the target area is compared with a preset light intensity threshold to obtain the light intensity comparison result; Based on the sound intensity comparison results and the light intensity comparison results, a lighting switch control command is generated.
[0014] Another objective of this invention is to provide an intelligent lighting control system for smart buildings, used to implement the aforementioned intelligent lighting control method for smart buildings, specifically including: The signal acquisition module is used to acquire multi-channel ambient sound signals and ambient light signals of the target area; The sound signal processing module is used to extract acoustic features and locate sound sources from the multi-channel environmental sound signals to obtain an acoustic feature matrix and at least one set of initial sound source spatial coordinates. The sound source location determination module is used to perform mirror sound source elimination processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates. The sound recognition module is used to determine the confidence level of containing valid human activity sound signals within a target area based on a preset sound recognition model, according to the acoustic feature matrix and the spatial coordinates of the direct sound source; An effective sound intensity calculation module is used to enhance the effective human activity sound signal based on the confidence level and calculate the effective human activity sound intensity in the target area. An ambient light signal processing module is used to determine the ambient light intensity of the target area based on the ambient light signal. The control command generation module is used to generate lighting switch control commands based on the sound intensity of effective personnel activity and the ambient light intensity in the target area.
[0015] This invention provides an intelligent lighting control method for smart buildings. By eliminating mirrored sound sources to obtain the clean spatial coordinates of the direct sound source, and then using this coordinates for beamforming enhancement, it can extract effective human activity sound signals with a high signal-to-noise ratio from complex multipath reverberation environments. This ensures that the calculated effective human activity sound intensity accurately reflects the actual sound energy of the activity, allowing the sound intensity threshold to be set far below the traditional total sound pressure level threshold. This invention can accurately identify effective human activity sound signals, eliminate environmental noise interference, and thus lower the sound intensity threshold for triggering lighting switch control. This allows users to trigger lighting switch control during normal conversations, slow walking, and other quiet activities, achieving a seamless, natural, and energy-saving lighting wake-up experience. Attached Figure Description
[0016] Figure 1 This is a flowchart illustrating the intelligent lighting control method for smart buildings provided in an embodiment of the present invention.
[0017] Figure 2 This is a flowchart illustrating step S300 in the intelligent lighting control method for intelligent buildings provided in an embodiment of the present invention.
[0018] Figure 3 This is a flowchart illustrating step S500 in the intelligent lighting control method for intelligent buildings provided in an embodiment of the present invention.
[0019] Figure 4 This is a flowchart illustrating step S600 in the intelligent lighting control method for intelligent buildings provided in an embodiment of the present invention.
[0020] Figure 5 This is a schematic diagram of the structure of an intelligent lighting control system for smart buildings provided in an embodiment of the present invention. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0022] like Figure 1 As shown, in one embodiment of the present invention, a method for intelligent lighting control in a smart building is provided, comprising the following steps: S100: Acquires multi-channel ambient sound and ambient light signals of the target area; S200. Perform acoustic feature extraction and sound source localization on the multi-channel environmental sound signal to obtain an acoustic feature matrix and at least one set of initial sound source spatial coordinates; S300. Perform mirror sound source elimination processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates; S400. Based on the preset sound recognition model, determine the confidence level that the target area contains effective human activity sound signals according to the acoustic feature matrix and the spatial coordinates of the direct sound source; S500. Based on the confidence level, enhance the effective personnel activity sound signal and calculate the effective personnel activity sound intensity in the target area. S600. Determine the ambient light intensity of the target area based on the ambient light signal; S700 generates lighting switch control commands based on the sound intensity of effective personnel activity and ambient light intensity in the target area.
[0023] In practical applications, the above method can be applied to lighting control in target areas such as office building corridors, hallways, and open-plan office areas. For example, in a typical nighttime application scenario in an office building, a circular microphone array consisting of multiple (e.g., four) MEMS microphones, along with a light sensor integrated into the lighting fixture panel, is installed on the ceiling of the target area (such as a corner of an open-plan office area). When an employee returns to the open-plan office area late at night, the sound of the door opening and subsequent footsteps propagate through the space. Simultaneously, the microphone array continuously executes step S100, synchronously acquiring four ambient sound signals at a certain sampling frequency and outputting multi-channel ambient sound signals; the light sensor, on the other hand, acquires moonlight from outside the window and floodlighting from adjacent buildings at a certain frequency, outputting ambient light signals.
[0024] In a preferred embodiment of the present invention, in step S200, the method for acoustic feature extraction is as follows: adaptive filtering and noise reduction are performed on the multi-channel ambient sound signal to obtain a noise-reduced multi-channel ambient sound signal; Mel frequency cepstral coefficients and constant Q transform features are extracted from the noise-reduced multi-channel ambient sound signal to form an acoustic feature matrix.
[0025] Specifically, firstly, adaptive filtering and noise reduction are performed on the multi-channel ambient sound signal. The multi-channel ambient sound signal is synchronously acquired by each element of a microphone array installed in the target area, with each channel corresponding to one element. A normalized least mean square adaptive filter is used, with one element signal as a reference, to adaptively cancel noise in the remaining channels, suppressing steady-state background noise such as air conditioners and fans, and outputting the noise-reduced multi-channel ambient sound signal.
[0026] Secondly, each channel of the denoised multi-channel ambient sound signal is framed and windowed, and its spectrum is obtained through Fast Fourier Transform. Then, the logarithmic energy is obtained by passing it through a Mel filter bank and performing a Discrete Cosine Transform to obtain the Mel frequency cepstral coefficient vector for each frame. Simultaneously, a constant Q-transform with 12 frequency points / octaves is applied to the spectrum of this frame, and the logarithmic amplitude spectrum is used as the constant Q-transform feature vector. The Mel frequency cepstral coefficient vectors of all channels in the same frame are concatenated with the constant Q-transform feature vector to form a row vector. The row vectors of each frame within a time segment are then stacked to form an acoustic feature matrix with time as the row and the feature dimension as the column.
[0027] In this embodiment of the invention, adaptive noise reduction effectively suppresses steady-state environmental interference, making subsequent features purer. In addition, the Mel frequency cepstral coefficients simulate the characteristics of human auditory perception, while the constant Q transform precisely captures low-frequency transient details. The acoustic feature matrix formed by the two together contains both short-term timbre information of the sound and fully expresses the subtle distribution of frequency components, laying a solid feature foundation for accurately distinguishing human activity sounds (such as door opening sounds and footsteps) from mechanical noise.
[0028] In a preferred embodiment of the present invention, in step S200, the method for sound source localization is as follows: the sound source is located by using a controllable beamforming algorithm based on phase transformation weighting on the noise-reduced multi-channel ambient sound signal to obtain the initial spatial coordinates of the sound source.
[0029] Specifically, firstly, the target region is discretized into a series of candidate spatial grid points. For each candidate spatial grid point p (represented in three-dimensional spatial coordinates), based on its relationship with the positions m of each array element... i Theoretical time delay required to calculate geometric distance (represented in three-dimensional spatial coordinates) ,in, Let c be the Euclidean norm and c be the speed of sound in air.
[0030] Next, a short-time Fourier transform is performed on each channel of the denoised multi-channel ambient sound signal to obtain the spectrum X of the i-th channel in the l-th frame. i (ω,l), where ω is the angular frequency, ω=2πf, f is the sound frequency, and l is the time frame index. The spatial energy spectrum value E, weighted by phase transform, is calculated for candidate spatial grid point p. PHA (p):
[0031] In the formula, is the complex conjugate of the spectrum of the k-th channel in the l-th frame; j is the imaginary unit; ω represents the theoretical time delay from candidate spatial grid point p to the i-th and k-th array elements, respectively; min and ωmax These are the lower and upper limits of the angular frequency of the processed signal, respectively. By introducing a phase transformation weighting term of the cross spectrum, only the phase difference information of the two channel signals is retained, and the amplitude is normalized to 1, thereby suppressing the interference of reverberation and noise amplitude fluctuations on positioning.
[0032] Then, all P candidate spatial grid points within the target area are traversed, and the spatial energy spectrum value of each candidate spatial grid point is calculated to form a spatial energy spectrum map. Local maxima exceeding a preset energy threshold are searched in this spatial energy spectrum map, and the candidate spatial grid points corresponding to these local maxima are used as the initial spatial coordinates of the sound source. In practical applications, since there are many sound sources in the environment, multiple sets of initial spatial coordinates of the sound source are generally obtained.
[0033] In this embodiment of the invention, the interference of signal amplitude difference on localization is effectively eliminated by phase transformation weighting, so that the sound source localization has a sharper spatial spectral peak in the reverberant environment, which significantly improves the accuracy of the initial sound source spatial coordinates and provides a more reliable input for the subsequent mirror sound source elimination step.
[0034] like Figure 2 As shown, in a preferred embodiment of the present invention, the multi-channel ambient sound signal is acquired through a microphone array installed in the target area; the step of performing mirror sound source cancellation processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates, i.e., step S300, specifically includes: S310. Obtain the acoustic geometry parameters of the target area and the microphone array installation position, and construct an acoustic mirror model of the target area; S320. Based on the acoustic mirror model of the target area, calculate the mirror sound source coordinates of each order corresponding to the initial sound source spatial coordinates, and form a mirror sound source candidate set. S330. Calculate the measured arrival time difference vector from the initial sound source spatial coordinates to each element of the microphone array; S340. The measured arrival time difference vector is matched with the theoretical arrival time difference vector of each order of mirror sound source coordinates in the mirror sound source candidate set to obtain the sound source similarity; S350. The initial sound source spatial coordinates with sound source similarity exceeding the preset elimination threshold are marked as mirror sound sources and eliminated, and the unmarked initial sound source spatial coordinates are used as direct sound source spatial coordinates.
[0035] In practical applications, wall reflection is used as an example. First, the acoustic geometric parameters include the length, width, and height of the target area, as well as the sound absorption coefficients of each wall surface. The microphone array installation positions are the three-dimensional coordinates of each element in the target area's coordinate system. These parameters are pre-stored in memory to construct an acoustic mirror model of the target area. This model treats each wall surface as a symmetrical reflective surface. For a real sound source coordinate S=(x s,y s ,z s ), its first-order mirror image sound source coordinates about the nth wall. The following mirror equation gives the result: ; In the formula, Let n be the unit normal vector of the nth wall surface to the target area. Let be the coordinates of any reference point on the wall. This represents the vector inner product; by recursively using this mirror formula, second-order and higher-order mirror source coordinates can be generated. In this embodiment, a second-order mirror is used to achieve a balance between computational complexity and completeness.
[0036] Subsequently, for a given initial sound source spatial coordinates, first-order and second-order mirror sound source coordinates with respect to the surrounding walls, ceiling, and floor are generated, forming a candidate set of mirror sound sources.
[0037] Next, for each initial sound source spatial coordinate, the measured arrival time difference between the sound source reaching the i-th array element and reaching the reference array element (denoted as the 1st array element) is estimated using the existing generalized cross-correlation-phase transform method, forming a measured arrival time difference vector. For each mirror sound source coordinate in the mirror sound source candidate set, the theoretical arrival time difference is calculated based on its geometric distance from each array element, forming a theoretical arrival time difference vector. The theoretical arrival time difference is calculated as follows: the difference between the geometric distance from the mirror sound source coordinate to the current array element and the geometric distance from the mirror sound source coordinate to the reference array element is calculated, and then divided by the speed of sound, which gives the theoretical arrival time difference.
[0038] Then, the cosine similarity between the measured arrival time difference vector and the theoretical arrival time difference vector is calculated to obtain the sound source similarity. The closer the value is to 1, the more consistent the directions of the two vectors are, that is, the higher the probability that the initial sound source is a mirror sound source.
[0039] Finally, a preset rejection threshold (e.g., 0.85) is set. If the similarity between the spatial coordinates of an initial sound source and any mirror sound source coordinate in the candidate set of mirror sound sources is greater than the rejection threshold, the initial sound source is determined to be a mirror sound source formed by wall reflection, and the corresponding initial sound source spatial coordinates are rejected. Otherwise, it is determined to be a real physical sound source, and the direct sound source spatial coordinates are retained and output.
[0040] In this embodiment of the invention, by establishing an acoustic mirror model and introducing arrival time difference vector similarity matching, false sound sources caused by reflections from walls and other objects can be accurately identified and eliminated, retaining only the direct spatial coordinates of the real physical sound source. This purifies the spatial input from the root, providing unambiguous spatial location information for subsequent sound recognition of human activities, and greatly reducing the misjudgment rate.
[0041] In a preferred embodiment of the present invention, the sound recognition model is trained based on a two-stream convolutional recurrent neural network. Its first input is an acoustic feature matrix, its second input is the spatial coordinates of the direct sound source, and its output is the confidence level that the target area contains sound signals of effective human activity.
[0042] Specifically, the sound recognition model employs an existing two-stream convolutional recurrent neural network as its main architecture, comprising two input branches: an acoustic feature stream and a spatial coordinate stream. The acoustic feature stream takes the acoustic feature matrix as its first input. This stream extracts local time-frequency patterns and high-level semantic features of sound events through three 2D convolutional layers (each followed by batch normalization and ReLU activation functions) and two max-pooling layers, ultimately outputting a fixed-length acoustic feature vector. The spatial coordinate stream takes the spatial coordinates directly from the sound source as its second input. It encodes the 3D spatial coordinates into a distributed spatial feature vector through two fully connected layers (each followed by a ReLU activation function). The output feature vectors from the two branches are concatenated in a fusion layer to obtain a joint feature vector. This joint feature vector is then fed into a gated recurrent unit layer with multiple hidden units to model the dynamic evolution of sound events and spatial locations over time, outputting a temporal feature vector. Finally, a fully connected layer and a sigmoid output layer map the temporal feature vector to a scalar value between 0 and 1, which is the confidence level that the target area contains valid human activity sound signals. This confidence level represents the probability value that the collected multi-channel environmental sound signals contain valid human activity sound signals.
[0043] Training this two-stream convolutional recurrent neural network requires collecting a large amount of labeled lighting scene data. Positive samples include sounds of human activity such as walking, talking, and door opening, along with their corresponding direct source coordinates; negative samples include sounds of non-human activity such as air conditioning operation, traffic noise outside the door, thunder, and rain, along with their corresponding direct source coordinates. Binary cross-entropy is used as the loss function, and a stochastic gradient descent optimization algorithm is employed to supervise the training of all network parameters until convergence.
[0044] like Figure 3 As shown, in a preferred embodiment of the present invention, the step of enhancing the effective personnel activity sound signal based on the confidence level and calculating the effective personnel activity sound intensity in the target area, i.e., step S500, specifically includes: S510. Determine whether the confidence level is higher than the preset confidence level threshold; if yes, proceed to step S520; otherwise, return to step S100 to re-acquire the multi-channel ambient sound signal and ambient light signal of the target area. S520. Mark the spatial coordinates of the direct sound source corresponding to the effective personnel activity sound signal as the spatial coordinates of the target direct sound source. It should be noted that if multiple direct sound source spatial coordinates are output at the same time, each direct sound source spatial coordinate and its corresponding acoustic feature matrix are input into the sound recognition model, and the direct sound source spatial coordinate with the highest confidence level and higher than the preset confidence threshold is selected and marked as the target direct sound source spatial coordinate.
[0045] S530. Using the spatial coordinates of the target direct sound source as the target direction, beamforming enhancement is performed on the multi-channel environmental sound signal to obtain the enhanced effective human activity sound signal. In practical applications, the target directly reaches the spatial coordinates p of the sound source. target For the target direction, a time-delay summation beamforming method is used to apply time delay compensation matching the target direction to the original multi-channel ambient sound signals (i.e., the undenoised multi-channel signals that retain the original phase relationship) of each channel, and then sum them with equal weights. Specifically, let the original ambient sound signal of the i-th channel be x. i (t), the time delay compensation amount corresponding to this channel is Then the enhanced effective human activity sound signal y(t) is: ; Where t is the sampling time index; after this delay summation operation, the sound signal from the direction of the target directly to the sound source is enhanced because the channels are superimposed in phase, while the environmental noise from other directions is suppressed by average because the phases are different, thus realizing spatial filtering.
[0046] S540. Calculate the sound pressure level of the enhanced effective human activity sound signal as the effective human activity sound intensity; Specifically, the sound pressure level adopts the existing equivalent continuous A-weighted sound pressure level, which is calculated as follows: First, the enhanced effective human activity sound signal is subjected to A-weighted filtering to obtain the A-weighted sound pressure signal; then, within the preset effective sound integration time window (such as within 0.5s), the equivalent continuous A-weighted sound pressure level is calculated as the effective human activity sound intensity.
[0047] In this embodiment of the invention, directional enhancement is performed only when the confidence level is high, avoiding invalid calculations; beamforming accurately extracts the effective human activity sound from the noisy background, greatly improving the signal-to-noise ratio, so that the intensity of the effective human activity sound calculated subsequently can truly reflect the energy of extremely quiet activities, thereby setting the sound intensity threshold used for subsequent comparison at a level far lower than the traditional total sound pressure level threshold, so that users can trigger the lighting with natural behaviors such as normal conversation and slow walking.
[0048] like Figure 4As shown, in a preferred embodiment of the present invention, the ambient light signal is collected by a light sensor installed in the target area; the step of determining the ambient light intensity of the target area based on the ambient light signal, i.e., step S600, specifically includes: S610. Determine the initial light intensity based on the currently collected ambient light signal; S620: Collect the occlusion status signal of the light sensor and determine the occlusion coefficient; S630. Determine the corresponding self-illuminating light contribution value based on the current lighting switch status; S640. The initial light intensity is corrected according to the occlusion coefficient and the self-illumination light contribution value to obtain the light intensity correction value, which is used as the ambient light intensity of the target area.
[0049] In practical applications, a light sensor contains a photodiode, an analog-to-digital converter, a standard light source for self-testing, and a self-testing photodiode. First, the photodiode within the light sensor converts the received ambient light into a photocurrent, which is then amplified by a transimpedance amplifier and converted into a raw digital light intensity value by the analog-to-digital converter. Based on the linear mapping relationship calibrated at the factory of the light sensor, the raw digital light intensity value is converted into an initial light intensity in lux.
[0050] Secondly, a self-test circuit consisting of a standard light source (low-power LED) and a self-test photodiode is installed inside the light sensor's light-transmitting window. At a preset calibration time (e.g., every 15 minutes or after each light switch), the system triggers the standard light source to emit a self-test light pulse of constant intensity. The self-test photodiode receives the light signal transmitted through the dust-collecting window and outputs a self-test electrical signal V. cur Read the reference electrical signal V stored in the non-volatile memory under factory clean condition. ref Calculate the shading coefficient H=V cur / V ref .
[0051] Next, the self-illumination light contribution value of the lighting circuit to the light sensor after reflection from the luminaire onto the wall was pre-calibrated through experiments at different dimming levels, and a switch state-light contribution mapping table was established. Let the current lighting switch state be 'state', which can take discrete values such as "off", "on -1 level", "on -2 level", or a continuous dimming duty cycle from 0% to 100%. The self-illumination light contribution value L of the sensor in this state is obtained by looking up the table. self (state), unit same as light intensity.
[0052] Finally, based on the occlusion coefficient H and the self-illumination light contribution value L... self (state) relative to the initial light intensity E raw After correction, the light intensity correction value E is obtained. env, which is the ambient light intensity of the target area. The specific correction formula is: .
[0053] In this embodiment of the invention, by introducing a joint correction mechanism of shading coefficient and self-illumination light contribution value, the ambient light intensity of the target area can accurately reflect the real natural light level, unaffected by long-term dust accumulation on the sensor and light emission feedback from the lamps. This ensures the accuracy of subsequent comparisons with the preset light intensity threshold and avoids the phenomenon of accidental daytime light turning on due to sensor dirt or switching oscillations caused by light feedback.
[0054] In a preferred embodiment of the present invention, the step of generating a lighting switch control command based on the effective human activity sound intensity and ambient light intensity in the target area specifically includes: The sound intensity of effective human activity in the target area is compared with a preset sound intensity threshold to obtain a sound intensity comparison result; the ambient light intensity of the target area is compared with a preset light intensity threshold to obtain a light intensity comparison result; and a lighting switch control command is generated based on the sound intensity comparison result and the light intensity comparison result.
[0055] Specifically, if the effective human activity sound intensity in the target area is greater than the preset sound intensity threshold, the sound intensity comparison result is "the activation condition is met"; otherwise, it is "the activation condition is not met".
[0056] Meanwhile, the preset light intensity threshold can be dynamically set according to the building's orientation, window area, and current time. For example, the preset value is higher during the day and lower at night. If the ambient light intensity of the target area is less than the preset light intensity threshold, the light intensity comparison result is "meets the opening conditions"; otherwise, it is "does not meet the opening conditions".
[0057] Finally, the control logic uses an AND operation: a lighting switch control command to turn on the lighting circuit is generated only if both the sound intensity comparison result and the light intensity comparison result are met. If the sound intensity comparison result is not met, a lighting switch control command is generated to maintain the current switch state or delay turning off the lighting, regardless of the light intensity comparison result. If the light intensity comparison result is not met (i.e., sufficient ambient light), a lighting switch control command is generated to turn off the lighting circuit or remain off, regardless of whether there is any valid human activity sound, to avoid unnecessary daytime lighting. The lighting switch control command can be sent to the corresponding lighting circuit's relay or LED driver controller via building bus or wireless communication to execute the switching action.
[0058] like Figure 5As shown, in another embodiment of the present invention, a smart building lighting intelligent control system is also provided to implement the above-mentioned smart building lighting intelligent control method, specifically including: Signal acquisition module 10 is used to acquire multi-channel ambient sound signals and ambient light signals of the target area; The sound signal processing module 20 is used to extract acoustic features and locate sound sources from the multi-channel environmental sound signals to obtain an acoustic feature matrix and at least one set of initial sound source spatial coordinates. The sound source location determination module 30 is used to perform mirror sound source elimination processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates. The sound recognition module 40 is used to determine the confidence level of containing valid human activity sound signals within the target area based on a preset sound recognition model, according to the acoustic feature matrix and the spatial coordinates of the direct sound source; The effective sound intensity calculation module 50 is used to enhance the effective human activity sound signal based on the confidence level and calculate the effective human activity sound intensity in the target area. An ambient light signal processing module 60 is used to determine the ambient light intensity of the target area based on the ambient light signal. The control command generation module 70 is used to generate lighting switch control commands based on the sound intensity of effective personnel activity and the ambient light intensity in the target area.
[0059] It should be noted that the above modules and units can be implemented as a computer program, which can run on a computer device. The computer device's memory can store the computer program that makes up the modules or units, enabling the processor to execute the various steps of the above method.
[0060] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps.
[0061] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.
[0062] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.
Claims
1. A method for intelligent lighting control in smart buildings, characterized in that, Includes the following steps: Collect multi-channel ambient sound and ambient light signals from the target area; Acoustic feature extraction and sound source localization are performed on the multi-channel environmental sound signal to obtain an acoustic feature matrix and at least one set of initial sound source spatial coordinates; The initial sound source spatial coordinates are subjected to mirror sound source elimination processing to obtain the direct sound source spatial coordinates; Based on a preset sound recognition model, the confidence level of containing valid human activity sound signals within the target area is determined according to the acoustic feature matrix and the spatial coordinates of the direct sound source. Based on the confidence level, the effective personnel activity sound signal is enhanced, and the effective personnel activity sound intensity in the target area is calculated. The ambient light intensity of the target area is determined based on the ambient light signal. Based on the sound intensity of effective human activity and the ambient light intensity in the target area, generate lighting switch control commands.
2. The intelligent lighting control method for intelligent buildings according to claim 1, characterized in that, The method for acoustic feature extraction is as follows: adaptive filtering and noise reduction are performed on the multi-channel ambient sound signal to obtain a noise-reduced multi-channel ambient sound signal; Mel frequency cepstral coefficients and constant Q-transform features are extracted from the noise-reduced multi-channel ambient sound signal to form an acoustic feature matrix.
3. The intelligent lighting control method for intelligent buildings according to claim 2, characterized in that, The method for sound source localization is as follows: a controllable beamforming algorithm based on phase transformation weighting is used to locate the sound source in the noise-reduced multi-channel ambient sound signal to obtain the initial spatial coordinates of the sound source.
4. The intelligent lighting control method for intelligent buildings according to claim 1, characterized in that, The multi-channel ambient sound signal is acquired through a microphone array installed in the target area; The step of performing mirror sound source cancellation processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates specifically includes: Obtain the acoustic geometry parameters of the target area and the microphone array installation location, and construct an acoustic mirror model of the target area; Based on the acoustic mirror model of the target area, calculate the mirror sound source coordinates of each order corresponding to the initial sound source spatial coordinates, and form a candidate set of mirror sound sources. Calculate the measured time difference vector from the initial sound source spatial coordinates to each element of the microphone array; The measured arrival time difference vector is matched with the theoretical arrival time difference vector of each order of mirror sound source coordinates in the mirror sound source candidate set to obtain the sound source similarity. Initial sound source spatial coordinates with similarity exceeding a preset rejection threshold are marked as mirror sound sources and rejected. Unmarked initial sound source spatial coordinates are used as direct sound source spatial coordinates.
5. The intelligent lighting control method for intelligent buildings according to claim 1, characterized in that, The sound recognition model is trained based on a two-stream convolutional recurrent neural network. Its first input is the acoustic feature matrix, the second input is the spatial coordinates of the direct sound source, and the output is the confidence level that the target area contains sound signals of valid human activity.
6. The intelligent lighting control method for intelligent buildings according to claim 1, characterized in that, The step of enhancing the effective human activity sound signal based on the confidence level and calculating the effective human activity sound intensity in the target area specifically includes: Determine whether the confidence level is higher than a preset confidence threshold; If so, then mark the spatial coordinates of the direct sound source corresponding to the effective personnel activity sound signal as the spatial coordinates of the target direct sound source; Using the spatial coordinates of the target sound source as the target direction, beamforming enhancement is performed on the multi-channel environmental sound signal to obtain an enhanced effective human activity sound signal. Calculate the sound pressure level of the enhanced effective human activity sound signal, which is taken as the effective human activity sound intensity.
7. The intelligent lighting control method for intelligent buildings according to claim 1, characterized in that, The ambient light signal is collected by a light sensor installed in the target area; The step of determining the ambient light intensity of the target area based on the ambient light signal specifically includes: Determine the initial light intensity based on the currently collected ambient light signal; Collect the occlusion status signal of the light sensor to determine the occlusion coefficient; Determine the corresponding self-illuminating light contribution value based on the current lighting switch status; The initial light intensity is corrected based on the occlusion coefficient and its own illumination contribution value to obtain a corrected light intensity value, which is used as the ambient light intensity of the target area.
8. The intelligent lighting control method for intelligent buildings according to claim 1, characterized in that, The steps for generating lighting switch control commands based on the sound intensity of effective human activity and ambient light intensity in the target area specifically include: The sound intensity of effective human activity in the target area is compared with a preset sound intensity threshold to obtain the sound intensity comparison result; The ambient light intensity of the target area is compared with a preset light intensity threshold to obtain the light intensity comparison result; Based on the sound intensity comparison results and the light intensity comparison results, a lighting switch control command is generated.
9. A smart building lighting intelligent control system, used to implement the smart building lighting intelligent control method according to any one of claims 1-8, characterized in that, include: The signal acquisition module is used to acquire multi-channel ambient sound signals and ambient light signals of the target area; The sound signal processing module is used to extract acoustic features and locate sound sources from the multi-channel environmental sound signals to obtain an acoustic feature matrix and at least one set of initial sound source spatial coordinates. The sound source location determination module is used to perform mirror sound source elimination processing on the initial sound source spatial coordinates to obtain the direct sound source spatial coordinates. The sound recognition module is used to determine the confidence level of containing valid human activity sound signals within a target area based on a preset sound recognition model, according to the acoustic feature matrix and the spatial coordinates of the direct sound source; An effective sound intensity calculation module is used to enhance the effective human activity sound signal based on the confidence level and calculate the effective human activity sound intensity in the target area. An ambient light signal processing module is used to determine the ambient light intensity of the target area based on the ambient light signal. The control command generation module is used to generate lighting switch control commands based on the sound intensity of effective personnel activity and the ambient light intensity in the target area.