An adaptive color-changing lighting control system based on sound control sensing

Abstract

This invention relates to the field of intelligent lighting control technology, and discloses an adaptive color-changing lighting control system based on sound-controlled sensing, solving the problems of existing systems being susceptible to acoustic reverberation interference and causing visual flicker. The system includes a main control module, a sensor network, and an execution network. Sensor nodes collect ambient acoustic signals, and the main control module extracts acoustic features and sound source angles of arrival based on these signals, generating target lighting vectors and determining target groups. During system initialization, node binding and weight allocation are achieved using radio frequency signals and the direct sound energy ratio. During operation, based on transient discrimination factors, step-limited walk and smooth interpolation approach control are performed in steady-state and transient states, respectively. Simultaneously, a dynamically variable fractional-order PID controller is introduced, adjusting the calculus order according to the proportion of high-frequency acoustics, combined with nonlinear decoupling to output the drive signal. This invention can automatically adapt to local acoustic quality, reduce false triggering rate, suppress chromaticity drift and illuminance overshoot, and eliminate visual flicker.

Description

Technical Field

[0001] This invention relates to the field of intelligent lighting control technology, and more specifically, to an adaptive color-changing lighting control system based on sound-sensing. Background Technology

[0002] With the popularization of intelligent control and Internet of Things (IoT) technologies, sound and light linkage lighting systems have been widely used in commercial spaces, performance stages, and smart home environments. Traditional sound and light control systems typically use microphones to collect ambient audio, extract basic acoustic parameters such as volume or frequency band energy, and directly map them to the brightness and color of the lights. This control method can achieve basic sound and light interaction in ideal acoustic environments, but its limitations become increasingly apparent when facing complex physical environments.

[0003] In complex spatial acoustic environments, existing sound-light linkage systems are highly susceptible to interference from physical interface reverberation and multipath noise. Due to the lack of effective acoustic signal separation and spatial resolution mechanisms, these systems often misinterpret reflected sound or ambient noise as valid commands, leading to frequent trigger misjudgments and spatial source location misalignment. Furthermore, in distributed systems, sound acquisition nodes and lighting execution nodes are typically spatially discrete, and current technologies lack methods for spatial topological mapping of these two types of hardware nodes. This results in the system's inability to implement precise targeted lighting adjustments for specific sound source areas, leading to poor spatial tracking performance in sound-light linkage.

[0004] In terms of underlying light-emitting drive control, traditional systems often employ fixed-parameter control models or direct open-loop control strategies, failing to fully consider the inherent nonlinear electro-optic conversion characteristics of light-emitting devices. When the ambient sound field experiences energy transients or sudden high-frequency acoustic events, the fixed control logic cannot adaptively adjust damping and tracking speed, easily leading to illuminance overshoot and light intensity oscillations at the drive end, resulting in noticeable visual flickering. Existing systems not only fail to provide naturally varying light effects when the background sound is stable, but also lack control mechanisms that perform smooth approximation and anti-abrupt control based on the dynamic characteristics of the sound field. These deficiencies directly result in the harsh light effects of current sound-light linkage systems, easily causing visual adaptation dulling and fatigue in users, making it difficult to meet the requirements for constructing high-quality intelligent lighting environments. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an adaptive color-changing lighting control system based on sound-controlled sensing. The technical problem solved by this invention is that existing sound-light linkage lighting systems are susceptible to physical interface reverberation and multipath noise interference in complex spatial acoustic environments, leading to trigger misjudgment and spatial positioning deviation. At the same time, traditional systems often adopt fixed-order or open-loop control strategies in the underlying light-emitting driving stage. When dealing with the nonlinear electro-optical conversion characteristics of the light source and transient jumps in sound field energy, they are prone to illuminance overshoot, light intensity oscillation, and visual flicker. They lack a mechanism for adaptive topology reconstruction and smooth approximation control based on the dynamic characteristics of the sound field.

[0006] To address the above problems, the present invention provides the following technical solution: This invention provides an adaptive color-changing lighting control system based on sound-controlled sensing, comprising: Main control module, sensor network and execution network; The main control module includes a signal processor; The sensor network establishes a data communication connection with the main control module, and includes multiple discretely distributed sensor nodes. Each sensor node integrates a microphone array for collecting multi-channel environmental acoustic signals. The execution network communicates with the main control module and includes multiple execution nodes. Each execution node includes a microcontroller, a driver module, a second communication module, and a photoelectric feedback module. The photoelectric feedback module is used to collect the actual lighting state vector of the area where the execution node is located. The signal processor is used to extract acoustic features and the spatial angle of arrival of the sound source based on the multi-channel environmental acoustic signal, generate a target lighting state vector based on the acoustic features, determine the target broadcast group according to the spatial angle of arrival of the sound source and the spatial topology coordinate table stored in the system, generate a target broadcast control data frame according to the deviation between the target lighting state vector and the actual lighting state vector of the corresponding execution node, and send it to the execution node in the target broadcast group to drive the underlying light-emitting device.

[0007] Furthermore, the actual lighting state vector includes CIE 1931 chromaticity horizontal axis components, chromaticity vertical axis components, and luminance components; The photoelectric feedback module includes at least one of an illuminance sensor, a color sensor, or an integrated colorimetric sensor; When generating the target broadcast control data frame, the signal processor calculates the control deviation vector between the target lighting state vector and the actual lighting state vector of the corresponding execution node; A fractional-order proportional-integral-derivative controller is used to process the control deviation vector to generate a corrected lighting state vector; The correction lighting state vector is converted into a pulse width modulation duty cycle, and the pulse width modulation duty cycle is encapsulated into a target broadcast control data frame and then sent to the corresponding execution node.

[0008] Furthermore, the main control module is configured to enter offline calibration mode during the initialization phase and perform the following operations: A preset broadband test signal is sent to the physical space, and each sensor node collects the acoustic response signal; the acoustic transfer function of the corresponding sensor node location is calculated and the direct sound energy ratio is calculated; The system sends a wireless ranging command to the execution network, constructs a received signal strength indication matrix based on the received signal strength indication value extracted from the radio frequency test frame between the execution node and the sensor node, and generates an equivalent received signal strength indication matrix through moving average filtering. Based on the equivalent received signal strength indication matrix, each execution node agent is bound to the sensor node with the largest equivalent received signal strength. The normalized acoustic feature value obtained by normalizing the direct sound energy ratio of the bound sensor node is used to assign node convergence weights to each execution node.

[0009] Furthermore, the signal processor extracts acoustic features and sound source spatial angle of arrival based on multi-channel environmental acoustic signals in the following ways: Multi-channel environmental acoustic signals are converted into frequency domain signal vectors, a spatial cross-correlation matrix is ​​constructed based on the frequency domain signal vectors, and eigenvalue decomposition is performed on the spatial cross-correlation matrix. Based on the established information theory criteria, the signal subspace and noise subspace are separated, and a spatial pseudospectral function is constructed using the noise subspace to extract the spatial angle of arrival of the sound source. The multi-channel environmental acoustic signal is weighted and synthesized using the signal feature vector corresponding to the signal subspace to generate a backbone audio sequence. The backbone audio energy envelope is reconstructed based on the backbone audio sequence, and the centroid frequency value of the spectrum is extracted by frequency domain transformation. The backbone audio energy envelope and the centroid frequency value of the spectrum are concatenated to construct an acoustic time-frequency feature vector, which is used as the acoustic feature.

[0010] Furthermore, the signal processor generates the target illumination state vector based on acoustic features in the following ways: The core audio energy envelope is mapped to the target luminance component, and the spectral centroid frequency value is mapped to the correlated color temperature value, which is then converted into the target chromaticity coordinate component in the CIE 1931 chromaticity space. The components are combined to generate the target illumination state vector. Based on the spatial vector pointed to by the spatial arrival angle of the sound source and the spatial topology coordinate table stored in the system, the targeted broadcast group is determined, and the targeted broadcast control data frame generated based on the target lighting state vector is sent to the execution node in the targeted broadcast group.

[0011] Furthermore, the signal processor is also configured to execute transient sound field discrimination logic, specifically including: Calculate the absolute value of the difference between the backbone audio energy envelope of the current time frame and the long-term moving average background energy of the previous time frame, and extract the ratio of the absolute value of the difference to the long-term moving average background energy as the transient discrimination factor. Compare the transient discrimination factor with the preset transient change threshold; When the transient discrimination factor is greater than the transient change threshold, the sound field state indicator is configured to transient excitation state; when the transient discrimination factor is less than or equal to the transient change threshold, the sound field state indicator is configured to steady-state background state.

[0012] Furthermore, when the sound field state indicator is in a steady-state background state, the main control module controls the target illumination state vector corresponding to the execution node to perform step-size limited walk control: A pseudo-random sequence generator is invoked to generate a three-dimensional random perturbation vector; Calculate the difference between the preset base anchor point vector and the target illumination state vector of the previous frame, apply the center recovery coefficient, and combine the three-dimensional random perturbation vector and the walk step gain to generate an intermediate illumination state vector that converges to the base anchor point. Perform boundary truncation on the intermediate lighting state vector and output the target lighting state vector of the current time frame.

[0013] Furthermore, when the sound field status indicator is in a transient excitation state, the main control module performs physical weight approach control for the execution node: The transient discriminant factor is mapped to the approach control weight using a nonlinear saturation function; The comprehensive convergence coefficient of the execution node is calculated by using the node convergence weight, the approach control weight, and the Euclidean distance between the target illumination state vector of the previous frame and the transient target illumination vector generated by the acoustic time-frequency feature vector of the current time frame. Based on the comprehensive convergence coefficient, interpolation smoothing is performed on the target illumination state vector of the previous frame and the transient target illumination vector generated by the acoustic time-frequency feature vector of the current time frame to calculate the target illumination state vector of the execution node in the current time frame.

[0014] Furthermore, methods for generating a corrected lighting state vector by processing the control deviation vector using a fractional-order proportional-integral-derivative controller include: The signal processor constructs a fractional calculus operator based on a truncated Grünwald-Letnikov time-domain discrete approximation algorithm with finite memory length, and uses the fractional calculus operator to calculate fractional calculus terms; The control correction vector is generated by linearly combining the fractional calculus term and the proportional term, and then superimposed on the actual lighting state vector to obtain the corrected lighting state vector. Acquire the acoustic power spectral density data of the multi-channel environmental acoustic signal in the current time frame, and extract the ratio of the total energy in the high-frequency band to the total energy in the entire frequency band as the proportion of the high-frequency component; The target order of the integral and derivative elements is calculated by utilizing the proportion of high-frequency components. After performing time-domain smoothing filtering, the fractional order of the integral and derivative elements of the fractional calculus operator is dynamically overwritten and adjusted.

[0015] Furthermore, methods for converting the corrected lighting state vector into a pulse width modulation duty cycle include: After applying lower limit protection to the chromaticity ordinate component in the corrected lighting state vector, the corrected lighting state vector is converted into a tristimulus value vector in a linear mapping space. The tristimulus value vector is left-multiplied by a preset channel inverse mapping matrix, and linearly decoupled into the original primary color intensity components of each physical light emission channel. Bidirectional truncation constraints are applied to each original primary color intensity component, and an inverse gamma exponential function is introduced to perform inverse nonlinear correction compensation for the luminescence intensity. The pulse width modulation duty cycle of each physical light emission channel is then calculated.

[0016] This invention provides an adaptive color-changing lighting control system based on sound-controlled sensing. It has the following beneficial effects: 1. This invention utilizes the received signal strength indication to construct a spatial adjacency matrix to achieve node mapping and binding, thus solving the problem of spatial topological discontinuity; at the same time, it extracts the direct sound energy ratio by measuring the acoustic transfer function as the convergence weight of each node, so that the light intensity adjustment can automatically adapt to the local acoustic quality, effectively reducing the false triggering rate of lighting caused by environmental reverberation.

[0017] 2. This invention establishes a cross-modal nonlinear control logic based on acoustic characteristics. In steady state, random walks are used to simulate the gradual change of natural light to prevent visual dulling; when transient events are detected, smooth interpolation is performed by synthesizing the approach coefficients to limit the step rate of the drive signal and avoid overloading the constant current power supply.

[0018] 3. This invention designs a dynamically variable fractional-order controller and a nonlinear feedback decoupling mechanism. By dynamically adjusting the integral and derivative orders based on the proportion of high-frequency acoustic components, combined with algebraic decoupling and inverse gamma compensation, the photometric response is maintained in a critically damped state, effectively suppressing chromaticity drift and illuminance overshoot, and eliminating visual flicker. Attached Figure Description

[0019] Figure 1 This is a block diagram of the overall hardware and control architecture of the system of the present invention; Figure 2 This is a flowchart illustrating the overall control method of the present invention. Figure 3 This is a flowchart of the offline sound field calibration and node weight allocation process of the present invention; Figure 4 This is a flowchart of the acoustic feature extraction and target illumination mapping process of the present invention; Figure 5 This is a flowchart of the transient sound field discrimination and dynamic strategy control of the present invention; Figure 6 This is a flowchart illustrating the adaptive closed-loop correction and decoupling of the underlying driver in this invention. Figure 7 This is a state evolution curve diagram of the control process according to an embodiment of the present invention; Figure 8 This is a comparison curve of the step response of the algorithm in an embodiment of the present invention; Figure 9 This is a bar chart comparing the core performance indicators of embodiments of the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See attached document Figure 1 The present invention provides an adaptive color-changing lighting control system based on sound-controlled sensing, the system comprising a main control module, a sensor network, and an execution network.

[0022] The main control module includes a signal processor. The signal processor generates a corrected lighting state vector based on the control deviation vector between the target lighting state vector and the actual lighting state vector. The signal processor is used to perform matrix operations and sound field modeling algorithms. A sensor network is deployed in the physical space, containing multiple discretely distributed sensor nodes. Each sensor node integrates a microphone array and a first communication module. The microphone array is used to synchronously acquire multi-channel environmental acoustic signals. The first communication module is used for data exchange between nodes. The execution network contains multiple execution nodes distributed in the physical space. Each execution node includes a microcontroller, a driver module, a second communication module, and a photoelectric feedback module.

[0023] The microcontroller connects to the driver module and the photoelectric feedback module. The photoelectric feedback module acquires the actual lighting state vector of the area where the execution node is located. This vector includes the CIE 1931 chromaticity horizontal axis component, chromaticity vertical axis component, and luminance component. The photoelectric feedback module includes at least one of an illuminance sensor, a color sensor, or an integrated chromaticity sensor. It transmits the acquired optical feedback data to the microcontroller or uploads it to the main control module via a second communication module for closed-loop correction of the lighting output state. The execution node connects to the system communication network via the second communication module. The main control module establishes data communication connections with both the sensor network and the execution network.

[0024] It should be noted that, in this embodiment, the main control module is responsible for performing acoustic feature extraction, sound field state discrimination, target illumination state vector generation, node weight convergence control, fractional proportional-integral-derivative control, and pulse width modulation duty cycle calculation; the execution node is responsible for receiving configuration data frames or control data frames issued by the main control module, and writing the pulse width modulation duty cycle into the driver module through the microcontroller to drive the underlying light-emitting devices. Thus, the system forms a control architecture that combines centralized sound field analysis with distributed illumination execution.

[0025] See attached document Figure 2 This invention provides an adaptive color-changing lighting control system based on sound-controlled sensing, comprising the following steps: S1. During the initialization phase, the system enters the offline calibration state and sends a preset broadband test signal to the surrounding space. Each sensor node collects the response signal. The main control module calculates the acoustic transfer function of the physical location of the corresponding sensor node and calculates the direct sound energy ratio. It extracts the wireless received signal strength indication matrix between each execution node and each sensor node. Based on the wireless received signal strength indication matrix, the system binds each execution node to the sensor node with the largest equivalent received signal strength and uses this sensor node as the spatial neighbor proxy node of the corresponding execution node. Based on the direct sound energy ratio of the bound sensor nodes, the system assigns node convergence weights of the discrete wolf pack algorithm to each execution node. S2, after the system completes parameter configuration, it enters the online operation state. The sensor nodes collect the environmental acoustic sequence in real time. The main control module calculates the cross-correlation function between different channels of environmental acoustic sequences according to the preset sliding time window to construct the cross-correlation matrix. The cross-correlation matrix is ​​subjected to eigenvalue decomposition to extract principal component features. Based on the principal component features, the multi-channel environmental acoustic sequences are weighted and synthesized to generate the backbone audio sequence. The backbone audio energy envelope is reconstructed based on the backbone audio sequence. The frequency centroid frequency value of the current sound field is extracted through frequency domain transformation. S3, the main control module inputs the extracted backbone audio energy envelope and the spectral centroid frequency value into the system's built-in mapping function to calculate the target chromaticity coordinates and target brightness in the CIE 1931 chromaticity space, and encapsulates the target chromaticity coordinates and target brightness into a target illumination state vector; the main control module determines the target broadcast group based on the spatial angle of arrival of the sound source, and sends the target illumination state vector to the execution nodes in the target broadcast group, and maintains the previous frame's illumination state or performs steady-state walk control for execution nodes not included in the target broadcast group; S4, the main control module calculates the transient discriminant factor of the backbone audio energy envelope relative to the long-term moving average background energy. When the transient discriminant factor is less than or equal to the preset transient mutation threshold, the main control module controls the photometric state vector corresponding to the execution node to perform a step-limited random walk update within the current numerical neighborhood; when the transient discriminant factor is greater than the preset transient mutation threshold, the main control module retrieves the node convergence weight generated by the corresponding execution node in the offline calibration state from the node configuration table, or reads the node convergence weight from the corresponding execution node through the system communication network, and calculates the state approach update amount of the corresponding execution node using the node convergence weight, the transient discriminant factor, and the Euclidean distance between the current photometric state vector and the target illumination state vector; S5, the main control module acquires the actual lighting state vector collected by the photoelectric feedback module, calculates the control deviation between the updated photometric state vector and the actual lighting state vector, and inputs the control deviation into the fractional proportional-integral-derivative (FPI) controller; the main control module extracts the high-frequency component proportion data of the backbone audio energy envelope, and dynamically adjusts the integral order parameter and derivative order parameter of the FPI controller according to the high-frequency component proportion data; the FPI controller outputs the control correction vector, the main control module converts the control correction vector into the corrected lighting state vector, and linearly decouples it into a pulse width modulation (PWM) command before sending it to the execution node; the microcontroller of the execution node receives the PWM command and inputs it to the driver module to drive the underlying light-emitting device.

[0026] To gain a deeper understanding of the innovative logic and implementation mechanism of this invention, the following will elaborate on the specific operational rules, hardware control principles, and cross-domain collaborative technology details in the above system architecture and method steps.

[0027] See attached document Figure 3 In this embodiment of the invention, acoustic characteristics detection of the space physical environment is performed during the system initialization phase. The specific implementation includes the following steps.

[0028] S111: When the system powers on and initializes or receives an external calibration command, the main control module enters offline calibration mode. The main control module sends control signals to an audio playback device in the environment via the communication bus, triggering the playback of a broadband test signal with a known waveform. The audio playback device is either an external calibration sound source device or a built-in calibration sound source module controlled by the main control module, used to send a preset broadband test signal to the physical space during offline calibration.

[0029] To establish a time-domain alignment reference between the transmitting and receiving ends, the main control module sends control signals while simultaneously distributing synchronous acquisition timestamps to each sensor node via the system communication network. The broadband test signal employs a logarithmic sweep frequency signal. The frequency of the logarithmic sweep frequency signal changes exponentially with time, covering the audio sensing frequency band within a preset time window and stimulating multipath reflection characteristics in space. The digital waveform synthesis and analog-to-analog conversion output circuitry for the logarithmic sweep frequency signal is well-known in the field and will not be described in detail here.

[0030] S112, the sensor nodes are in monitoring mode. Each sensor node acquires acoustic response signals using a microphone array based on a synchronized acquisition timestamp. The spatial physical environment's effects on sound wave transmission and reflection are equivalent to a linear time-invariant system. The acoustic response signal includes the initial sound wave reaching the corresponding sensor node, as well as multi-order reverberant sound waves reflected by indoor obstructions. The microphone array processes the acoustic response signal through analog-to-digital conversion to generate a digital acoustic response sequence. The sensor nodes transmit the digital acoustic response sequence back to the main control module via the first communication module.

[0031] S113, the signal processor in the main control module receives the digital acoustic response sequences from each sensor node. The signal processor uses the original broadband test signal stored internally in the system and the digital acoustic response sequences to perform frequency domain deconvolution operations to calculate the acoustic transfer function at the location of each sensor node.

[0032] In the specific calculations, the signal processor performs discrete Fourier transforms on the digital acoustic response sequence and the original broadband test signal respectively, converting them to the frequency domain. For the first... Each sensing node, the signal processor calculates the frequency domain sequence of the acoustic transfer function according to the formula: In the formula, Indicates frequency domain index; Indicates the first The frequency domain sequence of the acoustic transfer function at the corresponding location of each sensing node; This represents the frequency domain sequence of a digital acoustic response sequence after discrete Fourier transform. This represents the frequency domain sequence of the original broadband test signal; Represents the complex conjugate of the frequency domain sequence of the original broadband test signal; This represents the auto-power spectral density of the original broadband test signal; This represents the regularization constant.

[0033] Regularization constant The value is set to 0.01% to 1% of the maximum power spectral density of the original broadband test signal, and is configured according to the measured variance of the ambient background noise. This is used to prevent the test signal from attenuating in a specific frequency band, causing the denominator to approach zero.

[0034] Signal processors acquire frequency domain sequences Then, an inverse discrete Fourier transform is performed on it to output the acoustic transfer function in the time domain. ,in Indicates time index. Acoustic transfer function. Characterizes the test signal generation source to the first The acoustic impulse response characteristics of each sensing node location are used to map the attenuation characteristics of sound waves in the spatial physical environment into a time series, which is then used to calculate the direct sound energy ratio.

[0035] After obtaining the acoustic transfer function of each sensor node, the main control module performs sound field feature quantization processing. The specific implementation includes the following steps.

[0036] S121, the signal processor in the main control module performs peak retrieval on the acoustic transfer function in the time domain to determine the arrival time of the direct sound. The acoustic transfer function reflects the energy attenuation process of sound waves propagating and reflecting in space. The first part to arrive and with concentrated energy corresponds to the initial wavefront of the sound source directly reaching the sensing node, and the subsequent energy tail corresponds to the reverberation component formed by reflection from the spatial interface.

[0037] The signal processor searches for the discrete-time index corresponding to the maximum absolute value in the sequence of acoustic transfer functions, and identifies this time index as the point of arrival of the direct sound. Using the point of arrival of the direct sound as the time reference, the signal processor defines the upper threshold of the direct sound time window. Upper limit threshold The value is taken as the time scale value of the direct sound arrival point plus a fixed time offset of 2.5ms to 5ms. For time window truncation and temporal envelope extraction of audio signals, those skilled in the art can use conventional digital signal windowing processing methods. The signal framing principle is a well-known technology in the field and will not be elaborated here.

[0038] S122, the signal processor calculates the time window of the direct sound according to the defined time window. The direct sound energy and reverberant sound energy of each sensing node. The direct sound energy is calculated from the square of the acoustic transfer function at time zero to the upper threshold. The reverberant energy is obtained by integrating the integrals between the two. The reverberant energy is derived from the square of the acoustic transfer function at an upper threshold. To the effective cut-off length The integral between them is obtained. The calculation formula is as follows: ; ; In the formula, Indicates the first Direct acoustic energy of each sensing node; Indicates the first The reverberant energy of each sensing node; Indicates the first Acoustic transfer function of each sensing node at its corresponding location; Indicates a time index; This indicates the upper threshold of the direct sound time window; Represents the effective cutoff length of the acoustic transfer function. Effective cutoff length The value is determined based on the decay characteristics of the system's impulse response, and is taken as the time point corresponding to the decay of the acoustic transfer function energy to the system's basic noise level.

[0039] S123, the signal processor utilizes direct acoustic energy With reverberation energy Calculate the first The direct sound energy ratio of each sensing node The formula for calculating the direct sound energy ratio is: In the formula, Indicates the first The direct sound energy ratio at the location of each sensing node; This represents a protection constant to prevent the denominator from approaching zero. It is used to avoid system division overflow errors caused by insufficient reverberation energy. Its value is configured based on the processor's calculation precision. Direct acoustic energy ratio. This ratio reflects the relative energy of the initial sound wave at the target location before environmental reflection to the energy of the reverberated sound wave after multiple reflections. This ratio is related to the spatial distance between the sensing node and the sound source, as well as the sound absorption characteristics of the surrounding reflective surfaces. The signal processor calculates the direct sound energy ratio for each sensing node. The multipath acoustic interference attenuation characteristics are transformed into quantitative feature indicators. These quantitative feature indicators serve as the basis for evaluating the acoustic signal quality at each physical location. They are written into the registers of the main control module and used in subsequent stages to generate and allocate the node convergence weights for each execution node in the discrete wolf pack algorithm.

[0040] S131, the main control module sends a wireless ranging command to the execution network through the system communication network. Each execution node, based on the wireless ranging command, periodically broadcasts radio frequency (RF) test frames using its second communication module. Each sensor node, in receiving mode, receives the RF test frames through its first communication module. The physical layer hardware within the sensor node extracts the received signal strength indication value when the RF test frame arrives. The sensor node packages the collected received signal strength indication values ​​from each execution node and sends them back to the main control module. The modulation and demodulation of the RF signal and the physical measurement of the received signal strength are well-known technologies in the field and will not be described in detail here.

[0041] S132, the signal processor within the main control module receives data packets from each sensor node and constructs a received signal strength indication matrix. In the indoor physical space, the electromagnetic energy of the radio frequency signal attenuates with increasing propagation distance; the received signal strength indication value conversely represents the equivalent spatial distance between nodes. To eliminate spatial multipath fading and measurement fluctuations caused by hardware thermal noise, the signal processor performs a moving average filtering operation on the multiple collected received signal strength indication values ​​to generate an equivalent received signal strength indication matrix. The filtering calculation formula is as follows: In the formula, Indicates the number of lines after filtering and smoothing. Each execution node and Equivalent received signal strength indication value between each sensing node; This indicates the total number of measurement samples taken within the test period; Indicates the index of the current measurement sampling batch; Indicates the first The first collection obtained The execution node and the first Instantaneous received signal strength indication value between sensor nodes.

[0042] S133, the signal processor traverses the equivalent received signal strength indication matrix and performs spatial proxy mapping calculation for each execution node. To avoid proxy mapping failure due to excessive spatial physical distance, the signal processor determines whether the value in the equivalent received signal strength indication matrix is ​​greater than a preset effective communication threshold before retrieval.

[0043] If all equivalent received signal strength indicators for a given execution node are below the effective communication threshold, the execution node is suspended and a topology anomaly is reported. If a value greater than or equal to the effective communication threshold exists, the signal processor uses the maximum signal strength principle to search the matrix for the maximum equivalent received signal strength indicator corresponding to that execution node and uses its corresponding sensor node as the binding target. The effective communication threshold is set based on the receiver sensitivity parameters of the wireless transceiver hardware.

[0044] When multiple sensor nodes have the same equivalent received signal strength indication value and are all at the maximum value, the signal processor selects the sensor node with the smaller index sequence number according to the preset arbitration rules to break the algorithm dead zone.

[0045] Based on the retrieval results, the signal processor establishes a binding mapping table between execution nodes and sensor nodes in system memory. The main control module also constructs a proxy adjacency topology between execution nodes based on this binding mapping table; execution nodes bound to the same sensor node are assigned to the same local neighborhood, and execution nodes bound to adjacent sensor nodes are assigned to adjacent local neighborhoods. The main control module determines the neighborhood information of each execution node in the subsequent swarm intelligence iteration process based on this proxy adjacency topology, realizing the topology reconstruction of the execution network. Since the execution nodes are not equipped with acoustic acquisition devices, the main control module assigns the direct sound energy ratio data corresponding to the bound sensor nodes to the corresponding execution nodes according to the binding mapping table. This mechanism utilizes the spatial attenuation characteristics of electromagnetic waves to characterize physical distance, compensating for the coverage gaps of the sound pickup devices in the hardware topology, enabling execution nodes to obtain acoustic environment benchmark data of their area, and establishing a data foundation for the subsequent allocation of node convergence weights in the discrete wolf pack algorithm.

[0046] After establishing the binding mapping relationship between the execution nodes and the sensing nodes, the main control module assigns node convergence weights of the discrete wolf pack algorithm to each execution node based on the binding acoustic characteristics. The specific implementation includes the following steps.

[0047] S141, the signal processor in the main control module retrieves the binding mapping table to obtain the first... The signal processor calculates the initial direct acoustic energy ratio of the sensor nodes it proxies and binds to. It then performs a linear normalization operation on the initial direct acoustic energy ratio, mapping it to a value between 0 and 1. To prevent division by zero errors and out-of-bounds values ​​during normalization, the signal processor presets a reference upper limit for the direct acoustic energy ratio. Compared with the reference lower limit value Furthermore, boundary truncation is applied before computation. Specifically, when the initial direct sound energy ratio is greater than the reference upper limit value... When, take the upper limit reference value. When the initial direct sound energy ratio is less than the reference lower limit value When, take the lower reference limit value. The value after boundary truncation is defined as the effective direct sound energy ratio. Normalization calculation formula: ; In the formula, Indicates the first Normalized acoustic feature values ​​corresponding to each execution node; Indicates the first The effective direct acoustic energy ratio corresponding to each execution node; This indicates the preset upper limit of the direct sound energy ratio reference; This indicates the preset direct sound energy ratio to the lower reference limit. The upper reference limit is... and reference lower limit value The maximum and minimum critical values ​​for the direct sound energy ratio in the corresponding space must be non-zero. Based on the physical attenuation characteristics of the indoor sound field, the upper limit of the direct sound energy ratio is defined as follows: The value range is configured to be from 10dB to 30dB, reaching the lower limit of the sound energy ratio reference value. The value range is configured to be -20dB to -10dB.

[0048] S142, the signal processor calculates the node convergence weights for the Discrete Wolf Pack Algorithm (DFP) for each execution node based on normalized acoustic eigenvalues. The DFP approximates the global optimum through collaborative search by multiple potential solution subjects. The degree of reverberation interference varies among nodes in the physical sound field. During spatial collaborative optimization, the convergence weights determine the step size of node state updates and the response weights for approaching the global optimum. A high direct sound energy ratio indicates less reverberation interference in the area, reflecting a higher accuracy in the true sound field; the corresponding execution node is assigned a higher convergence weight to act as the dominant node in the optimization process. A low direct sound energy ratio indicates a complex acoustic environment in the area; the corresponding convergence weight is reduced to suppress false triggering. The formula for calculating the node convergence weight is: ; In the formula, Indicates the first The node convergence weight of each execution node; This indicates the set upper limit parameter for the weight; This indicates the set lower limit parameter for the weight. The upper limit parameter for the weight. With weight lower bound parameter The convergence boundary for maintaining the stability of the algorithm's iteration is determined based on the transient response requirements of the system's photometric state adjustment. To prevent excessively low direct acoustic energy ratios in specific regions from causing the normalized acoustic eigenvalues ​​to reach zero, thus triggering the corresponding execution node's update step size to decrease to zero and halting the iteration, a lower weight limit parameter is used. Greater than zero. Weight cap parameter. The value range is configured to be from 0.8 to 1.0, and the lower limit parameter for weight is... The value range is configured to be from 0.1 to 0.3.

[0049] In step S143, the signal processor encapsulates the calculated node convergence weights into configuration data frames and distributes them to each execution node via the system communication network. The second communication module within each execution node receives the configuration data frames, and the microcontroller parses the data frames and extracts the node convergence weights. The microcontroller writes the node convergence weights into its internal non-volatile memory to permanently store the parameters. Simultaneously, the main control module writes the node convergence weights corresponding to each execution node into its node configuration table, or reads the node convergence weights from the corresponding execution node via the system communication network during online operation, for use in subsequent state approach update calculations.

[0050] Through this allocation mechanism, each execution node obtains algorithm iteration parameters that match the quality of its physical acoustic environment, completing the initial configuration for offline calibration and providing a static parameter basis for the system to enter online operation for dynamic photometric adjustments. The read / write control of non-volatile memory and the data frame encapsulation and parsing mechanism are well-known technologies in this field and will not be elaborated upon here.

[0051] The discrete wolf pack algorithm in this embodiment refers to a swarm intelligent iterative control mechanism that treats each execution node as a discrete individual, the target lighting state vector as the group's convergence target, uses surrogate adjacency topology constraints to define the neighborhood collaboration relationships between execution nodes, and adjusts the state update step size of each execution node using node convergence weights. During the iteration process, the main control module calculates the photometric state update for each execution node based on its corresponding node convergence weight, neighborhood topology relationships, and the target lighting state vector, enabling the execution network to form a collaborative convergence dynamic lighting response triggered by the spatial sound field.

[0052] See attached document Figure 4 After completing the initial configuration for offline calibration, the system enters online operation. The main control module controls the array signal acquisition and cross-correlation matrix construction, and the specific implementation includes the following steps.

[0053] S211, the sensor node is in online monitoring mode. When a target sound source signal is generated in the physical space, the sensor node uses its internally integrated microphone array to collect spatial sound waves. The microphone array contains multiple microphone units distributed according to a preset topology, which may include a uniform linear array or a uniform circular array. Each microphone unit synchronously picks up the target sound source signal and processes it into a multi-channel discrete-time acoustic sequence via an analog-to-digital converter. The sensor node transmits the discrete-time acoustic sequence to the main control module through the first communication module.

[0054] S212, the signal processor within the main control module receives multi-channel discrete-time domain acoustic sequences. To extract the frequency-domain spatial features of the sound source signal, the signal processor performs framing and windowing operations on the discrete-time domain acoustic sequence, and uses short-time Fourier transform to convert the time-domain data to the frequency domain. The positional differences of each microphone unit in physical space lead to different arrival times of sound waves, which manifests as phase differences at specific frequency points in the frequency domain. This phase difference is the basis for the system to resolve the spatial orientation of the sound source. Sound sources in real-world environments are mostly broadband signals. The signal processor calculates the spectral energy of each frequency band and selects effective frequency points with energy greater than a preset energy threshold as target frequencies. The signal processor obtains the frequency-domain signal vector at the target frequency. The dimension of the frequency-domain signal vector is consistent with the number of microphone units, representing the spatial phase and amplitude distribution of the array at that frequency point. Framing, windowing, and short-time Fourier transform processing of audio signals are well-known techniques in this field and will not be elaborated upon here.

[0055] S213, the signal processor constructs a spatial cross-correlation matrix based on the frequency domain signal vectors. The spatial cross-correlation matrix characterizes the statistical correlation of the multi-channel array signals in spatial distribution. To reduce random noise interference and obtain stable statistical characteristics, the signal processor performs a smooth average calculation of the autocorrelation tensor of the frequency domain signal vectors over multiple consecutive time frames. To prevent singular matrices caused by low signal-to-noise ratio or collinearity of multi-channel signals during subsequent matrix inversion operations, which could lead to data anomalies during inversion, the signal processor introduces diagonal loading terms on the main diagonal of the matrix. The formula for calculating the spatial cross-correlation matrix is: ; In the formula, Indicates target frequency Spatial cross-correlation matrix at the location; This represents the number of time-smoothing frames, determined based on the steady-state duration of the array signal, with a value range of 10 to 50. The index parameter represents the current time frame; Indicates the first Frame at target frequency The dimension is The frequency domain signal column vector, Indicates the number of microphone units; Represents the column vector of frequency domain signals The complex conjugate transpose; Indicates the diagonal loading coefficient; The dimension is The identity matrix. Diagonal loading coefficients. The value is greater than zero, and its value is set based on the measured average variance of the system's environmental noise floor energy. The signal processor will calculate the generated spatial cross-correlation matrix. The data is stored in the system cache of the main control module, providing a computational basis for subsequent sound source spatial spectrum estimation and angle of arrival analysis.

[0056] After completing the construction of the spatial cross-correlation matrix, the main control module performs eigenvalue decomposition and principal component purification. The specific implementation includes the following steps.

[0057] S221, the signal processor within the main control module retrieves the spatial cross-correlation matrix from the system buffer and performs eigenvalue decomposition. Based on the subspace orthogonality principle of array signal processing, the cross-correlation matrix is ​​decomposed into a signal subspace spanned by the target sound source and a noise subspace spanned by the ambient background noise; these two subspaces are mutually orthogonal. The eigenvalue decomposition operation decouples the multi-channel array coupled signal into orthogonal spatial eigenvectors and their corresponding eigenvalues. Calculation formula: In the formula, Represents the spatial cross-correlation matrix; This represents a unitary matrix containing all spatial eigenvectors; This represents a diagonal matrix composed of eigenvalues; This represents the complex conjugate transpose of a unitary matrix. The algebraic operations of matrix eigenvalue decomposition are well-known techniques in this field and will not be elaborated upon here.

[0058] S222, the signal processor from the diagonal matrix Feature values ​​are extracted and sorted in descending order of numerical value. Larger feature values ​​correspond to the spatial signal subspace where the target sound source is located, while smaller feature values ​​correspond to the spatial noise subspace where the ambient noise is located. The signal processor determines the dimension of the signal subspace based on predefined information theory criteria, including the Akaike information criterion or the minimum description length criterion. To prevent sudden changes in hardware thermal noise from causing confusion in feature value distribution and thus leading to subspace partitioning errors, the signal processor checks whether the ratio of the largest to the smallest feature value is greater than a preset signal-to-noise separation threshold before partitioning. If this ratio is less than or equal to the preset signal-to-noise separation threshold, it indicates that the current frame data is covered by background noise or there is no valid target sound source. The main control module discards the frame data and terminates the direction-finding calculation for the current cycle. The signal-to-noise separation threshold is set based on prior environmental signal-to-noise ratio data for the application scenario, with a value range of 3 to 5.

[0059] S223, after passing the threshold determination, the signal processor separates the signal subspace and noise subspace according to the determined signal subspace dimension. The signal processor extracts the feature vectors corresponding to the noise subspace to construct a noise subspace matrix. The extraction process strips away the signal feature vectors representing the principal components, purifying the noise features and providing an orthogonal basis for the subsequent construction of the spatial pseudospectrum. The construction logic of the noise subspace matrix is ​​as follows: ; In the formula, This represents the purified noise subspace matrix; The dimension of the defined signal subspace represents the number of concurrent independent target sound sources in the space; This indicates the total number of microphone units in the microphone array; This indicates the order after sorting in descending order. The spatial feature vector corresponding to each feature value, index The range is to The signal processor will construct the noise subspace matrix. The data is written into an internal temporary register to complete the principal component purification process and to establish a feature mapping matrix for the spectral peak search function that generates the spatial arrival angle.

[0060] S224, the signal processor constructs the array steering vector corresponding to the candidate direction based on the geometric topological coordinates of the microphone array, and constructs a spatial pseudospectral function using the noise subspace matrix. The signal processor calculates the spatial pseudospectral value point-by-point within a preset angle search range, determines the angle corresponding to the peak value of the spatial pseudospectral function as the spatial angle of arrival of the target sound source, and transforms this spatial angle of arrival to the physical space global coordinate system based on the installation attitude calibration parameters of the sensor node. When multiple spectral peaks exist in the spatial pseudospectral function, and the amplitude difference between the largest and second-largest spectral peaks is greater than or equal to a preset peak separation threshold, the signal processor selects the angle corresponding to the largest spectral peak as the spatial angle of arrival of the main target sound source; when the amplitude difference between the largest and second-largest spectral peaks is less than the preset peak separation threshold, the signal processor maintains the sound source spatial angle of arrival from the previous time frame, or marks the current frame as a multi-source uncertain state to avoid frequent jumps in the targeted broadcast group.

[0061] After completing subspace purification and spectral estimation feature construction, the main control module simultaneously performs energy envelope reconstruction and spectral centroid frequency value extraction on the extracted sound source signal to obtain time-frequency domain features characterizing the dynamic changes of the sound field. The specific implementation includes the following steps.

[0062] S231, the signal processor within the main control module extracts the multi-channel discrete-time domain acoustic sequences acquired by the sensor nodes and performs backbone audio energy envelope reconstruction. The signal processor extracts the signal eigenvector corresponding to the largest eigenvalue of the spatial cross-correlation matrix and uses this eigenvector as a spatial weighting vector to weight and synthesize the multi-channel discrete-time domain acoustic sequences, obtaining the backbone audio sequence. The signal processor performs windowed sum-of-squares processing on the backbone audio sequence to obtain the short-time energy distribution characteristics of the sound source signal. The windowed sum-of-squares calculation process of the short-time energy is equivalent in the time domain to performing low-pass smoothing filtering on the squared discrete signal, thereby reconstructing a smooth output energy envelope, characterizing the fluctuation state of sound wave energy in physical space. Calculation formula: ; In the formula, Indicates the first The energy envelope value of the frame; Indicates the length of the analysis window; Indicates the amplitude of the backbone audio sequence; Indicates the frame shift length between adjacent frames; Indicates the index of the sampling points within the window; The index parameter representing the current time frame. The length of the analysis window. Based on the minimum residence time of the acoustic signal, the value range is configured to be 256 to 1024 sampling points.

[0063] S232, the signal processor retrieves the amplitude spectrum data of the sound source signal after conversion to the frequency domain and performs centroid frequency value extraction. The centroid frequency value represents the position of the center of gravity of sound energy on the frequency axis, reflecting the offset state of the sound source's spectral energy distribution. Different physical sound sources have different spectral characteristics; the energy distribution of mechanical background noise is in the low-frequency band, while the energy distribution of sudden ambient noise is in the high-frequency band. This feature is used to distinguish between low-frequency interference and high-frequency target sound. The signal processor calculates the weighted average of the center frequency and corresponding amplitude of each frequency band in the amplitude spectrum. To avoid the algorithm malfunction caused by the sum of spectral energy being zero due to the processing frame being in a silent segment, the signal processor introduces a positive-zero energy bias constant in the denominator term. Calculation formula: ; In the formula, Indicates the first The centroid frequency value of the frame's spectrum; Indicates the total number of effective frequency bands in the spectrum; Indicates frequency band index; Indicates the first The center frequency of each frequency band; Indicates the first Frame in Amplitude spectrum values ​​at each frequency band; This represents the energy bias constant. The value is determined based on the quantization lower limit of the hardware noise floor of the system's analog-to-digital conversion.

[0064] In step S233, the signal processor concatenates the calculated energy envelope value with the spectral centroid frequency value to construct the acoustic time-frequency feature vector of the target sound source. To eliminate ambient noise interference, the signal processor compares the energy envelope value with a preset energy activation threshold. If the energy envelope value is greater than the preset energy activation threshold, the signal processor marks the acoustic time-frequency feature vector as validly activated and stores it in the system memory of the main control module, providing a feature basis for sound field state classification and illumination photometric adjustment; if the energy envelope value is less than or equal to the preset energy activation threshold, the signal processor determines that the current frame is invalid and discards the feature vector of that frame. The energy activation threshold is set based on the average steady-state ambient noise energy in the physical space, with a value range of -40dBFS to -20dBFS. The data processing operation of feature concatenation is a well-known technology in this field and will not be described in detail here.

[0065] After the system obtains the target lighting requirements, the main control module controls the system to perform colorimetric space coordinate mapping transformation, converting colorimetric indicators into underlying hardware driver parameters of the execution node. The specific implementation includes the following steps.

[0066] S311, the signal processor within the main control module acquires the CIE 1931 chromaticity coordinates and target luminance value corresponding to the target illumination state. The CIE 1931 chromaticity system is built upon human visual perception characteristics, expressing color brightness and chromaticity separately. To achieve the conversion to absolute physical stimulus quantities, the signal processor converts the chromaticity coordinates and target luminance value into CIE 1931 tristimulus values. During the conversion process, to prevent algorithmic anomalies caused by division by zero when the chromaticity coordinate ordinate approaches zero, the signal processor compares the input ordinate with a preset minimum ordinate threshold. When the ordinate is less than or equal to the minimum ordinate threshold, the ordinate is truncated and assigned the minimum ordinate threshold value. The formula for calculating the tristimulus value is as follows: ; ; In the formula, This represents the first stimulus value after conversion; This represents the converted value of the third stimulus; The target brightness value is used as the secondary stimulus value. Represents the x-coordinate of the target chromaticity coordinates; This represents the ordinate of the target chromaticity coordinates after truncation. The minimum ordinate threshold is fixed at 0.001.

[0067] In step S312, the signal processor maps the CIE 1931 tristimulus values ​​to the primary color intensity space corresponding to the execution node. Since the LED chips equipped in different execution nodes have different physical spectral characteristics, the signal processor retrieves a preset chromaticity representation matrix that matches the target execution node. This chromaticity representation matrix is ​​constructed from the test data of the red, green, and blue tristimulus values ​​of each primary color light source within the target execution node, and it is a non-singular matrix. The signal processor constructs a channel inverse mapping matrix by inverting the chromaticity representation matrix, converting the tristimulus values ​​into the red, green, and blue primary color control intensities. The calculation formula is as follows: ; In the formula, The dimension is The primary color control intensity vector, whose elements represent the control intensity of the three primary colors: red, green, and blue; This represents the channel inverse mapping matrix, which is the chromaticity representation matrix. Inverse matrix; The dimension is The tristimulus value vector, by , and The algebraic transformation mechanisms for matrix inversion and multiplication are well-known techniques in this field and will not be elaborated upon here.

[0068] The S313 signal processor performs nonlinear inverse gamma correction on the control intensity of the red, green, and blue primary colors and maps it to a pulse width modulation duty cycle signal. Due to the nonlinear characteristics between the luminous intensity and driving current of the LED chip, and the human visual perception of brightness, the signal processor introduces an electro-optical conversion correction coefficient for compensation. When the physical color corresponding to the target chromaticity coordinates exceeds the actual luminous color gamut of the execution node, the primary color control intensity output by the channel inverse mapping matrix is ​​negative. Directly performing exponentiation will lead to an algorithm anomaly with no real solution. To prevent the calculated control intensity from exceeding the hardware driving capability or causing anomalies in the negative value domain, the signal processor limits the control intensity of each primary color to between 0 and a preset maximum intensity threshold before correction. The formula for calculating the pulse width modulation duty cycle is as follows: ; In the formula, Indicates the first The pulse width modulation duty cycle of the roadbed color channel ranges from 0 to 1; This represents the channel index parameter, corresponding to the red, green, and blue channels; Represents the primary color control intensity vector The Middle Control intensity value of road passage; This indicates the preset maximum intensity threshold, which is set according to the full-scale value of the drive circuit. This represents the gamma correction factor, with a value range of 2.0 to 2.4. The signal processor transmits the converted pulse width modulation duty cycle configuration data for each channel to the execution node via the system network. The microcontroller within the execution node outputs the corresponding hardware drive current to complete the physical-level chroma and brightness adjustment.

[0069] After acquiring the acoustic time-frequency feature vector and the spatial orientation of the sound source, the main control module performs photometric brightness envelope mapping and targeted broadcasting to achieve spatial dynamic lighting adjustment with sound and light linkage. The specific implementation method includes the following steps.

[0070] S321, the signal processor within the main control module extracts the energy envelope value from the system memory and performs photometric brightness envelope mapping. Based on the principle of cross-modal perception mapping in psychophysics, this step converts the sound pressure energy distribution perceived by hearing into the change in ambient brightness perceived by vision. To establish a physical correlation between acoustic energy and optical brightness, the signal processor directly maps the energy envelope value to the target brightness value, making the target brightness value the sole brightness input in the subsequent CIE 1931 tristimulus value conversion, avoiding repeated scaling of brightness during the pulse width modulation stage. To prevent sudden environmental noise from causing the calculated target brightness value to exceed the threshold, triggering hardware anomalies such as drive power overload or duty cycle data overflow, the signal processor performs upper limit truncation on the input energy envelope value, limiting the dynamic range of the input energy and ensuring that the generated target brightness value is distributed within the set threshold range. The formula for calculating the target brightness value is: ; In the formula, Indicates the first The target brightness value of the frame; Indicates the base sustained brightness value; Indicates the acousto-optic conversion gain constant; Indicates the first The energy envelope value of the frame; This indicates the preset energy saturation cutoff threshold.

[0071] S322, the signal processor determines the target node group for targeted broadcasting based on the spatial angle of arrival of the sound source obtained through analysis and the system's pre-stored spatial topology coordinate table. The spatial topology coordinate table records the absolute three-dimensional coordinates of each execution node in the physical space.

[0072] The signal processor calculates the spatial vector pointed to by the spatial angle of arrival of the sound source and divides the physical space into multiple illumination sectors. The signal processor selects illumination sectors whose angle with the spatial vector of the sound source is less than a preset beamwidth and marks the execution nodes within these sectors as targeted broadcast groups. To prevent the execution nodes from repeatedly switching states and causing visual flickering of the illumination when the target sound source moves at the edge of the sector, the signal processor sets up spatial hysteresis buffers between adjacent illumination sectors.

[0073] The signal processor sets activation and release boundary angles for each lighting sector, and the difference between the two angles constitutes a spatial hysteresis buffer. When the sound source spatial vector crosses the activation boundary angle from the outside to the inside, the execution node is included in the targeted broadcast group; when the sound source spatial vector does not cross the release boundary angle from the inside to the outside, the signal processor maintains the targeted broadcast group state of the previous time frame, thus maintaining a smooth transition in lighting control.

[0074] S323, the signal processor encapsulates the pulse width modulation duty cycle of each primary color channel corresponding to the calculated target illumination state vector into an instruction, generating a targeted broadcast control data frame. The targeted broadcast control data frame includes a frame header, group address identifier, duty cycle configuration data, and a checksum. The targeted broadcast control data frame may also include a target brightness value as a state check or display synchronization parameter, but this target brightness value is no longer used to perform double multiplication scaling on the duty cycle.

[0075] The signal processor sends targeted broadcast control data frames to the corresponding execution nodes within the targeted broadcast group via the system network. For execution nodes not included in the targeted broadcast group, the main control module sends a hold control frame or a basic walk control frame to maintain the photometric state of the previous frame, or to slowly return to the basic anchor point state according to the step-limited walk logic under steady-state conditions. The microcontroller within the execution node receives the targeted broadcast control data frames and performs data verification and parsing. After receiving the targeted broadcast control data frames, the microcontroller extracts the pulse width modulation duty cycle of each primary color channel, writes the duty cycle into the comparison register of the corresponding hardware timer, and outputs the updated hardware drive current.

[0076] See attached document Figure 5 After acquiring the acoustic time-frequency feature vector, the main control module synchronously executes the state vector definition and sound field transient discrimination logic to establish a mathematical benchmark characterizing the evolution law of the sound field. The specific implementation method includes the following steps.

[0077] S411, the signal processor within the main control module acquires the energy envelope value and spectral centroid frequency value of the current time frame, and incorporates historical feature data from previous frames to construct a time-frequency state vector. Acoustic events in physical space exhibit specific energy fluctuations and spectral distribution patterns. Combining the energy envelope and spectral centroid frequency value allows for a joint characterization of the sound source's evolutionary trends in loudness and timbre, providing feature support for subsequent state recognition. The formula for constructing the time-frequency state vector is as follows: ; In the formula, Indicates the first The time-frequency state vector of a frame; Indicates the first The energy envelope value of the frame; Indicates the first The energy envelope value of the frame; Indicates the first The centroid frequency value of the frame's spectrum; Indicates the first The centroid frequency value of the frame's spectrum; This indicates the length of the time smoothing window and represents the number of historical data frames introduced. Represents the matrix transpose symbol; This parameter represents the index of the current time frame. The length of the time smoothing window is configured to range from 3 to 8. The data assembly mechanism using vector transposition and multidimensional array storage is well-known in the field and will not be elaborated upon here.

[0078] In S412, the signal processor performs transient sound field discrimination using the constructed time-frequency state vector. Sudden acoustic events can cause the current frame energy to deviate from the long-term background noise floor. By calculating the relative deviation ratio, interference from steady-state background noise can be eliminated. The signal processor calculates the absolute value of the difference between the energy envelope value of the current time frame and the long-term moving average background energy of the previous frame, and extracts the ratio of this absolute value to the long-term moving average background energy of the previous frame as the transient discrimination factor. To prevent the long-term moving average background energy from approaching zero due to a silent physical space, which could cause a division-by-zero error in the algorithm, the signal processor introduces a positive-zero stable bias constant in the denominator. The formula for calculating the transient discrimination factor is as follows: ; In the formula, Indicates the first Transient discrimination factor of the frame; Indicates the first The energy envelope value of the frame; express Long-term moving average background energy of the frame; This represents the stable bias constant. The value of the stable bias constant is set according to the system hardware quantization precision, and is configured to be 0.0001.

[0079] To avoid sudden transient noise contaminating the estimated background energy and causing the algorithm to lose sensitivity to subsequent acoustic events, the signal processor updates the long-term moving average background energy using a first-order recursive filtering algorithm with a smoothing forgetting factor after calculating the transient discrimination factor. The update calculation formula is as follows: In the formula, Indicates the updated number Long-term moving average background energy; This represents the smoothing forgetting factor. The smoothing forgetting factor is configured to range from 0.95 to 0.99, and is used to constrain the weights of historical states, maintaining the slowly varying characteristics of the background energy estimation.

[0080] S413, the signal processor compares the transient discrimination factor with the preset transient change threshold to determine the transient attributes of the current sound field and updates the sound field state indicator bit. If the transient discrimination factor is greater than the preset transient change threshold, it indicates that a sudden acoustic event exists in the sound field, and the signal processor configures the sound field state indicator bit to the first logic state, representing that the sound field is in a transient excitation state; if the transient discrimination factor is less than or equal to the preset transient change threshold, it indicates that the sound field is in an energy stable state, and the signal processor configures the sound field state indicator bit to the second logic state, representing that the sound field is in a steady-state background state. The transient change threshold is set based on the fluctuation variance data of the steady-state noise floor in the physical space, and the value range is configured to be 1.5 to 3.5. The signal processor associates the updated sound field state indicator bit with the time-frequency state vector and stores it in the system memory, providing a trigger determination basis for lighting strategy control.

[0081] After updating the sound field status indicator bit, the signal processor in the main control module performs step-limited walk control for the sound field in a steady background state in order to maintain the dynamic naturalness of the spatial light environment when there are no acoustic trigger events. The specific implementation includes the following steps.

[0082] S421, the signal processor retrieves the sound field state indicator bit from the system memory. When the sound field state indicator bit is in the second logic state representing the steady-state background state, the signal processor activates the step-limited walk control logic. Since long-term static fixed lighting easily induces visual fatigue and adaptive passivation, the signal processor introduces low-frequency random perturbations into the basic lighting parameters to simulate the gradual fluctuations of natural light. Specifically, the signal processor calls its internal pseudo-random sequence generator to generate a three-dimensional random perturbation vector with values ​​limited to between -1 and +1 and following a uniform distribution, providing independent walk excitations in the CIE 1931 chromaticity horizontal axis, chromaticity vertical axis, and photometric luminance dimensions. The pseudo-random number generation algorithm uses a conventional linear congruence generator or Mason rot algorithm; its sequence generation mechanism is well-known in the field and will not be elaborated upon here.

[0083] S422, the signal processor calculates the intermediate illumination state vector based on the 3D random perturbation vector of the current frame. To avoid wander divergence caused by the continuous accumulation of perturbations in a single dimension, the signal processor introduces a mean regression mechanism. This mechanism calculates the difference between the base anchor vector and the target illumination state vector of the previous frame and applies a center recovery coefficient to guide the wandering coordinates towards the anchor point. The formula for calculating the intermediate illumination state vector is as follows: ; In the formula, Indicates the first The intermediate lighting state vector of the frame; Indicates the first The target illumination state vector of the frame; This represents the preset base anchor point vector, which includes the set reference chromaticity coordinates and reference luminance values; This represents the center recovery coefficient, used to control the strength of the directional force moving towards the foundation anchor point, with a value range of 0.01 to 0.05. This represents the walk step gain, which is set based on the perceptible difference between brightness and chromaticity in the human eye and is used to constrain the incremental boundary of a single random perturbation. Indicates the first The three-dimensional random perturbation vector of the frame; The index parameter represents the current time frame.

[0084] S423, the signal processor performs boundary truncation on the intermediate illumination state vector and outputs the target illumination state vector of the current frame. The truncation process includes independent dimensional constraints and chromaticity joint constraints: the signal processor first restricts the components of the intermediate illumination state vector in the chromaticity x-coordinate, chromaticity y-coordinate, and luminance dimensions to preset lower and upper thresholds, respectively. The upper and lower thresholds for the chromaticity coordinate components are determined by the vertices of the color gamut polygon of the target execution node, and the upper threshold for the luminance component is set to the maximum rated lumen value of that node. Subsequently, the signal processor calculates the sum of the restricted chromaticity x-coordinate and chromaticity y-coordinate. If this sum is greater than or equal to a preset chromaticity physical boundary threshold, the x-coordinate and y-coordinate are truncated proportionally using the ratio of this physical boundary threshold to the current sum. The chromaticity physical boundary threshold is configured to 0.999.

[0085] In step S424, the signal processor uses the generated target illumination state vector as the target illumination index for the current frame, overwriting the original data in the system memory. The signal processor parses the target illumination state vector to obtain the updated CIE 1931 chromaticity coordinates and target luminance value, and sends them to the chromaticity space coordinate mapping module. After being solved by the channel inverse mapping matrix, the pulse width modulation duty cycle of each primary color channel is calculated. Finally, the microcontroller of the execution node controls the driver module to output the updated hardware drive current according to the pulse width modulation duty cycle issued by the main control module, completing the dynamic wandering adjustment of the light environment.

[0086] When the signal processor in the main control module detects that the sound field status indicator bit has switched to the first logic state, it suspends the steady-state walk logic and executes physical weight approach control under transient jumps to realize the dynamic response of the light environment to sudden sound field events. The specific implementation method includes the following steps.

[0087] S431, the signal processor retrieves the time-frequency state vector of the current frame, extracts the spectral centroid frequency value and energy envelope value, and establishes a cross-modal mapping from acoustic features to optical state. The signal processor applies bandpass truncation to the spectral centroid frequency value, limiting it to a set effective audible frequency band to prevent frequency exceedance from causing the mapped coordinates to overflow the physical luminous color gamut. The lower limit frequency of the effective audible frequency band is configured as 300 Hz, and the upper limit frequency is configured as 8000 Hz. Subsequently, the signal processor uses a mapping transformation function to linearly map the truncated spectral centroid frequency value to a correlated color temperature value; the linear mapping calculation process is as follows: calculate the difference between the current frequency value and the lower limit frequency, divide by the interval span between the upper and lower limit frequencies to obtain a frequency domain normalization coefficient, then multiply the frequency domain normalization coefficient by the set maximum and minimum color temperature difference and add the minimum color temperature value. The signal processor converts the correlated color temperature value into target chromaticity coordinate components in the CIE 1931 chromaticity space according to the Planck trajectory equation, and combines it with the target luminance component mapped by the current frame energy envelope value to generate a three-dimensional transient target illumination vector. The conversion mechanism of deriving Planck trajectory chromaticity coordinates from correlated color temperature values ​​is a well-known technique in the field and will not be elaborated here.

[0088] S432, the signal processor retrieves the transient discrimination factor output by the sound field transient discrimination logic and calculates the approach control weight for the current frame. The approach control weight determines the rate at which the current illumination state approaches the transient target illumination vector. The magnitude of the transient discrimination factor characterizes the energy step amplitude of the acoustic event. The signal processor uses a nonlinear saturation function to map the transient discrimination factor to normalized approach control weights. The formula for calculating the approach control weights is as follows: ; In the formula, Indicates the first Frame proximity control weights; Indicates the first Transient discrimination factor of the frame; This represents the response sensitivity coefficient, used to adjust the sensitivity of lighting changes to sudden sound changes, with a value range of 0.5 to 1.5. Represents the natural constant; The index parameter represents the current time frame.

[0089] S433, the signal processor uses the calculated approach control weights, combined with the first... The node convergence weight of the execution node and the Euclidean distance between the lighting state vector of the previous frame and the transient target lighting vector of the current frame are used to calculate the th execution node. The comprehensive convergence coefficient of the execution node is calculated, and based on the comprehensive convergence coefficient, interpolation smoothing is performed on the target illumination state vector of the previous frame and the transient target illumination vector of the current frame. The 1st execution node is then calculated. The target lighting state vector of each execution node in the current frame.

[0090] The comprehensive convergence coefficient is limited to a value between 0 and 1, and increases with the increase of node convergence weight, convergence control weight, and Euclidean distance. In one implementation, the comprehensive convergence coefficient is obtained by multiplicatively or weightedly combining the node convergence weight, convergence control weight, and distance response term, and is limited to a value between 0 and 1 by boundary truncation. The node convergence weight reflects the acoustic signal quality of the region where the corresponding execution node is located, the convergence control weight reflects the intensity of the transient change in the current sound field, and the Euclidean distance reflects the degree of deviation between the current illumination state and the transient target illumination vector. Through the physical weight convergence mechanism, the algorithm enables the illumination state to smoothly move towards the acoustically mapped target when a transient jump occurs, preventing the hardware drive signal from stepping and triggering power overload protection. The update calculation formula for the target illumination state vector of the current frame is as follows: The current frame target illumination state vector of each execution node is obtained by interpolating the target illumination state vector of the previous frame and the transient target illumination vector of the current frame using a comprehensive convergence coefficient: ; In the formula, Indicates the first The execution node is at the The target illumination state vector of the frame; Indicates the first The execution node is at the The target illumination state vector of the frame; Indicates the first The transient target illumination vector of the frame; Indicates the first The execution node is at the The overall convergence coefficient of the frame.

[0091] In step S434, the signal processor performs boundary truncation on the target illumination state vector of the current frame. The signal processor employs a dual-judgment logic of independent dimension constraints and chromaticity joint constraints under steady-state conditions, ensuring that the calculated color coordinates and luminance are within the physically realizable domain of the execution node. The signal processor updates the processed target illumination state vector to system memory and decouples it for input to the chromaticity space coordinate mapping and transformation module. The channel inverse mapping matrix solves the target illumination state vector into the pulse width modulation duty cycle of each primary color channel. The microcontroller of the execution node controls the driver module to output the updated hardware drive current based on the pulse width modulation duty cycle issued by the main control module, completing the transient jump control of the physical-level audio-visual linkage.

[0092] See attached document Figure 6 After determining the target lighting state vector, the signal processor in the main control module synchronously performs control deviation calculation and fractional-order controller construction to achieve smooth approximation control of the lighting state. The specific implementation includes the following steps.

[0093] In S511, the signal processor within the main control module retrieves the target illumination state vector for the current time frame from system memory and obtains the actual illumination state vector for the current time frame acquired by the photoelectric feedback module of the execution node. The signal processor performs a subtraction calculation on the corresponding dimensions of the target illumination state vector and the actual illumination state vector to obtain the control deviation vector for the current time frame. Both the target illumination state vector and the actual illumination state vector contain chromaticity horizontal axis components, chromaticity vertical axis components, and luminance components. Calculating these components independently ensures that the control deviation remains decoupled in the physical dimensions of chromaticity and luminance. The formula for calculating the control deviation vector is as follows: ; In the formula, Indicates the first The execution node is at the The frame's control deviation vector; Indicates the first The execution node is at the The target illumination state vector of the frame; Indicates the first The execution node is at the The actual lighting state vector of the frame; This parameter represents the index of the current time frame. Subtraction matrix operations on multidimensional vectors can be performed using a standard register bitwise subtraction procedure. For ease of description, the subsequent fractional-order controller construction process will be illustrated using a single execution node as an example, and the execution node index will be omitted. Its data processing mechanism is a well-known technology in this field and will not be elaborated here.

[0094] In S512, the signal processor constructs a fractional-order calculus operator based on a finite memory length, using the calculated control deviation vector. Fixed integer-order control is prone to overshoot and light intensity oscillations when dealing with the nonlinear electro-optic conversion characteristics of the light source. Introducing a fractional-order operator allows for the use of its historical memory characteristics to provide damping constraints. Fractional calculus possesses a slowly decaying memory capability for historical error weights. To prevent system memory overflow caused by the accumulation of historical memory, the signal processor employs a truncated Grünwald-Letnikov time-domain discrete approximation algorithm. The signal processor allocates a sliding buffer in system memory to dynamically store the control deviation vector of a specific number of historical frames. Simultaneously, the signal processor introduces the discrete sampling period of the control system as a time scale reference, giving the operator physical dimensions. The calculation formula for the fractional-order calculus operator is as follows: ; In the formula, Indicates the first Frame in The values ​​of fractional calculus terms in each dimension; Indicates the fractional order of the corresponding control element; This represents the discrete sampling period of a digital control system. This represents a time-scale scaling factor constructed by combining the sampling period and the fractional order. This indicates the set finite memory truncation length, with a value range of 5 to 15; Indicates the first The first frame of the control deviation vector The component values ​​of each dimension; Indicates the time offset index of the historical frame; Let represent the binomial coefficient term. The binomial coefficient term is solved using a recursive formula, which is: and initial value .

[0095] In S513, the signal processor linearly combines the calculated fractional integral and proportional terms to construct a fractional controller, generating a control correction vector for the actuator. This control correction vector is used for closed-loop correction of the actual lighting state vector. The signal processor superimposes the control correction vector onto the actual lighting state vector of the current time frame to obtain the corrected lighting state vector, which is then input into the linear decoupling and pulse width modulation output process of the drive signal. The signal processor calculates the fractional integral and fractional derivative terms for each dimension of the control deviation vector and reconstructs them into a multi-dimensional control correction vector after scalar combination. The fractional controller adjusts the gain constants and orders of each component to make the illuminance dynamic tracking curve of the actuator node approach the critical damping state, suppressing visual flicker during the lighting state adjustment process. The formula for calculating the control correction vector is as follows: ; In the formula, Indicates the first The frame's control correction vector; Indicates the first The frame's control deviation vector; Indicates the first The fractional integral vector of a frame, whose internal dimensional components are derived from fractional calculus operators of order 1. Calculated time-by-time component This indicates the order of the fraction in the integration process, with a value range of 0.2 to 0.8. Indicates the first The fractional differential vector of a frame, whose internal dimensional components are derived by fractional calculus operators of order 1. Calculated time-by-time component This represents the fractional order of the differential term, with a value range of 0.4 to 0.9. Represents the proportional gain constant; Represents the integral gain constant; This represents the differential gain constant. The proportional gain constant, integral gain constant, and differential gain constant are tuned and set using the critical proportional gain method based on the hardware response hysteresis time of the execution node in the physical space. The signal processor converts the corrected lighting state vector into a pulse width modulation duty cycle and sends it to the execution node, driving its internal microcontroller to adjust the lighting output.

[0096] When constructing a fractional-order controller, the signal processor in the main control module synchronously executes dynamic adjustment logic of the calculus order based on the proportion of high-frequency components. By analyzing the frequency domain energy distribution characteristics of the acoustic signal, it adaptively corrects the order parameters of the fractional-order calculus operator. The specific implementation includes the following steps.

[0097] S521, the signal processor in the main control module acquires the acoustic power spectral density data of the current time frame, calculates the total energy in the high-frequency band and the total energy across the entire frequency band, and extracts the proportion of high-frequency components. High-frequency transient acoustic events cause energy concentration in the high-frequency band of the frequency domain; quantifying the proportion of high-frequency components is used to characterize the concentrated distribution of energy in the sound field frequency domain. To prevent the system from outputting random values ​​due to microphone hardware noise interference in a silent state, the signal processor determines whether the total energy across the entire frequency band is greater than a set effective noise floor threshold before calculation; if it is less than or equal to the effective noise floor threshold, the proportion of high-frequency components in the current time frame is set to zero; if it is greater than the effective noise floor threshold, a division calculation of the proportion of high-frequency components is performed. Simultaneously, to prevent the denominator from approaching zero and causing calculation overflow, the signal processor adds a stabilizing anti-bias constant to the denominator. The formula for calculating the proportion of high-frequency components is: ; In the formula, Indicates the first The proportion of high-frequency components in the frame, Indicates the first The frame acoustic signal in the first Power spectral density values ​​at discrete frequency points; This indicates the frequency index corresponding to the highest cutoff frequency across the entire frequency band, set according to the system's discrete sampling rate. This indicates the starting frequency index of the set high-frequency band, corresponding to a physical frequency configuration of 4000 Hz; This represents the stable anti-bias constant, with a value configured as 0.0001; This represents the index parameter of the current time frame. The Discrete Fourier Transform and power spectral density solution mechanism are well-known techniques in this field and will not be elaborated upon here.

[0098] S522: Based on the calculated proportion of high-frequency components, the signal processor uses a nonlinear activation function to calculate the order adjustment factor and solves for the target order of the integral and derivative components in the current time frame. When the acoustic signal exhibits high-frequency dominance, the system increases the proportion of the derivative component to provide greater damping force and suppress overshoot during the illumination response; simultaneously, it reduces the proportion of the integral component to weaken the integral saturation effect. The signal processor establishes a reverse-linked order mapping model, and the formula for calculating the target order is as follows: ; ; ; In the formula, Indicates the first The order adjustment factor of the frame, whose value range is mapped between 0 and 1; Indicates the first The proportion of high-frequency components in a frame; This indicates the set center threshold for the percentage, configured as 0.35; This represents the mapping gain slope, used to control the steepness of the nonlinear transition, and is configured to 10.0; Indicates the first The target order of the frame integration process; and These represent the upper and lower limits of the order of the integral element, respectively, and are configured as 0.8 and 0.2 in conjunction with the stability boundary of the control system. Indicates the first The target order of the differential element of the frame; and These represent the upper and lower limits of the order of the differential element, respectively, and are configured as 0.9 and 0.4.

[0099] S523, the signal processor performs time-domain smoothing filtering on the calculated target order, outputting the updated fractional orders of the integral and derivative elements. The inter-frame fluctuations of acoustic characteristics exhibit transient jumps; directly substituting the target order into the fractional-order operator causes discontinuous jumps in the controller's internal state variables, leading to sudden changes in the drive output current. The signal processor employs a first-order low-pass filtering algorithm with a forgetting mechanism to constrain the order. The smoothing update formula is as follows: ; ; In the formula, λ and They represent the first Frame and the The order of the integral step after frame update; and They represent the first Frame and the Fractional order of the differential element after frame update; This represents the order smoothing coefficient, used to determine the degree of hysteresis in order evolution, with a value range of 0.85 to 0.95. The signal processor will smooth the updated fractional order of the integrator. Fractional order of the differential element Overwrite it into system memory to replace the fixed-order parameters in the Grünwald-Laitnikov discrete approximation operator, so that the solution of the multidimensional control correction vector has dynamic damping characteristics that adapt to the acoustic frequency domain distribution.

[0100] After constructing a fractional-order controller and calculating the multidimensional control correction vector, the signal processor in the main control module executes the linear decoupling and pulse width modulation output logic of the drive signal, converting the digital control quantity in the color space into the underlying electrical signal that the microcontroller directly drives the light-emitting hardware. The specific implementation includes the following steps.

[0101] S531, the signal processor within the main control module acquires the corrected illumination state vector generated after closed-loop correction by the fractional-order controller. The corrected illumination state vector includes chromaticity horizontal axis components, chromaticity vertical axis components, and luminance components. The signal processor uses this as the illumination state quantity to be converted under the CIE 1931 chromaticity system. The CIE 1931 chromaticity system belongs to a non-uniform visual psychophysical quantity space. To suppress the non-linear modulation effect of the non-linear chromaticity space on the control gain, the signal processor converts the corrected illumination state components into tristimulus values ​​in a linearly mapped space.

[0102] The conversion calculation involves a division operation with the chromaticity ordinate component in the corrected lighting state vector as the denominator. To avoid division by zero errors when the chromaticity ordinate component approaches zero, the signal processor performs lower limit protection on the chromaticity ordinate component in the corrected lighting state vector before conversion. The lower limit protection process is as follows: compare the set lower limit threshold for ordinate protection with the chromaticity ordinate component in the corrected lighting state vector, and take the maximum value of the two as the chromaticity ordinate component after limit protection; the set lower limit threshold for ordinate protection is configured to 0.001.

[0103] Subsequently, the signal processor calculates the first stimulus value, which is equal to the chromaticity abscissa component in the corrected lighting state vector divided by the chromaticity ordinate component after limiting protection, and then multiplied by the luminance component in the corrected lighting state vector. The signal processor calculates the third stimulus value, which is equal to a value minus the sum of the chromaticity abscissa component and the chromaticity ordinate component after limiting protection in the corrected lighting state vector, divided by the chromaticity ordinate component after limiting protection, and then multiplied by the luminance component in the corrected lighting state vector. The luminance component in the corrected lighting state vector is directly used as the second stimulus value in the linear space.

[0104] S532, the signal processor combines the calculated first stimulus value, second stimulus value and third stimulus value into a tristimulus value vector, and then decouples it linearly into the original primary color intensity components of each physical light emission channel by left multiplying it by the set channel inverse mapping matrix, which is the inverse of the forward color mixing transformation matrix.

[0105] Since the underlying light-emitting hardware of the execution node relies on the physical superposition of independent primary colors, this embodiment uses the red, green, and blue LED channels as an example for linear decoupling. The channel inverse mapping matrix elements are obtained by extracting the basic CIE 1931 chromaticity coordinates and luminous flux ratios of the red, green, and blue LEDs within the execution node under rated luminous current, constructing a forward color mixing transformation matrix, and then performing matrix inversion on this matrix. Due to overadjustment in the dynamic adjustment of the control closed loop or the target chromaticity exceeding the boundary of the hardware light-emitting color gamut polygon of the execution node, the intensity components of each decoupled original primary color may exhibit negative values ​​or exceed their physical rated maximum values.

[0106] To prevent integer overflow in pulse width modulation calculations due to values ​​exceeding physical boundaries, the signal processor performs bidirectional truncation limits on the decoupled red, green, and blue original primary color intensity components. The truncation limit process is as follows: take the smaller value between the original primary color intensity component and the set normalized value of the channel's rated maximum luminous intensity, then compare this smaller value with zero and take the maximum value, outputting the truncated channel luminous intensity; the normalized value of the channel's rated maximum luminous intensity is configured to 1.0.

[0107] The S533 signal processor performs inverse gamma nonlinear compensation on the luminous intensity of each channel after truncation and limitation, converting it into a pulse width modulation duty cycle digital control quantity that matches the characteristics of the physical actuator. The principle of nonlinear compensation lies in the fact that there is a nonlinear electro-optic conversion response between the input drive duty cycle of the light-emitting diode and the output physical luminous intensity. Directly performing linear mapping causes a decrease in visual accommodation resolution and brightness gradient distortion in the low-brightness range.

[0108] The signal processor introduces an inverse gamma exponential function to perform inverse nonlinear correction on the luminous intensity, ensuring a linear correspondence between the perceived luminous intensity and the internal control vector. The calculation process for inverse gamma nonlinear compensation is as follows: the ratio of the truncated channel luminous intensity to the normalized value of the channel's rated maximum luminous intensity is calculated; this ratio is then raised to the power of a set electro-optic conversion nonlinear correction coefficient; the result of the power operation is multiplied by a set maximum pulse width modulation (PWM) count value to obtain the PWM duty cycle for each physical luminous channel. The set electro-optic conversion nonlinear correction coefficient is configured as 2.2 based on the electro-optic measurement response curves of each color light source within the execution node; the set maximum PWM count value is determined based on the counting clock resolution of the internal timer of the physical layer microcontroller and is configured as 65535.

[0109] In the S534, the signal processor within the main control module sends the calculated pulse width modulation duty cycles of the red, green, and blue channels to the target execution node via a hardware communication bus. The microcontroller within the target execution node receives the corresponding data frames through the bus interface and performs low-level decoding to extract the duty cycle values ​​for each channel.

[0110] The microcontroller extracts the duty cycle values ​​of each channel and writes them into the comparison register of its internal hardware timer. It then adjusts the square wave duty cycle of the corresponding hardware output pin, controlling the conduction time of the constant current driver chip to achieve pulse width modulation of the physical drive current of the LED, thus realizing smooth control of the chromaticity and brightness of the physical space lighting. Regarding the underlying waveform generation mechanism of the microcontroller outputting the pulse width modulation drive signal by configuring the underlying hardware timer register, those skilled in the art can use conventional embedded control programs; its register configuration is well-known technology in the field and will not be elaborated upon here.

[0111] Specific application examples: This embodiment provides an adaptive color-changing lighting control system and method based on sound-controlled sensing. In the offline calibration phase, an acoustic transfer function and an equivalent received signal strength indication matrix are constructed to complete the proxy binding of execution nodes and sensing nodes, as well as the allocation of node convergence weights. During the online operation phase, the system extracts the energy envelope and spectral centroid frequency values ​​of the sound field in real time, calculates the target lighting state vector, and uses a discrete wolf pack algorithm combined with a transient discrimination factor to perform physical weight convergence control. Simultaneously, the system introduces a fractional-order controller based on the adaptive adjustment of the calculus order according to the proportion of high-frequency components to perform closed-loop correction of lighting control deviations. Finally, the underlying pulse width modulation duty cycle is output after linear decoupling and inverse gamma compensation. Test data shows that the response time for tracking transient changes in the sound field is 95ms, the transient overshoot is controlled within 1.5%, and the visual flicker index is 0.012, improving the visual oscillation and overload problems existing in open-loop control and fixed-order PID control.

[0112] Taking a multi-functional conference room with dimensions of 8m×6m×3m and a gridded deployment of 12 execution nodes and 4 sensor nodes as an example. During the offline calibration phase, the main control module, based on the received signal strength indication matrix, proxy-binds execution node 5, located directly above the conference table, to sensor node 2, and assigns it a node convergence weight of 0.736 based on the 15dB direct sound energy ratio at that location. When the conference room is in a quiet steady state, the long-term moving average background energy remains stable at 0.05. When a speaker turns on their microphone, the sensor node extracts the backbone audio energy envelope value of the current time frame in real time, which is 0.25. The main control module then calls the formula accordingly: ; Substituting the numerical values, the transient discrimination factor is calculated to be approximately 3.99. This value is greater than the preset abrupt change threshold of 2.0, and the sound field state indicator switches to transient excitation state. Subsequently, the main control module maps the extracted 2500 Hz spectral centroid frequency value to the correlated color temperature, and combines it with the target brightness mapped by the energy envelope to generate a transient target illumination vector. The system substitutes the transient discrimination factor into a nonlinear saturation function to map it to a convergence control weight, and combines it with the node convergence weight of execution node 5 to calculate a comprehensive convergence coefficient of 0.706. Finally, the main control module uses the convergence update formula: ; Control execution node 5 transitions from a steady state of standard white light at 40% brightness to a transient target of warm white light at 80% brightness. The accompanying diagrams in the manual, combined with experimental verification data, illustrate the system's control state.

[0113] See attached document Figure 7 The horizontal axis represents the time frame, and the vertical axis represents the normalized acoustic and optical parameter amplitudes. In the legend, the solid line represents the change of the backbone audio energy envelope, the dashed line represents the resolved target photometric brightness, and the dotted line represents the physical output brightness of the execution node, showing the process of the physical output brightness conforming to the target trajectory according to the approach coefficient when facing the sudden change of acoustic energy.

[0114] See attached document Figure 8 The horizontal axis represents time (ms), and the vertical axis represents normalized illuminance. The legend contains three curves with different line types, which respectively represent the illuminance adjustment response trajectories of open-loop linear mapping control, traditional integer-order PID control, and the adaptive fractional-order control of this invention when facing the same sudden broadband noise excitation.

[0115] See attached document Figure 9 The horizontal axis lists the three different control algorithms mentioned above, and the vertical axis corresponds to the specific performance evaluation index values. In the legend, the dark gray bars represent the brightness adjustment time (ms), and the light gray bars represent the transient overshoot percentage. The comparison reflects the differences in adjustment time and overshoot suppression among the various schemes.

Claims

1. An adaptive color-changing lighting control system based on sound-controlled sensing, characterized in that, This includes the main control module, sensor network, and execution network; The main control module includes a signal processor; The sensor network and the execution network are respectively communicatively connected to the main control module; The sensor network includes multiple sensor nodes, each of which is equipped with a first communication module and a microphone array for acquiring multi-channel environmental acoustic signals. The execution network includes multiple execution nodes, each of which is equipped with a microcontroller, a driver module, a second communication module, and a photoelectric feedback module for acquiring actual lighting state vectors; The signal processor is used to extract acoustic features and the spatial angle of arrival of the sound source, generate a target lighting state vector based on the acoustic features, determine a target broadcast group based on the spatial angle of arrival of the sound source and a pre-stored spatial topology coordinate table, generate a corrected lighting state vector based on the control deviation vector between the target lighting state vector and the actual lighting state vector of the corresponding execution node, convert the corrected lighting state vector into a pulse width modulation duty cycle, encapsulate it into a target broadcast control data frame and send it to the target broadcast group to drive the underlying light-emitting device.

2. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 1, characterized in that, The actual lighting state vector includes the CIE 1931 chromaticity horizontal axis component, chromaticity vertical axis component, and luminance component; The photoelectric feedback module includes at least one of an illuminance sensor, a color sensor, or an integrated colorimetric sensor; The signal processor calculates the control deviation vector between the target lighting state vector and the actual lighting state vector of the corresponding execution node, generates a control correction vector using a fractional proportional-integral-derivative controller, superimposes the control correction vector onto the actual lighting state vector to obtain the corrected lighting state vector, and encapsulates the pulse width modulation duty cycle corresponding to the corrected lighting state vector into the target broadcast control data frame.

3. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 2, characterized in that, During the initialization phase, the main control module enters an offline calibration state and sends a preset broadband test signal to the physical space. Each sensor node collects the acoustic response signal and calculates the acoustic transfer function and direct sound energy ratio at the location of each sensor node. A wireless ranging command is sent to the execution network, and a received signal strength indication matrix is ​​constructed based on the received signal strength indication value of the radio frequency test frame. An equivalent received signal strength indication matrix is ​​generated by moving average filtering. Each execution node is bound to the sensor node with the highest equivalent received signal strength, and the normalized acoustic feature value obtained by normalizing the direct sound energy ratio of the bound sensor node is used to assign node convergence weights to each execution node.

4. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 3, characterized in that, The signal processor converts the multi-channel ambient acoustic signal into a frequency domain signal vector, constructs a spatial cross-correlation matrix, and performs eigenvalue decomposition. Based on the established information theory criteria, the signal subspace and noise subspace are separated, and a spatial pseudospectral function is constructed using the noise subspace to extract the spatial angle of arrival of the sound source. The signal feature vector corresponding to the largest eigenvalue of the spatial cross-correlation matrix is ​​extracted. The multi-channel environmental acoustic signal is weighted and synthesized to obtain the backbone audio sequence. The backbone audio energy envelope is reconstructed and the spectral centroid frequency value is extracted. The backbone audio energy envelope and the spectral centroid frequency value are concatenated to form an acoustic time-frequency feature vector.

5. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 4, characterized in that, The signal processor maps the backbone audio energy envelope to a target luminance component, maps the spectral centroid frequency value to a correlated color temperature value and converts it into a target chromaticity coordinate component in the CIE 1931 chromaticity space, and combines them to generate the target illumination state vector. The targeted broadcast group is determined based on the spatial vector pointed to by the spatial arrival angle of the sound source and the spatial topology coordinate table, and closed-loop correction control of the lighting output state is performed on the execution nodes within the targeted broadcast group.

6. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 5, characterized in that, The signal processor calculates the absolute value of the difference between the backbone audio energy envelope of the current time frame and the long-term moving average background energy of the previous time frame, and uses the ratio of the absolute value of the difference to the long-term moving average background energy as a transient discrimination factor. When the transient discrimination factor is greater than the preset transient change threshold, the sound field state indicator bit is configured to transient excitation state; When the transient discrimination factor is less than or equal to the preset transient change threshold, the sound field state indicator bit is configured as a steady-state background state.

7. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 6, characterized in that, When the sound field state indicator is in a steady-state background state, the signal processor performs step-limited walk control on execution nodes not included in the targeted broadcast group: A pseudo-random sequence generator is invoked to generate a three-dimensional random perturbation vector; The intermediate lighting state vector is generated based on the difference between the base anchor point vector and the target lighting state vector of the previous frame, the center recovery coefficient, the three-dimensional random perturbation vector, and the walk step gain. Boundary truncation is performed on the intermediate lighting state vector to obtain the target lighting state vector of the current time frame.

8. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 6, characterized in that, When the sound field state indicator is in transient excitation state, the signal processor generates a transient target illumination vector based on the backbone audio energy envelope and the spectral centroid frequency value of the current time frame, and performs physical weight convergence control on the execution nodes within the targeted broadcast group: The transient discriminant factor is mapped to a convergent control weight using a nonlinear saturation function; Based on the node convergence weight of the corresponding execution node, the approach control weight, and the Euclidean distance between the target illumination state vector of the previous frame and the transient target illumination vector, a comprehensive approach coefficient with a value limited to 0 to 1 is calculated. The target illumination state vector of the previous frame and the transient target illumination vector are interpolated and smoothed according to the comprehensive convergence coefficient to obtain the target illumination state vector of the current time frame.

9. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 2, characterized in that, The signal processor uses a truncated Grünwald-Letnikov time-domain discrete approximation algorithm with finite memory length to construct a fractional calculus operator, calculates fractional integral terms and fractional derivative terms, and linearly combines them with proportional terms to generate a control correction vector, which is then superimposed on the actual lighting state vector to obtain the corrected lighting state vector. Acquire the acoustic power spectral density data of the current time frame, and use the ratio of the total energy in the high-frequency band to the total energy in the entire frequency band as the proportion of the high-frequency component. Solve the target order of the integral element and the target order of the differential element, and overwrite the fractional order of the integral element and the fractional order of the differential element of the fractional calculus operator after time-domain smoothing filtering.

10. The adaptive color-changing lighting control system based on sound-controlled sensing according to claim 2, characterized in that, The signal processor performs lower limit amplitude protection on the chromaticity ordinate component in the corrected lighting state vector and calculates the first stimulus value, the second stimulus value and the third stimulus value in the linear mapping space. The first stimulus value, the second stimulus value, and the third stimulus value are combined into a tristimulus value vector, and the tristimulus value vector is left-multiplied by the set channel inverse mapping matrix to linearly decouple it into the original primary color intensity components of the three physical light emission channels: red, green, and blue. By performing bidirectional truncation constraints and inverse gamma nonlinear compensation on each original primary color intensity component, the pulse width modulation duty cycle of the corresponding physical light emission channel is obtained.

Images