Intelligent light control method and system based on sound change

Abstract

The invention discloses a light intelligent control method and system based on sound change, and relates to the technical field of light intelligent control, and the method comprises the steps: obtaining an environment sound signal containing a sound pressure level time sequence in a target sound field region; importing the voiceprint data into a voiceprint feature analysis engine, constructing a time pane matrix which stretches and retracts along with the sound intensity, and generating a voiceprint energy streamline; capturing a slope abrupt change event of the voiceprint energy streamline, and extracting an acoustic disturbance feature packet; the feature packet drives a brightness controller to output a light intensity compensation deformation instruction, compensation deformation is applied to a time pane matrix, and new voiceprint energy streamline distribution is generated; and identifying the steady-state voiceprint channel and mapping the steady-state voiceprint channel into a target illumination value instruction and a color temperature offset According to the method, accurate adaptation of lamplight and sound changes can be realized, and the continuity and stability of lamplight control are improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent lighting control technology, specifically a method and system for intelligent lighting control based on sound changes. Background Technology

[0002] Currently, sound-controlled lighting technology is widely used in various scenarios. Existing technologies mostly collect sound signals from the environment, extract fixed-dimensional features such as sound pressure level and spectrum, slice the sound signal using a fixed-duration time window, and then directly output lighting control commands based on the extracted features to achieve functions such as turning lights on and off and adjusting brightness. This type of technology relies on a fixed time window setting, dividing the sound signal into uniform time intervals, ignoring the dynamic changes in sound intensity itself.

[0003] Existing technologies with fixed time windows cannot adapt to the dynamic changes in sound intensity. When sound intensity changes abruptly, fixed time windows struggle to accurately capture the instantaneous changes in sound characteristics, leading to deviations in voiceprint feature analysis and consequently affecting the accuracy of lighting control commands. Furthermore, existing sound-based lighting technologies often employ unidirectional control logic, generating lighting control commands directly from sound signals without a feedback adjustment mechanism. This prevents the correction of voiceprint analysis deviations based on the actual effect of lighting control, resulting in poor consistency and stability in lighting adjustments. Consequently, these technologies struggle to match the dynamic changes in sound and fail to achieve precise, adaptive intelligent lighting control. Summary of the Invention

[0004] This invention aims to solve at least one of the technical problems existing in the prior art; Therefore, this invention proposes a smart lighting control method based on sound changes, comprising: Acquire ambient sound signals within the target sound field area, wherein the ambient sound signals include a time series of sound pressure levels generated by air vibrations; The ambient sound signal is imported into the voiceprint feature analysis engine for dynamic slicing. A time window matrix that expands and contracts with the sound intensity is constructed in the voiceprint feature analysis engine. Voiceprint energy streamlines are generated between the grids of the time window matrix according to the sound pressure level time series. Track and capture slope abrupt events of the acoustic energy streamline. When the local slope of the acoustic energy streamline exceeds a preset slope threshold, extract the waveform segment of the acoustic energy streamline and its corresponding time pane matrix region to form an acoustic disturbance feature package. Using the acoustic disturbance feature packet as input, a brightness controller based on acoustic deformation feedback is driven. The brightness controller based on acoustic deformation feedback outputs a light intensity compensation deformation command according to the spectral migration mode of the waveform segment in the acoustic disturbance feature packet and the energy density distribution of the time pane matrix region. According to the light intensity compensation deformation command, the corresponding compensation deformation is applied to the time window matrix in the voiceprint feature analysis engine, and the time window matrix after compensation deformation generates a new voiceprint energy streamline distribution. The steady-state audio channel is identified in the new audio energy streamline distribution, and the steady-state audio channel is mapped back to the physical light control space and converted into target illuminance value command and color temperature offset command.

[0005] Furthermore, the ambient sound signal is imported into a voiceprint feature analysis engine for dynamic slicing. A time-pane matrix that scales and adjusts with sound intensity is constructed within the voiceprint feature analysis engine, including: The sound intensity level and the centroid frequency of the spectrum are calculated at each sampling time based on the sound pressure level time series. A two-dimensional acoustic field is constructed with the sound intensity level value as the horizontal axis coordinate and the frequency of the spectral centroid as the vertical axis coordinate, and the acoustic state point at each sampling time is plotted in the two-dimensional acoustic field. Connecting the voiceprint state points at adjacent sampling times in the two-dimensional voiceprint field forms a dynamic voiceprint trajectory line; Based on the spatial density of the dynamic voiceprint trajectory line, a non-uniform rectangular grid is automatically generated in the two-dimensional voiceprint field to cover the dynamic voiceprint trajectory line. The time width of the rectangular grid is positively correlated with the sound intensity level value. The rectangular grid is the time pane matrix in the voiceprint feature analysis engine, which adjusts its grid size and arrangement relationship as the voiceprint state point changes.

[0006] Further, the step of generating acoustic energy streamlines between the grids of the time pane matrix based on the sound pressure level time series includes: For each rectangular grid cell in the time pane matrix, calculate the gradient vector between the sound intensity level value at its four vertices and the frequency value at the centroid of the spectrum; The gradient vectors of the four vertices within each rectangular grid cell are vector synthesized to obtain the acoustic flow vector of the rectangular grid cell; Along the sound signature flow vector field, streamlines from high sound signature density regions to low sound signature density regions are drawn on the time window matrix. These streamlines are the sound signature energy streamlines that characterize the real-time convergence and diffusion path of sound energy.

[0007] Furthermore, the tracking and capturing of slope abrupt events in the acoustic signature energy streamline includes: The local slope of each acoustic energy streamline is calculated with a fixed step size to form a slope sequence of acoustic energy streamlines; Apply sliding window mean calculation to the slope sequence to detect the time points in the slope sequence that deviate from the mean by more than the mutation detection threshold, and mark them as the time of occurrence of slope mutation events; Back to the moment of occurrence, extract the acoustic energy streamline segments within a preset time window before and after the moment of occurrence, and simultaneously record the set of rectangular grid cells in the time window matrix through which the acoustic energy streamline segments pass. The acoustic energy streamline segments and the set of rectangular grid cells together constitute the acoustic disturbance feature package.

[0008] Furthermore, using the acoustic disturbance feature packet as input, a brightness controller based on acoustic pattern deformation feedback is driven, comprising: The waveform segments within the acoustic disturbance feature package are subjected to joint time-frequency analysis to extract the dominant peak frequency and harmonic attenuation coefficient in the time spectrum, which are used as spectral migration mode features. Analyze the time pane matrix region within the acoustic disturbance feature package, and calculate the variance of the sound intensity level values ​​of all rectangular grid cells within the time pane matrix region and the dispersion of the spectral centroid frequency as energy density distribution characteristics. The spectral migration mode features and energy density distribution features are input into a pre-trained deformation decision network. The deformation decision network outputs instructions to apply expansion, contraction or translation deformation to the target region of the time pane matrix, as well as the deformation amplitude and direction corresponding to each deformation instruction, which together constitute the light intensity compensation deformation instructions.

[0009] Furthermore, the step of applying corresponding compensation deformation to the time pane matrix in the voiceprint feature parsing engine according to the light intensity compensation deformation instruction includes: Locate the target area specified by the light intensity compensation deformation command in the two-dimensional acoustic field, wherein the target area is identified by a set of rectangular grid cells; According to the deformation type and deformation amplitude in the light intensity compensation deformation command, the center coordinates of all rectangular grid cells in the target area are synchronously transformed. After the center coordinate transformation, the physical boundaries and adjacency relationships of all affected rectangular grid cells are recalculated and updated to complete the construction of the time pane matrix after deformation compensation.

[0010] Furthermore, the identification of steady-state voiceprint channels in the new voiceprint energy streamline distribution includes: the steady-state voiceprint channel is defined by the slope of the voiceprint energy streamline being continuously lower than a steady-state threshold; On the time pane matrix after deformation compensation, the gradient vector of each grid cell is calculated and the acoustic signature flow vector field is synthesized. The acoustic signature energy streamline from the high acoustic signature density region to the low acoustic signature density region is regenerated along the flow field. Monitor the slope change of newly generated voiceprint energy streamlines within a preset observation period. If the maximum slope of a certain voiceprint energy streamline is consistently lower than the steady-state threshold throughout the entire observation period, then the voiceprint energy streamline segment is determined to be a steady-state voiceprint channel. Record the starting point coordinates, ending point coordinates, and the sequence of raster cells traversed by all steady-state acoustic channels in the two-dimensional acoustic field.

[0011] Furthermore, the process of mapping the steady-state acoustic channel back to the physical light control space and converting it into target illuminance value instructions and color temperature offset instructions includes: For each steady-state acoustic text channel, the average sound intensity level and average spectral centroid frequency of the steady-state acoustic text channel in the two-dimensional acoustic text field are calculated based on its starting point coordinates and ending point coordinates. By using a preset sound-light mapping relationship, the average sound intensity level value is converted into an illuminance value command for the target area, and the average spectral centroid frequency value is converted into a color temperature offset command relative to the reference white light.

[0012] Furthermore, it also includes: The steady-state light signal defined by the target illuminance value command and the color temperature offset command is fused with the residual noise oscillation mode to synthesize the light driving waveform signal; Based on the light driving waveform signal, a pulse width modulation signal for the lighting device is generated to realize intelligent control of the light based on sound changes; The environmental sound signal is subjected to overtone mode stripping to separate the fundamental frequency energy mode and the residual noise oscillation mode; The process of performing overtone mode stripping on the environmental sound signal to separate the fundamental frequency energy mode and the residual noise oscillation mode includes: Hilbert-Huang transform is applied to the sound pressure level time series of the ambient sound signal to obtain a series of intrinsic mode function components. Based on the instantaneous frequency of the intrinsic mode function components, intrinsic mode function components whose instantaneous frequency is within the audio fundamental frequency range are selected. The intrinsic mode function components are reconstructed into fundamental frequency energy modes. The original ambient sound signal is subtracted from the fundamental frequency energy modes, and the remaining waveform components are the residual noise oscillation modes. The step of fusing the steady-state light signal defined by the target illuminance value command and the color temperature offset command with the residual noise oscillation mode to synthesize the light driving waveform signal includes: Based on the target illuminance value command and the color temperature offset command, a composite reference waveform with DC bias and AC modulation is generated as a steady-state light signal. The initial light waveform is obtained by convolving and superimposing the steady-state light signal with the residual noise oscillation mode in the time domain. The initial light waveform is subjected to nonlinear normalization and high-frequency jitter suppression to ensure that its value range meets the driving specifications of the lighting equipment and that the waveform is smooth and continuous. The processed waveform is the final light driving waveform signal used to generate the pulse width modulation signal.

[0013] Furthermore, the present invention also includes a sound-change-based intelligent lighting control system, the system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the steps of the sound-change-based intelligent lighting control method described above.

[0014] Compared with the prior art, the beneficial effects of the present invention are: A time-pane matrix that scales and expands with sound intensity is constructed. Based on the sound pressure level time series, acoustic energy streamlines are generated between the grids of this time-pane matrix. These acoustic energy streamlines serve as the carriers of sound features, replacing conventional methods of extracting acoustic features using fixed-duration time windows and fixed-dimensional dimensions. The dynamically scaling time-pane matrix adjusts the size and distribution of the time windows in real time according to changes in sound intensity, accurately capturing the instantaneous features of sounds at different intensities. This avoids the loss or redundancy of acoustic features caused by fixed time windows, making acoustic feature analysis more closely aligned with the actual dynamic changes in sound. The accuracy and adaptability of acoustic feature analysis are optimized, and the generation of light control commands remains consistent with the patterns of sound changes.

[0015] An acoustic disturbance feature package drives a brightness controller based on acoustic pattern deformation feedback to output a light intensity compensation deformation command. This command is then applied inversely to the acoustic pattern feature analysis engine, performing corresponding compensation deformation on the time pane matrix. The deformed time pane matrix generates a new acoustic pattern energy streamline distribution. Steady-state acoustic pattern channels are then identified from this new distribution and mapped to physical lighting control commands, forming a closed-loop control mechanism based on acoustic pattern deformation feedback. This replaces the conventional unidirectional control logic of "sound signal → direct output lighting command". The closed-loop feedback mechanism can correct deviations in the acoustic pattern analysis process in real time, ensuring that the distribution of acoustic pattern energy streamlines always adapts to the dynamic changes in sound. The illuminance and color temperature of the lights can be adjusted in real time to follow sound changes, avoiding control lag or deviation. The consistency and stability of lighting control are optimized, resulting in a precise match between light and sound changes. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the steps of the intelligent lighting control method based on sound changes described in this invention. Figure 2 A flowchart for constructing a time pane matrix that scales as sound intensity changes; Figure 3 A flowchart for generating acoustic signature energy streamlines based on sound pressure level time series; Figure 4 A summary diagram of the acousto-optic feature mapping of multiple stable acoustic channels; Figure 5 This is a monitoring diagram of abrupt changes in the energy flow lines and slope of the voiceprint. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0018] See Figure 1 The system acquires ambient sound signals within the target sound field area, including a time series of sound pressure levels generated by air vibrations. These signals are then imported into a voiceprint feature analysis engine for dynamic slicing, where a time-pane matrix that automatically scales with changes in sound intensity is constructed. Based on the sound pressure level time series, voiceprint energy streamlines characterizing energy dynamics are generated between the grids of the time-pane matrix. The system continuously tracks and captures slope abrupt events of the voiceprint energy streamlines. When a local slope of the voiceprint energy streamline exceeds a preset slope threshold, the waveform segment corresponding to the abrupt change point and its corresponding time-pane matrix region are extracted and encapsulated into an acoustic disturbance feature package. Using this acoustic disturbance feature package as input, a brightness controller based on voiceprint deformation feedback is driven. This controller calculates and outputs corresponding light intensity compensation deformation commands by analyzing the spectral migration patterns of the waveform segments within the feature package and the energy density distribution of the time-pane matrix region. Based on the received light intensity compensation deformation command, the specified compensation deformation operation is applied to the time pane matrix in the voiceprint feature parsing engine. The deformed time pane matrix will generate a new voiceprint energy streamline distribution. In the new voiceprint energy streamline distribution, the system identifies steady-state voiceprint channels with gentle slope changes and maps the geometric and energy characteristics of these steady-state voiceprint channels back to the physical light control space, ultimately converting them into target illuminance value commands and color temperature offset commands for adjusting lighting equipment.

[0019] In one embodiment of the present invention, see [reference] Figure 2 The sound intensity level and spectral centroid frequency for each sampling moment are calculated based on the sound pressure level time series. A two-dimensional acoustic signature field is constructed using the calculated sound intensity level as the horizontal axis and the spectral centroid frequency as the vertical axis, and the acoustic signature state point for each sampling moment is plotted within this field. Within the two-dimensional acoustic signature field, acoustic signature state points from adjacent sampling moments are connected in chronological order to form a dynamic acoustic signature trajectory line. Based on the spatial distribution density of this dynamic acoustic signature trajectory line within the two-dimensional acoustic signature field, the system automatically generates a set of non-uniform rectangular grids to cover the entire trajectory line. The width of each rectangular grid in the time axis direction is positively correlated with the average sound intensity level of the trajectory line segment it covers. These adaptively generated rectangular grids together constitute the temporal pane matrix in the acoustic signature feature parsing engine.

[0020] In practical implementation, after acquiring the environmental sound signal within the target sound field area, the sound pressure level time series of the signal is processed. Based on the sound pressure level time series, the sound intensity level value at each sampling moment is calculated. The sound intensity level value is expressed in decibels (dB) and represents the intensity of the sound at that moment. Simultaneously, the centroid frequency value of the sound signal's spectrum at each sampling moment is calculated. The centroid frequency value reflects the center of sound energy distribution in the frequency domain, and its calculation method can be expressed as: in: Represents the centroid frequency value of the spectrum. This represents the frequency value at the k-th frequency point. Representing frequency point The corresponding power spectral density, This represents the total number of frequency points.

[0021] In some embodiments, a two-dimensional acoustic signature field is constructed using the calculated sound intensity level as the horizontal axis coordinate and the spectral centroid frequency value as the vertical axis coordinate. The acoustic signature state point at each sampling moment is plotted in the two-dimensional acoustic signature field based on its corresponding sound intensity level and spectral centroid frequency value. Connecting all temporally adjacent acoustic signature state points in the two-dimensional acoustic signature field forms a dynamic, continuous acoustic signature trajectory line, which visually demonstrates the evolution path of sound characteristics in terms of intensity and spectral centroid. Optionally, the sound intensity level is calculated based on a preset reference sound pressure level, obtained by multiplying the ratio of the effective sound pressure value at the sampling moment to the reference sound pressure level by 20 after taking the logarithm to base 10. It can be understood that the spectral centroid frequency value, as a global statistical feature of the sound spectrum, is related to the perceived pitch of the sound; sounds with a higher proportion of high-frequency components have a higher spectral centroid frequency value.

[0022] In practical implementation, based on the spatial distribution density of the dynamic voiceprint trajectory lines in the two-dimensional voiceprint field, the system automatically performs rasterization coverage. A set of non-uniform rectangular grids covering the dynamic voiceprint trajectory lines is generated. The time width of each rectangular grid in the time axis direction is not fixed, but rather positively correlated with the average sound intensity level of the trajectory line segment covered by the grid. That is, when the average sound intensity level of a certain segment of the dynamic voiceprint trajectory line is large, the time width of the rectangular grid covering that segment is also large; conversely, when the average sound intensity level is small, the time width of the corresponding rectangular grid is also small. It can be understood that these adaptively generated, variable-sized rectangular grids based on the dynamic voiceprint trajectory lines constitute the time pane matrix used for subsequent analysis in the voiceprint feature parsing engine.

[0023] In one embodiment of the present invention, see [reference] Figure 3For each rectangular grid cell in the time pane matrix, the gradient vectors of the sound intensity level and the centroid frequency of the spectrum at its four vertices in the two-dimensional acoustic field are calculated. The gradient vectors of the four vertices within each rectangular grid cell are vector-synthesized to obtain the acoustic flow vector representing the overall trend of acoustic energy within that grid cell. Based on the vector field formed by the acoustic flow vectors of all rectangular grid cells in the entire time pane matrix, continuous streamlines are drawn on the matrix from high acoustic density regions to low acoustic density regions. These streamlines represent the acoustic energy streamlines characterizing the real-time convergence and diffusion paths of sound energy. Along each generated acoustic energy streamline, its local slope is calculated with a fixed step size, forming a slope sequence. A sliding window mean operation is applied to the slope sequence, detecting time points in the sequence that deviate from the current window mean by more than a preset abrupt change detection threshold, and these time points are marked as the occurrence times of slope abrupt change events. The system traces back to the moment of occurrence, extracts a segment of acoustic energy flow line within a preset time window before and after the moment of occurrence, and records all rectangular grid cells in the time window matrix through which the segment passes. The extracted acoustic energy flow line segment and the set of recorded rectangular grid cells together constitute an acoustic disturbance feature package.

[0024] In practical implementation, for each rectangular grid cell in the time pane matrix, it is necessary to calculate the gradient vectors corresponding to its four vertices in the two-dimensional acoustic signature field. Each vertex of the rectangular grid cell corresponds to a sound intensity level value and a spectral centroid frequency value. By calculating the rate of change of the vertex in the sound intensity level dimension and the rate of change in the spectral centroid frequency value dimension, the gradient vector characterizing the direction and rate of change of acoustic signature features at that vertex can be obtained. Specifically, for each vertex... Its gradient vector It can be obtained by the numerical difference between it and its adjacent vertices in the two-dimensional field, where and Used to identify the position index of a vertex in the matrix.

[0025] In some embodiments, the gradient vectors of the four vertices of the rectangular grid cell are obtained. , , , Then, these vectors are combined. Vector combination is achieved through vector addition, and the resulting vector is the voiceprint flow vector of the rectangular grid cell. , The sound signature flow vector represents the overall movement trend of sound energy within the spatiotemporal range covered by the rectangular grid cell. It can be understood that, based on the sound signature flow vectors of all rectangular grid cells on the entire time pane matrix, a complete vector field can be constructed, which describes the flow pattern of sound energy at different locations in the two-dimensional sound signature field.

[0026] In specific implementation, soundprint energy streamlines are drawn along the soundprint flow vector field. The drawing process begins with a preset or automatically identified high soundprint density region. A high soundprint density region refers to an area in the two-dimensional soundprint field where the dynamic soundprint trajectory line traverses a large number of times per unit area or where the sound intensity level remains consistently high. The drawing algorithm determines the tangent direction of the streamline based on the soundprint flow vector direction at each location, and iteratively integrates with a fixed step size to form a continuous, smooth curve pointing from the high soundprint density region to the low soundprint density region; these curves are the soundprint energy streamlines. In some embodiments, a low soundprint density region refers to an area where the dynamic soundprint trajectory line is sparsely distributed or where the sound intensity level is low. The soundprint energy streamlines visually represent the real-time transmission and diffusion path of sound energy from concentrated areas to dispersed areas. Optionally, after the soundprint energy streamlines are generated, the system calculates the local slope of each soundprint energy streamline with a fixed step size, forming a slope sequence corresponding to the spatial coordinate point sequence of that streamline. ,in To calculate the number of points. Slope value. This reflects the rate of change of streamline direction at that point and demonstrates the calculation method based on the central difference method: in: Indicates the first on the streamline Local slope at each point, and They represent the first Point and the The centroid frequency value of the spectrum corresponding to the point in the two-dimensional acoustic field. and They represent the first Point and the The sound intensity level value corresponding to the point.

[0027] In practice, a sliding window mean operation is applied to the slope sequence to detect abrupt slope changes. A width of [value missing] is set. Using a sliding window, calculate the arithmetic mean of all slope values ​​within the window. When a certain slope value is within the window With window mean The absolute value of the deviation exceeds the preset mutation detection threshold. When, that is, satisfied Then The corresponding time point is marked as the time when the slope abruptly changes. This can be understood as the mutation detection threshold. It is a pre-defined positive constant whose magnitude determines the system's sensitivity to abrupt changes in the direction of the sound signature energy flow.

[0028] In practice, after marking the occurrence time, the system backtracks to that point in time. A segment of length is extracted both forward and backward from the occurrence time. Within a given time window, the system extracts a segment of acoustic energy streamlines. Simultaneously, it records all rectangular grid cells within the time window matrix that this acoustic energy streamline segment traverses in the two-dimensional acoustic field; these rectangular grid cells form a set. The extracted acoustic energy streamline segment and the recorded set of rectangular grid cells are collectively encapsulated into a complete data structure. This data structure constitutes the acoustic perturbation feature package used to drive the subsequent brightness controller.

[0029] In one embodiment of the invention, waveform segments within the acoustic disturbance feature package undergo time-frequency joint analysis to extract the dominant peak frequency and harmonic attenuation coefficient in the time spectrum. These parameters collectively constitute the spectral migration mode features. The time pane matrix region recorded within the acoustic disturbance feature package is analyzed, and the variance of the sound intensity level values ​​of all rectangular grid cells within this region and the dispersion of the spectral centroid frequency are calculated. These statistics collectively constitute the energy density distribution features. The extracted spectral migration mode features and energy density distribution features are input into a pre-trained deformation decision network. This network outputs instructions to apply specific deformation operations to the target region in the time pane matrix. The deformation operation types include expansion, contraction, or translation. The output instructions also include the deformation amplitude and direction corresponding to each deformation operation. This information collectively constitutes a complete light intensity compensation deformation instruction. In the two-dimensional acoustic field, the system first locates the target region specified by the light intensity compensation deformation instruction. This target region is determined by the identifiers of a set of rectangular grid cells. Subsequently, according to the deformation type and deformation amplitude specified in the instruction, a synchronous geometric transformation is performed on the center coordinates of all rectangular grid cells within the target region. After the center coordinate transformation of all affected raster cells is completed, the system recalculates and updates the physical boundaries of these raster cells and their spatial relationships with adjacent raster cells, thereby completing the construction of the time pane matrix after deformation compensation.

[0030] In practical implementation, the acoustic disturbance feature package is input to a brightness controller based on acoustic pattern deformation feedback. The controller first performs time-frequency joint analysis on the waveform segments within the feature package to extract spectral migration mode features characterizing the transient properties of the sound. Spectral migration mode features include the dominant peak frequency and harmonic attenuation coefficient in the time spectrum of the waveform segment. The dominant peak frequency refers to the frequency component with the highest power spectral density within a preset frequency band, and the harmonic attenuation coefficient describes the rate attenuation of the amplitude of each harmonic at the dominant peak frequency as the frequency increases. In practical implementation, the controller simultaneously analyzes the time pane matrix region recorded within the acoustic disturbance feature package to calculate the energy density distribution features. The energy density distribution features include the variance of the sound intensity level values ​​of all rectangular grid cells within the time pane matrix region, and the dispersion of the spectral centroid frequency of all rectangular grid cells within the time pane matrix region. Variance and dispersion quantify the degree of non-uniformity in the spatial distribution of sound intensity and spectral centroid, respectively.

[0031] In some embodiments, the extracted spectral migration pattern features and energy density distribution features are combined into a feature vector, which is then input into a pre-trained deformation decision network. The deformation decision network is an artificial neural network whose output layer corresponds to different combinations of deformation operation types and parameters. Based on the input feature vector, the deformation decision network calculates and outputs specific instructions for applying deformation to the target region of the time pane matrix. Deformation operation types include grid region expansion, contraction, or translation. The deformation decision network also outputs the deformation amplitude and deformation direction corresponding to each deformation operation type. The deformation amplitude is a scalar value, and the deformation direction is defined by a two-dimensional vector in the two-dimensional acoustic field. The deformation operation type, deformation amplitude, and deformation direction together constitute a complete light intensity compensation deformation instruction. Optionally, a forward calculation process of the deformation decision network can be described as follows: in: This represents the input feature vector composed of spectral migration mode features and energy density distribution features. The weight matrix represents the deformation decision network. This represents the bias vector of the deformation decision network. The activation function representing the deformation decision network. The original output vector represents the deformation decision network. After subsequent decoding and mapping, it is converted into specific deformation type, deformation amplitude, and deformation direction parameters.

[0032] In practical implementation, after receiving the light intensity compensation deformation command, the system locates the target area specified by the command in the two-dimensional acoustic field. The target area is determined by a unique identifier of a set of rectangular grid cells. According to the deformation type contained in the light intensity compensation deformation command, a synchronous geometric transformation is performed on the center coordinates of all rectangular grid cells within the target area. It can be understood that if the deformation type is expansion, the center coordinates of the rectangular grid cells move along the deformation direction, and the moving distance is proportional to the deformation amplitude and the distance from the center to the region's reference point; if the deformation type is contraction, the moving direction is opposite to expansion; if the deformation type is translation, the center coordinates of all rectangular grid cells are added with the same displacement vector defined by the deformation amplitude and deformation direction.

[0033] In some embodiments, after transforming the center coordinates of all rectangular grid cells within the target area, the system recalculates the physical boundary of each affected rectangular grid cell. The physical boundary of a rectangular grid cell is determined by its updated center coordinates, adjacent grid relationships, and preset grid generation rules. After recalculating the physical boundaries, the system updates the adjacency relationships between all affected rectangular grid cells and between affected and unaffected rectangular grid cells, thereby completing the construction of the time pane matrix after deformation compensation. It can be understood that the grid layout of the time pane matrix after deformation compensation reflects the proactive structural adjustment of the original sound feature space in response to acoustic disturbances.

[0034] In one embodiment of the present invention, after the deformation compensation operation is completed, the acoustic energy streamlines are recalculated and drawn according to the acoustic energy streamline generation method based on the new time pane matrix. The system monitors the slope change of the newly generated acoustic energy streamlines within a preset observation period. If the maximum slope value of a certain acoustic energy streamline is continuously lower than a preset steady-state threshold throughout the entire observation period, the acoustic energy streamline is determined to be a steady-state acoustic channel. The system records the starting point coordinates, ending point coordinates, and grid cell sequence traversed by the streamlines of all identified steady-state acoustic channels in the two-dimensional acoustic field. For each identified steady-state acoustic channel, the average sound intensity level and average spectral centroid frequency value of the channel in the two-dimensional acoustic field are calculated based on the recorded starting and ending point coordinates. Through a preset acousto-optic mapping function, the calculated average sound intensity level value is converted into the target illuminance value command required for the target illumination area, and the average spectral centroid frequency value is converted into a color temperature offset command relative to the reference white light source.

[0035] In practical implementation, based on the new time pane matrix constructed after the deformation compensation operation, the system recalculates and draws the soundprint energy streamlines according to the generation method of the soundprint energy streamlines. The grid distribution of the time pane matrix after deformation compensation is different from that before deformation, resulting in changes in the path and distribution of the newly generated soundprint energy streamlines. The system monitors the slope change of each newly generated soundprint energy streamline within a preset observation period. The observation period is a fixed time length calculated from the streamline's starting point, and the slope value is calculated in real time based on the coordinates of the points on the streamline. In practical implementation, a steady-state threshold is set. The steady-state threshold is a preset positive constant used to determine the stability of the streamline direction. If the maximum slope value of all calculated points of a certain soundprint energy streamline is continuously less than the steady-state threshold throughout the preset observation period, then this soundprint energy streamline is determined to be a steady-state soundprint channel. A steady-state soundprint channel represents a stable and gently changing transmission path of sound energy in a two-dimensional soundprint field. The system records the geometric and path information of all identified steady-state acoustic text channels, including the starting point coordinates, ending point coordinates, and the sequence of rectangular grid cells that the channel passes through in the two-dimensional acoustic text field for each steady-state acoustic text channel.

[0036] In some embodiments, for each recorded steady-state acoustic channel, it is necessary to calculate the acoustic feature values ​​required for its mapping. Based on the starting and ending coordinates of the steady-state acoustic channel, the average sound intensity level and average spectral centroid frequency of the channel in the two-dimensional acoustic field are calculated. The average sound intensity level is the arithmetic mean of the sound intensity levels corresponding to all points along the steady-state acoustic channel path, and the average spectral centroid frequency is the arithmetic mean of the spectral centroid frequencies corresponding to all points along the steady-state acoustic channel path. The calculation is expressed as follows: in: This represents the average centroid frequency value of the spectrum. This represents the average sound intensity level. Represents the first steady-state voiceprint channel The frequency values ​​of the spectral centroid corresponding to each point. Represents the first steady-state voiceprint channel The sound intensity level values ​​corresponding to each point. This represents the total number of feature points used for calculation on this steady-state acoustic channel.

[0037] In practical implementation, the calculated acoustic feature values ​​are converted into lighting control commands through a preset acoustic-optical mapping relationship. This preset mapping relationship defines the conversion rules from average sound intensity level values ​​to target illuminance values ​​and from average spectral centroid frequency values ​​to color temperature offset commands. The target illuminance value command is a value in lux (lx) used to specify the required illuminance of the lighting equipment. The color temperature offset command is a value in Kelvin (K) representing the adjustment amount required relative to the system-set reference white light color temperature; a positive value increases the color temperature, and a negative value decreases it. Optionally, a linear acoustic-optical mapping relationship can be implemented through a lookup table (see Table 1), which establishes an example of the correspondence between steady-state acoustic channel features and lighting control commands. It can be understood that the preset acoustic-optical mapping relationship can be configured and adjusted according to the lighting requirements of different application scenarios, and the mapping function can be linear, piecewise, or nonlinear.

[0038] Table 1: Mapping Table of Steady-State Acoustic Channel Features to Optical Control Commands In some embodiments, the system independently calculates and outputs a set of target illuminance value commands and color temperature offset commands for each identified steady-state acoustic fingerprint channel. When multiple steady-state acoustic fingerprint channels exist simultaneously, the system may output multiple sets of light control commands, which can be applied to the lighting system through weighted fusion or zone control strategies. It can be understood that the target illuminance value command and color temperature offset command together constitute a digital description of the lighting state of the target area, used to directly drive or adjust the operating parameters of the lighting equipment.

[0039] See Figure 4 This is a summary diagram of the acoustic-optical features of five steady-state acoustic channels, used to visually demonstrate the mapping relationship between the acoustic characteristics of the five steady-state acoustic channels and their corresponding lighting control commands. For illuminance control, the higher the average sound intensity level, the greater the target illuminance value. For example, channel 5 has the highest average sound intensity level, and its corresponding illuminance value is also the highest. For color temperature control, the higher the average spectral frequency, the more the color temperature shift is towards warm white. Channels 4 and 5 have spectral frequencies exceeding 600Hz, therefore their color temperature shift is +300K, which is completely consistent with the mapping rules in Table 1. All channels in the diagram are steady-state acoustic channels identified after deformation compensation, and their slopes are consistently below the steady-state threshold, representing a stable and smooth transmission path of sound energy in the acoustic field. The acoustic features of these channels are independently mapped to lighting control commands, which can be used for zone control or weighted fusion to achieve more precise lighting adjustment.

[0040] In one embodiment of the present invention, a Hilbert-Huang transform is applied to the sound pressure level time series of the ambient sound signal to obtain a series of intrinsic mode function components. Based on the instantaneous frequencies of these intrinsic mode function components, all components whose instantaneous frequencies fall within the fundamental frequency range of the audio signal are selected. These selected components are reconstructed into a new signal, which is the fundamental frequency energy mode. The fundamental frequency energy mode is subtracted from the original ambient sound signal, and the remaining signal component is defined as the residual noise oscillation mode. According to the target illuminance value command and color temperature offset command, a composite reference waveform consisting of a DC bias component superimposed with a specific AC modulation component is generated, which serves as the steady-state light signal. The steady-state light signal and the separated residual noise oscillation mode are convolved and superimposed in the time domain to obtain an initial light waveform. This initial light waveform is subjected to nonlinear normalization and high-frequency jitter suppression processing to ensure that its amplitude range conforms to the driving voltage or current specifications of the lighting equipment and that the waveform is smooth and continuous. The processed waveform is the final light driving waveform signal used to generate the driving signal. A pulse width modulation signal for controlling the brightness of the lighting equipment is generated based on the light driving waveform signal.

[0041] In practice, the Hilbert-Huang Transform is applied to the sound pressure level time series of the ambient sound signal. The Hilbert-Huang Transform decomposes the signal into a series of intrinsic mode function (IMF) components, which are oscillating modes with instantaneous frequency physical meaning. Based on the instantaneous frequency of each IMF component, IMF components whose instantaneous frequencies fall within the audio fundamental frequency range (a preset frequency interval, such as 80 Hz to 400 Hz) are selected. All selected IMF components with instantaneous frequencies within the audio fundamental frequency range are linearly superimposed to reconstruct a new signal; this reconstructed signal is the fundamental frequency energy mode. The fundamental frequency energy mode is subtracted from the sound pressure level time series of the original ambient sound signal. The remaining waveform components after subtraction include harmonics, non-harmonic components, and environmental noise; these remaining waveform components are defined as the residual noise oscillation mode. It can be understood that the fundamental frequency energy mode mainly carries the tonal information of the sound, while the residual noise oscillation mode contains timbre characteristics and environmental disturbance information.

[0042] In some embodiments, a steady-state optical signal is generated based on a target illuminance value command and a color temperature offset command. The target illuminance value command determines the average intensity or DC bias component of the steady-state optical signal, while the color temperature offset command affects the spectral distribution or AC modulation component characteristics of the steady-state optical signal. Specifically, the steady-state optical signal is represented as a composite reference waveform with a DC bias superimposed on a specific AC modulation component, and its mathematical expression can be considered as follows: in: This represents the amplitude of the steady-state optical signal at time t. This represents the DC bias amplitude determined by the target illuminance value command. and This represents the AC modulation amplitude and modulation frequency determined by the color temperature offset command. This represents the initial phase. In practice, the steady-state optical signal and the previously separated residual noise oscillation mode are convolved and superimposed in the time domain. The convolution and superposition operation maps the time-varying characteristics of the residual noise oscillation mode onto the waveform of the steady-state optical signal, generating an initial light waveform. It can be understood that the convolution and superposition operation achieves the dynamic fusion of acoustic disturbance characteristics in the optical driving signal.

[0043] In practical implementation, the initial light waveform obtained after convolution and superposition undergoes nonlinear normalization. This nonlinear normalization maps the amplitude range of the initial light waveform to the voltage or current range allowed by the lighting device's driving circuit, such as 0 volts to 5 volts, using a preset compression function. Simultaneously, high-frequency jitter suppression is applied to the initial light waveform. This suppression uses a low-pass filter function to attenuate frequency components above a preset cutoff frequency, ensuring a smooth and continuous final waveform and preventing perceptible flickering caused by abrupt changes in the driving signal. The waveform after nonlinear normalization and high-frequency jitter suppression is the final light driving waveform signal used to generate the pulse width modulation signal. Optionally, the compression function used for nonlinear normalization can be a Sigmoid function or a variant thereof, and the low-pass filter function used for high-frequency jitter suppression can be a Butterworth filter or a Chebyshev filter.

[0044] In some embodiments, a pulse width modulation (PWM) signal for controlling the brightness of the lighting device is generated based on the light-driving waveform signal. The instantaneous amplitude of the light-driving waveform signal determines the duty cycle of the PWM signal at the corresponding moment; a high amplitude corresponds to a high duty cycle, and a low amplitude corresponds to a low duty cycle. It can be understood that the PWM signal is a standard control signal recognizable by the lighting device. By adjusting the duty cycle, it controls the average current flowing through the light source, thereby achieving precise and dynamic adjustment of the light brightness and color temperature.

[0045] See Figure 5This is a monitoring map of acoustic energy streamlines and slope abrupt changes, showing five acoustic energy streamlines generated on a time pane matrix, with a slope abrupt change point marked on one of the streamlines. The map contains five periodically fluctuating acoustic energy streamlines, each corresponding to a different energy baseline level. The waveforms are stable and phase-consistent, representing the normal transmission path of sound energy in the acoustic field. All streamlines exhibit sinusoidal periodic changes with a period of approximately four time units, consistent with the fundamental frequency energy mode characteristics of ambient sound signals. The red dot marks a slope abrupt change point on streamline 3, located at approximately time 5, with an acoustic energy of approximately 4.9. This point represents the moment when the streamline slope deviates from the mean by more than the abrupt change detection threshold, signifying the occurrence of an acoustic disturbance event and serving as a key basis for extracting acoustic disturbance feature packets. Each streamline represents the flow path of sound energy from a high acoustic density region to a low acoustic density region in the two-dimensional acoustic field, and its slope change reflects the rate of energy convergence and diffusion.

[0046] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent lighting control based on sound changes, characterized in that, include: Acquire ambient sound signals within the target sound field area, wherein the ambient sound signals include a time series of sound pressure levels generated by air vibrations; The ambient sound signal is imported into the voiceprint feature analysis engine for dynamic slicing. A time window matrix that expands and contracts with the sound intensity is constructed in the voiceprint feature analysis engine. Voiceprint energy streamlines are generated between the grids of the time window matrix according to the sound pressure level time series. Track and capture slope abrupt events of the acoustic energy streamline. When the local slope of the acoustic energy streamline exceeds a preset slope threshold, extract the waveform segment of the acoustic energy streamline and its corresponding time pane matrix region to form an acoustic disturbance feature package. Using the acoustic disturbance feature packet as input, a brightness controller based on acoustic deformation feedback is driven. The brightness controller based on acoustic deformation feedback outputs a light intensity compensation deformation command according to the spectral migration mode of the waveform segment in the acoustic disturbance feature packet and the energy density distribution of the time pane matrix region. According to the light intensity compensation deformation command, the corresponding compensation deformation is applied to the time window matrix in the voiceprint feature analysis engine, and the time window matrix after compensation deformation generates a new voiceprint energy streamline distribution. The steady-state audio channel is identified in the new audio energy streamline distribution, and the steady-state audio channel is mapped back to the physical light control space and converted into target illuminance value command and color temperature offset command.

2. The intelligent lighting control method based on sound changes according to claim 1, characterized in that, The ambient sound signal is imported into a voiceprint feature analysis engine for dynamic slicing. A time-pane matrix that scales and adjusts with sound intensity is constructed within the voiceprint feature analysis engine, including: The sound intensity level and the centroid frequency of the spectrum are calculated at each sampling time based on the sound pressure level time series. A two-dimensional acoustic field is constructed with the sound intensity level value as the horizontal axis coordinate and the frequency of the spectral centroid as the vertical axis coordinate, and the acoustic state point at each sampling time is plotted in the two-dimensional acoustic field. Connecting the voiceprint state points at adjacent sampling times in the two-dimensional voiceprint field forms a dynamic voiceprint trajectory line; Based on the spatial density of the dynamic voiceprint trajectory line, a non-uniform rectangular grid is automatically generated in the two-dimensional voiceprint field to cover the dynamic voiceprint trajectory line. The time width of the rectangular grid is positively correlated with the sound intensity level value. The rectangular grid is the time pane matrix in the voiceprint feature analysis engine, which adjusts its grid size and arrangement relationship as the voiceprint state point changes.

3. The intelligent lighting control method based on sound changes according to claim 2, characterized in that, The step of generating acoustic energy streamlines between the grids of the time pane matrix based on the sound pressure level time series includes: For each rectangular grid cell in the time pane matrix, calculate the gradient vector between the sound intensity level value at its four vertices and the frequency value at the centroid of the spectrum; The gradient vectors of the four vertices within each rectangular grid cell are vector synthesized to obtain the acoustic flow vector of the rectangular grid cell; Along the sound signature flow vector field, streamlines from high sound signature density regions to low sound signature density regions are drawn on the time window matrix. These streamlines are the sound signature energy streamlines that characterize the real-time convergence and diffusion path of sound energy.

4. The intelligent lighting control method based on sound changes according to claim 3, characterized in that, The tracking and capturing of slope abrupt events in the acoustic energy streamline includes: The local slope of each acoustic energy streamline is calculated with a fixed step size to form a slope sequence of acoustic energy streamlines; Apply sliding window mean calculation to the slope sequence to detect the time points in the slope sequence that deviate from the mean by more than the mutation detection threshold, and mark them as the time of occurrence of slope mutation events; Back to the moment of occurrence, extract the acoustic energy streamline segments within a preset time window before and after the moment of occurrence, and simultaneously record the set of rectangular grid cells in the time window matrix through which the acoustic energy streamline segments pass. The acoustic energy streamline segments and the set of rectangular grid cells together constitute the acoustic disturbance feature package.

5. The intelligent lighting control method based on sound changes according to claim 4, characterized in that, Using the acoustic disturbance feature packet as input, a brightness controller based on acoustic pattern deformation feedback is driven, comprising: The waveform segments within the acoustic disturbance feature package are subjected to joint time-frequency analysis to extract the dominant peak frequency and harmonic attenuation coefficient in the time spectrum, which are used as spectral migration mode features. Analyze the time pane matrix region within the acoustic disturbance feature package, and calculate the variance of the sound intensity level values ​​of all rectangular grid cells within the time pane matrix region and the dispersion of the spectral centroid frequency as energy density distribution characteristics. The spectral migration mode features and energy density distribution features are input into a pre-trained deformation decision network. The deformation decision network outputs instructions to apply expansion, contraction or translation deformation to the target region of the time pane matrix, as well as the deformation amplitude and direction corresponding to each deformation instruction, which together constitute the light intensity compensation deformation instructions.

6. The intelligent lighting control method based on sound changes according to claim 5, characterized in that, The step of applying corresponding compensation deformation to the time pane matrix in the voiceprint feature parsing engine according to the light intensity compensation deformation command includes: Locate the target area specified by the light intensity compensation deformation command in the two-dimensional acoustic field, wherein the target area is identified by a set of rectangular grid cells; According to the deformation type and deformation amplitude in the light intensity compensation deformation command, the center coordinates of all rectangular grid cells in the target area are synchronously transformed. After the center coordinate transformation, the physical boundaries and adjacency relationships of all affected rectangular grid cells are recalculated and updated to complete the construction of the time pane matrix after deformation compensation.

7. The intelligent lighting control method based on sound changes according to claim 6, characterized in that, The identification of steady-state audio channels in the new audio energy streamline distribution includes: the steady-state audio channel is defined by the slope of the audio energy streamline being continuously lower than a steady-state threshold; On the time pane matrix after deformation compensation, the gradient vector of each grid cell is calculated and the acoustic signature flow vector field is synthesized. The acoustic signature energy streamline from the high acoustic signature density region to the low acoustic signature density region is regenerated along the flow field. Monitor the slope change of newly generated voiceprint energy streamlines within a preset observation period. If the maximum slope of a certain voiceprint energy streamline is consistently lower than the steady-state threshold throughout the entire observation period, then the voiceprint energy streamline segment is determined to be a steady-state voiceprint channel. Record the starting point coordinates, ending point coordinates, and the sequence of raster cells traversed by all steady-state acoustic channels in the two-dimensional acoustic field.

8. The intelligent lighting control method based on sound changes according to claim 7, characterized in that, The process of mapping the steady-state acoustic channel back to the physical light control space and converting it into target illuminance value instructions and color temperature offset instructions includes: For each steady-state acoustic text channel, the average sound intensity level and average spectral centroid frequency of the steady-state acoustic text channel in the two-dimensional acoustic text field are calculated based on its starting point coordinates and ending point coordinates. By using a preset sound-light mapping relationship, the average sound intensity level value is converted into an illuminance value command for the target area, and the average spectral centroid frequency value is converted into a color temperature offset command relative to the reference white light.

9. The intelligent lighting control method based on sound changes according to claim 8, characterized in that, Also includes: The steady-state light signal defined by the target illuminance value command and the color temperature offset command is fused with the residual noise oscillation mode to synthesize the light driving waveform signal; Based on the light driving waveform signal, a pulse width modulation signal for the lighting device is generated to realize intelligent control of the light based on sound changes; The environmental sound signal is subjected to overtone mode stripping to separate the fundamental frequency energy mode and the residual noise oscillation mode; The process of performing overtone mode stripping on the environmental sound signal to separate the fundamental frequency energy mode and the residual noise oscillation mode includes: Hilbert-Huang transform is applied to the sound pressure level time series of the ambient sound signal to obtain a series of intrinsic mode function components. Based on the instantaneous frequency of the intrinsic mode function components, intrinsic mode function components whose instantaneous frequency is within the audio fundamental frequency range are selected. The intrinsic mode function components are reconstructed into fundamental frequency energy modes. The original ambient sound signal is subtracted from the fundamental frequency energy modes, and the remaining waveform components are the residual noise oscillation modes. The step of fusing the steady-state light signal defined by the target illuminance value command and the color temperature offset command with the residual noise oscillation mode to synthesize the light driving waveform signal includes: Based on the target illuminance value command and the color temperature offset command, a composite reference waveform with DC bias and AC modulation is generated as a steady-state light signal. The initial light waveform is obtained by convolving and superimposing the steady-state light signal with the residual noise oscillation mode in the time domain. The initial light waveform is subjected to nonlinear normalization and high-frequency jitter suppression to ensure that its value range meets the driving specifications of the lighting equipment and that the waveform is smooth and continuous. The processed waveform is the final light driving waveform signal used to generate the pulse width modulation signal.

10. A sound-based intelligent lighting control system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the intelligent lighting control method based on sound changes as described in any one of claims 1 to 9.

Images