Adaptive sound field optimization control system based on reinforcement learning

Abstract

The invention discloses a self-adaptive sound field optimization control system based on reinforcement learning, relates to the technical field of intelligent control and acoustic engineering, and is used for solving the problems of weak abnormal recognition capability and lack of an intelligent regulation and control mechanism caused by high environmental complexity and diverse spatial structure changes. A sound field characteristic model is established by adopting a support vector machine through obtaining sound wave propagation path length, spatial reflectivity, sound field vibration frequency, an energy consumption fluctuation value and sound field inclination angle information, a sound field state is analyzed in combination with a hierarchical clustering method, and an inclination adjustment threshold value is generated. Judging whether the sound field is inclined or not through the inclination angle, monitoring the energy distribution density in real time to obtain the energy loss proportion, and analyzing the abnormal level of the sound field by adopting fuzzy logic in combination with the airflow disturbance intensities at different heights. And the inclination angle information is compared with an adjustment threshold value, and accurate adjustment and control of the sound field are realized according to a comparison result or an abnormal level.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control and acoustic engineering, and more specifically, to an adaptive sound field optimization control system based on reinforcement learning. Background Art

[0002] Sound field control technology, as a key enabler for achieving a high-quality auditory experience, has been widely researched and applied in a variety of application scenarios, such as intelligent conferencing systems, virtual reality spaces, voice recognition environments, and immersive audio systems. Traditional sound field control systems typically rely on static acoustic modeling, linear gain adjustment, and limited feedback control strategies to adjust sound propagation paths, reflection angles, and energy distribution.

[0003] The existing technology has the following deficiencies:

[0004] At present, due to the high complexity of the environment, diverse changes in spatial structure, and the dynamic nature of sound sources and receiving points, traditional methods generally adopt static modeling strategies based on empirical parameters. These methods cannot accurately reflect the dynamic coupling relationship between sound wave propagation path, reflectivity and vibration frequency, thereby reducing modeling accuracy, having weak anomaly recognition ability, and lacking intelligent control mechanism. Therefore, an adaptive sound field optimization control system based on reinforcement learning is proposed.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an adaptive sound field optimization control system based on reinforcement learning, which solves the problems raised in the above-mentioned background technology by applying reinforcement learning strategy, adaptive modeling algorithm and multi-source fusion perception mechanism.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an adaptive sound field optimization control system based on reinforcement learning, comprising a sound field data acquisition module, a parameter control module, an abnormal state detection module, and a feedback execution module;

[0008] The sound field data acquisition module is used to obtain the sound wave propagation path length and spatial reflectivity of the sound field, monitor the sound field vibration frequency and energy consumption fluctuation value during the interaction with the target sound field, transmit the sound wave propagation path length, spatial reflectivity, sound field vibration frequency and energy consumption fluctuation value to the parameter control module, and simultaneously collect the sound field tilt angle information and transmit it to the abnormal state detection module and feedback execution module;

[0009] The parameter control module uses the support vector machine algorithm to build a sound field characteristic model based on the length of the sound wave propagation path and the spatial reflectivity. It uses the hierarchical clustering method to analyze the sound field state based on the comprehensive sound field vibration frequency and energy consumption fluctuation value. The sound field characteristic model and the sound field state are combined to generate the tilt adjustment threshold and pass it to the feedback execution module.

[0010] The abnormal state detection module determines whether the target sound field is tilted based on the sound field tilt angle information, and monitors the sound field energy distribution density in real time to obtain the energy loss ratio. It uses fuzzy logic to analyze the sound field abnormality level based on the airflow disturbance intensity corresponding to different heights, and transmits the sound field abnormality level to the feedback execution unit;

[0011] The feedback execution module receives the sound field tilt angle information and compares it with the tilt adjustment threshold, and adjusts the target sound field according to the comparison result or the sound field abnormality level.

[0012] In a preferred embodiment, the sound field data acquisition module collects the time difference between the emission and reception of the sound wave to calculate the length of the sound wave propagation path;

[0013] Collect the sound pressure amplitude on the incident and reflected paths to calculate the spatial reflectivity;

[0014] Perform fast Fourier transform on the sound wave signal to obtain the spectrum structure, and extract the maximum frequency in the spectrum structure as the sound field vibration frequency;

[0015] The energy consumption fluctuation value is the variation range of the power of the sound field, and the power value of the sound field is collected to calculate the energy consumption fluctuation value;

[0016] The sound field tilt angle is the angle at which the actual propagation direction of the sound field energy is offset from the preset propagation direction. The sound field tilt angle is obtained by collecting the sound pressure value, calculating the sound energy density distribution, and deducing the actual propagation direction of the sound field energy and comparing it with the preset propagation direction.

[0017] In a preferred embodiment, the parameter control module uses the support vector machine algorithm to construct a sound field characteristic model based on the sound wave propagation path length and spatial reflectivity:

[0018] Set input parameters: collect the sound wave propagation path lengths corresponding to N sound wave receivers at the same time, and the spatial reflectivity of the sound wave receivers. Construct a two-dimensional vector and merge it into a feature vector set as input data.

[0019] Set kernel function: Select radial basis function as the kernel function to perform feature transformation on the two-dimensional vectors in the feature vector set;

[0020] Output sound field characteristics: Calculate the kernel function value and support vector coefficient to construct a prediction function and obtain the predicted value of the sound field characteristics.

[0021] In a preferred embodiment, the sound field vibration frequency and energy consumption fluctuation value are comprehensively classified using a hierarchical clustering method:

[0022] Data preparation: Within the preset period, the data are evenly divided into M acquisition moments to obtain the sound field vibration frequency and energy consumption fluctuation values ​​and normalize them. The same acquisition moment is merged into a sample vector and recorded as z i , each sample vector is regarded as an independent cluster, and the preset target number of clusters is recorded as K;

[0023] Calculate the inter-cluster distance: Calculate the Euclidean distance between any two clusters as the inter-cluster distance;

[0024] Cluster merging: In each iteration, two clusters with the smallest distance are selected for merging, and the mean of their sample vectors is used to update the new cluster. This process is repeated until the number of clusters is equal to K.

[0025] Output sound field status: The sound field status is divided into K clusters, and the sound field status index is calculated based on the standard deviation of the sound field vibration frequency and the mean of the energy consumption fluctuation value in each cluster.

[0026] In a preferred embodiment, the predicted value of the sound field characteristics is fused with the sound field state characteristics through a linear weighting function to calculate the sensitivity of the current sound field to the tilt angle;

[0027] The sensitivity is calculated using a nonlinear activation function to obtain the tilt adjustment threshold.

[0028] In a preferred embodiment, the abnormal state detection module compares the sound field tilt angle with the tilt adjustment threshold: if the sound field tilt angle is greater than the tilt adjustment threshold, it is determined that the sound field is tilted; otherwise, it is determined that the sound field is not tilted.

[0029] In a preferred embodiment, the energy loss ratio is the attenuation ratio of the sound energy density in the sound field, and the energy loss ratio is obtained by calculating the sound energy density at the current moment and comparing it with the historical average sound energy density.

[0030] In a preferred embodiment, the air pressure values ​​at different heights of the sound field are detected, and the air pressure variation amplitude is calculated as the airflow disturbance intensity corresponding to the different heights;

[0031] The energy loss ratio and the airflow disturbance intensity at different heights are combined to evaluate the abnormal level of the sound field using fuzzy reasoning.

[0032] The energy loss ratio and the airflow disturbance intensity corresponding to different heights are defined as input variables and divided into different fuzzy sets respectively.

[0033] The sound field abnormality level is defined as the output variable and divided into fuzzy sets;

[0034] A set of fuzzy rules is formulated to describe the influence of different input variables on output variables, and the abnormality level of the sound field is evaluated based on fuzzy reasoning;

[0035] Fuzzy reasoning is performed based on fuzzy rules to divide the abnormal level of the sound field into three levels: low, medium and high.

[0036] In a preferred embodiment, the feedback execution module adjusts the target sound field according to whether the sound field is tilted and the level of sound field abnormality:

[0037] If the sound field is tilted and the sound field abnormality level is high, the target sound field is adjusted, the sound energy density is injected into the low sound energy density area, and the actual propagation direction of the sound field energy is adjusted to the preset propagation direction;

[0038] If the sound field is not tilted and the sound field abnormality level is high, the target sound field is adjusted to inject sound energy density into the low sound energy density area;

[0039] If the sound field does not tilt and the sound field abnormality level is medium or low, no adjustment is made to the target sound field and the sound field state is stable;

[0040] If the sound field is tilted and the sound field abnormality level is medium or low, the target sound field is adjusted and the actual propagation direction of the sound field energy is adjusted to the preset propagation direction.

[0041] Technical effects and advantages of the present invention:

[0042] The present invention obtains the sound wave propagation path length, spatial reflectivity, sound field vibration frequency, energy consumption fluctuation value and sound field tilt angle information of the sound field, and uses a support vector machine algorithm to construct a sound field characteristic model based on the sound wave propagation path length and spatial reflectivity. The sound field vibration frequency and energy consumption fluctuation value are comprehensively analyzed using a hierarchical clustering method. A tilt adjustment threshold is generated in combination with the sound field characteristic model. Whether the target sound field is tilted is judged according to the sound field tilt angle information, and the sound field energy distribution density is monitored in real time to obtain the energy loss ratio. The sound field anomaly level is analyzed through fuzzy logic in combination with the airflow disturbance intensity corresponding to different heights. The sound field tilt angle information is compared with the tilt adjustment threshold. The target sound field is adjusted according to the comparison result or the sound field anomaly level, thereby realizing accurate judgment and control of the sound field tilt state. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart for implementing the adaptive sound field optimization control system based on reinforcement learning of the present invention.

[0044] Figure 2 Schematic diagram of the module architecture of the adaptive sound field optimization control system based on reinforcement learning of the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] Example 1

[0047] See also Figures 1 to 2 ,The adaptive sound field optimization control system based on reinforcement learning, includes a sound field data acquisition module, a parameter control module, an abnormal state detection module and a feedback execution module, and each module is connected with electrical signals;

[0048] The functions of each module are as follows:

[0049] The sound field data acquisition module is used to obtain the sound wave propagation path length and spatial reflectivity of the sound field, monitor the sound field vibration frequency and energy consumption fluctuation value during the interaction with the target sound field, transmit the sound wave propagation path length, spatial reflectivity, sound field vibration frequency and energy consumption fluctuation value to the parameter control module, and simultaneously collect the sound field tilt angle information and transmit it to the abnormal state detection module and feedback execution module;

[0050] The parameter control module uses the support vector machine algorithm to build a sound field characteristic model based on the length of the sound wave propagation path and the spatial reflectivity. It uses the hierarchical clustering method to analyze the sound field state based on the comprehensive sound field vibration frequency and energy consumption fluctuation value. The sound field characteristic model and the sound field state are combined to generate the tilt adjustment threshold and pass it to the feedback execution module.

[0051] The abnormal state detection module determines whether the target sound field is tilted based on the sound field tilt angle information, and monitors the sound field energy distribution density in real time to obtain the energy loss ratio. It uses fuzzy logic to analyze the sound field abnormality level based on the airflow disturbance intensity corresponding to different heights, and transmits the sound field abnormality level to the feedback execution unit;

[0052] The feedback execution module receives the sound field tilt angle information and compares it with the tilt adjustment threshold, and adjusts the target sound field according to the comparison result or the sound field abnormality level.

[0053] The specific implementation is as follows:

[0054] The length of the sound wave propagation path in the sound field data acquisition module is the total length of the spatial path from the sound source to the receiver. The difference in sound wave propagation time between the sound wave transmitter and the sound wave receiver is recorded as Δt. The sound wave propagation path length is: L = v·Δt, where v is the speed of sound and L is the length of the sound wave propagation path.

[0055] The spatial reflectivity is the ratio of the sound wave reflected by the environment during the propagation process. The sound pressure amplitude of the sound wave transmitter and the sound wave receiver are recorded as P and P, respectively. in and P ref , the spatial reflectivity is expressed as: Among them, R s is the spatial reflectivity;

[0056] After continuous sampling of the acoustic signal in the acoustic transmitter, a fast Fourier transform is performed to obtain the spectrum structure, and the maximum frequency in the spectrum structure is extracted as the sound field vibration frequency, which is recorded as f;

[0057] The energy consumption fluctuation value is the amplitude of the change in the energy consumption of the sound field. The power value of the sound field is obtained through the power sensor at the preset collection time, and the fluctuation degree of the energy consumption of the sound field is calculated: Among them, P(t k ) is t k The power value at the moment, is the average power value during the acquisition time, N is the total number of sampling moments, and ΔE is the energy consumption fluctuation value;

[0058] The sound wave propagation path length, spatial reflectivity, sound field vibration frequency and energy consumption fluctuation value are transmitted to the parameter control module.

[0059] The sound field tilt angle is the angle at which the actual propagation direction of the sound field energy deviates from the preset propagation direction. The sound pressure value at each measuring point is obtained by using acoustic sensors at different positions in the sound field to calculate the sound energy density: Among them, p i is the sound pressure value, c is the speed of sound, ρ is the air density, E i is the sound energy density, and the actual propagation direction is calculated based on the sound energy density: in, is the spatial position vector relative to the center of the sound source, is the actual propagation direction of the sound field energy; the sound field tilt angle is: in, is the preset propagation direction, θ s is the sound field tilt angle;

[0060] The sound wave propagation path length, spatial reflectivity, sound field vibration frequency and energy consumption fluctuation value are transmitted to the parameter control module, and the sound field tilt angle information is transmitted to the abnormal state detection module and the feedback execution module.

[0061] It should be noted that the sound wave transmitter records the time, frequency and pulse shape of the sound wave; the sound wave receiver captures the actual signal after the sound wave propagates in space; the sound pressure amplitude is the maximum value of the pressure fluctuation caused by the sound wave per unit time; the Fourier transform is a mathematical tool used to analyze the frequency characteristics of the sound wave signal, extract the vibration frequency of the sound field, and construct the spectrum distribution; the inclination sensor is a sensor used to measure the inclination angle or attitude angle of an object relative to the direction of gravity, and is usually used to monitor the tilt state of equipment, structure or platform.

[0062] In the parameter control module, the sound wave propagation path length and spatial reflectivity are used to construct a sound field characteristic model using the support vector machine algorithm to reflect the matching degree between the current sound field and the target sound field. The specific steps are as follows:

[0063] Set the input parameters: The sound wave propagation path lengths corresponding to the N sound wave receivers collected at the same time are normalized and marked as {L1, L2, L3...L N}, and the spatial reflectivity of the corresponding acoustic receiver is normalized and marked as {R1, R2, R3...R N}, with [L i ,R i ] to construct a two-dimensional vector labeled x i , merged into a feature vector set as input data;

[0064] Set kernel function: Select radial basis function as the kernel function to perform feature transformation on the two-dimensional vector in the feature vector set. The feature transformation formula is: Among them, e is the natural base, γ is the hyperparameter, K(x i ,x j ) is to substitute any two two-dimensional vectors into the calculation of their similarity in the high-dimensional space;

[0065] It should be explained that the hyperparameter is a parameter used to adjust the width of the kernel function. The hyperparameter can be set to the inverse of the input feature dimension. The specific setting is done by professionals and will not be elaborated here.

[0066] Output sound field characteristics: The output result is a function that reflects the sound field characteristics, denoted as Y. The kernel function value and support vector coefficient calculated using the training sample are used to construct the support vector machine prediction function: Among them, α i 、 is the support vector coefficient learned by the support vector machine model, b is the bias term, x is the current input two-dimensional vector, and Y is the predicted value of the sound field characteristics;

[0067] For example, assuming that the two-dimensional vectors in the feature vector set include x1 = [5.0, 0.8] and x2 = [7.0, 0.5], the parameters automatically learned through the model training process are: The bias term is: b = 0.1. If the current two-dimensional vector is: x = [6.0, 0.7], then the kernel function values ​​of the current two-dimensional vector and the feature vector set are: K(x1, x) = exp(-0.5×1.01) = exp(-0.505) ≈ 0.604, K(x2, x) = exp(-0.5×1.04) = exp(-0.52) ≈ 0.594, then the sound field characteristic prediction value of the current sample is: Y = 0.6×0.604+(-0.4)×0.594+0.1 = 0.2248.

[0068] The hierarchical clustering method is used to classify the sound field status categories based on the comprehensive sound field vibration frequency and energy consumption fluctuation value:

[0069] Data preparation: Within the preset period, the data is evenly divided into M acquisition moments to obtain the sound field vibration frequency and energy consumption fluctuation value. The same acquisition moment is merged into a sample vector and recorded as z i , the sample vectors of M acquisition moments are normalized, each sample vector is regarded as an independent cluster, and the preset target cluster number is recorded as K;

[0070] Calculate the inter-cluster distance: Calculate the Euclidean distance between any two clusters as the inter-cluster distance: Among them, d(z i ,z j ) is z i and z j The distance between clusters;

[0071] Cluster merging: Select the two clusters with the smallest distance and merge them into a new cluster. Take the average of the sample vectors in the new cluster and recalculate the distances of all clusters. Repeat this process until the number of clusters is equal to K, and then construct a hierarchical clustering tree.

[0072] Output sound field status: The sound field status is divided into K clusters. The standard deviation of the sound field vibration frequency in each cluster is calculated as m, and the mean of the energy consumption fluctuation value is recorded as e. The sound field status index is calculated based on the standard deviation of the sound field vibration frequency and the mean of the energy consumption fluctuation value: μ = 1-α·m-β·e, where α and β are preset weight coefficients and μ is the sound field status characteristic;

[0073] The real-time collected vibration frequency and energy consumption fluctuation feature vectors are mapped to the cluster corresponding to the sound field state, and the current sound field state index is calculated.

[0074] For example, when the preset target cluster number K is 2, the sound field state is divided into two categories, recorded as S1 and S2 respectively. When the sound field state is S1, it indicates that the sound field is in a stable state. When the sound field state is S2, it indicates that the sound field is in an abnormal state. The preset target cluster number and the preset weight coefficient are set by professionals and will not be elaborated here.

[0075] The predicted sound field characteristics are combined with the sound field state characteristics through a linear weighting function to calculate the sensitivity of the current sound field to the tilt angle: F = ω1·Y + ω2·μ, where ω1 and ω2 are weight coefficients and F is the sensitivity.

[0076] A nonlinear activation function is used to map the sensitivity to the tilt adjustment threshold: Among them, F0 is the preset sensitivity neutral point, T max is the preset maximum tilt adjustment range, and T is the tilt adjustment threshold;

[0077] It should be explained that F-F0 is the deviation of sensitivity from the neutral point, which determines the size of the adjustment threshold. The farther away from the preset sensitivity neutral point, the greater the risk of sound field tilt and the higher the adjustment threshold. max Used to limit the maximum intensity of adjustment to prevent exceeding the adjustment range.

[0078] The abnormal state detection module compares the sound field tilt angle with the tilt adjustment threshold: if the sound field tilt angle is greater than the tilt adjustment threshold, it is determined that the sound field is tilted; otherwise, it is determined that the sound field is not tilted;

[0079] The energy loss ratio is the attenuation ratio of the sound energy density in the sound field. The average sound energy density of each sampling point is recorded as E c , select a period of time in the past and calculate the average sound energy density of the target sound field, recorded as E ref , then the energy loss ratio is: Among them, ΔE is the energy loss ratio.

[0080] The acquisition period is preset, and the vertical direction of the sound field is divided into multiple height layers. The air pressure values ​​at different heights are detected by the micro-air pressure disturbance sensor, and the air pressure change amplitude is calculated as the airflow disturbance intensity corresponding to different heights: Among them, P(h i ,t) is the altitude layer h at time t i The corresponding air pressure value is is the altitude layer h i The corresponding average air pressure value, T is the preset collection period, D(h i ) is the airflow disturbance intensity corresponding to different heights.

[0081] The energy loss ratio and the airflow disturbance intensity at different heights are combined to evaluate the abnormal level of the sound field using fuzzy reasoning. The specific steps are as follows:

[0082] The energy loss ratio and the airflow disturbance intensity corresponding to different heights are defined as input variables and divided into different fuzzy sets respectively.

[0083] For example, "high", "medium", and "low" are for energy loss ratio, and "strong", "medium", and "weak" are for airflow disturbance intensity;

[0084] The sound field abnormality level is defined as the output variable and divided into fuzzy sets, for example, "high", "medium", and "low" for the water well adjustment level.

[0085] Formulate a set of fuzzy rules to describe the impact of different input variables on output variables, and evaluate the abnormal level of the sound field based on fuzzy reasoning. Fuzzy rules can be formulated according to actual conditions, for example:

[0086] Rule 1: If the energy loss ratio is high and the airflow disturbance intensity is strong, the sound field abnormality level is high;

[0087] Rule 2: If the energy loss ratio is medium and the airflow disturbance intensity is medium, the sound field abnormality level is medium;

[0088] Rule 3: If the energy loss ratio is low and the airflow disturbance intensity is weak, the sound field abnormality level is low;

[0089] Rule 4: If the energy loss ratio is high but the airflow disturbance intensity is weak, the sound field abnormality level is medium;

[0090] Rule 5: If the energy loss ratio is medium and the airflow disturbance intensity is strong, the sound field abnormality level is high; ...

[0092] Fuzzy reasoning is performed based on fuzzy rules. When the sound field anomaly level is high, it indicates that the actual propagation direction of the current sound field has a spatial offset exceeding the maximum allowable offset angle relative to the preset propagation direction; when the sound field anomaly level is medium, it indicates that the sound field tilt angle is within the allowable range; when the sound field anomaly level is low, it indicates that the sound field structure is stable.

[0093] It should be noted that the micro-air pressure disturbance sensor is a highly sensitive sensing device used to monitor in real time the subtle air pressure changes or local disturbance pressure fluctuations in the environment, reflecting the non-steady-state disturbances and local aerodynamic flow field changes in the sound field operating environment; the distributed sound pressure sensor array refers to a collection of multiple sound pressure sensor units arranged in a predetermined geometric layout in the target sound field space, which realizes real-time monitoring of the sound pressure distribution in different spatial positions of the sound field by synchronously collecting sound pressure signals at multiple points; the division of fuzzy sets can be adjusted according to actual conditions, and the comprehensive energy loss ratio and the airflow disturbance intensity corresponding to different heights can be subdivided into five fuzzy sets of "extremely high", "high", "medium", "low" and "extremely low" according to historical statistical data, thereby improving the classification granularity, enhancing the system's sensitivity to abnormal conditions and adjustment accuracy, so as to facilitate better precise adjustment according to different conditions. The specific settings are made by professionals and will not be elaborated here.

[0094] The feedback execution module receives the sound field tilt angle information, compares it with the tilt adjustment threshold, and adjusts the target sound field based on the sound field abnormality level:

[0095] If the sound field is tilted and the sound field abnormality level is high, the target sound field is adjusted, the sound energy density is injected into the low sound energy density area, and the actual propagation direction of the sound field energy is adjusted to the preset propagation direction;

[0096] If the sound field is not tilted and the sound field abnormality level is high, the target sound field is adjusted to inject sound energy density into the low sound energy density area;

[0097] If the sound field does not tilt and the sound field abnormality level is medium or low, no adjustment is made to the target sound field and the sound field state is stable;

[0098] If the sound field is tilted and the sound field abnormality level is medium or low, the target sound field is adjusted and the actual propagation direction of the sound field energy is adjusted to the preset propagation direction.

[0099] It needs to be explained that when the sound field is tilted and the sound field anomaly level is high, it means that the sound field structure is offset and the distribution of sound energy density in space is unbalanced. The sound energy density at different positions is detected, and sound energy is injected into the area with low sound energy density and the actual propagation direction of the sound field energy is corrected. When the sound field is not tilted and the sound field anomaly level is high, it means that the sound field structure is stable, but the local or overall sound energy is unbalanced, and sound energy is injected into the area with low sound energy density. When the sound field is tilted and the sound field anomaly level is medium or low, it means that the sound field density distribution is balanced, the sound field structure is offset, and the actual propagation direction of the sound field energy is corrected.

[0100] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0101] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0102] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0103] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0104] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0105] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0106] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0107] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0108] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0109] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. An adaptive sound field optimization control system based on reinforcement learning, characterized by: It includes sound field data acquisition module, parameter control module, abnormal state detection module and feedback execution module; The sound field data acquisition module is used to obtain the sound wave propagation path length and spatial reflectivity of the sound field, monitor the sound field vibration frequency and energy consumption fluctuation value during the interaction with the target sound field, transmit the sound wave propagation path length, spatial reflectivity, sound field vibration frequency and energy consumption fluctuation value to the parameter control module, and simultaneously collect the sound field tilt angle information and transmit it to the abnormal state detection module and feedback execution module; The parameter control module uses the support vector machine algorithm to build a sound field characteristic model based on the length of the sound wave propagation path and the spatial reflectivity. It uses the hierarchical clustering method to analyze the sound field state based on the comprehensive sound field vibration frequency and energy consumption fluctuation value. The sound field characteristic model and the sound field state are combined to generate the tilt adjustment threshold and pass it to the feedback execution module. The abnormal state detection module determines whether the target sound field is tilted based on the sound field tilt angle information, and monitors the sound field energy distribution density in real time to obtain the energy loss ratio. It uses fuzzy logic to analyze the sound field abnormality level based on the airflow disturbance intensity corresponding to different heights, and transmits the sound field abnormality level to the feedback execution unit; The feedback execution module receives the sound field tilt angle information and compares it with the tilt adjustment threshold, and adjusts the target sound field according to the comparison result or the sound field abnormality level.

2. The adaptive sound field optimization control system based on reinforcement learning according to claim 1, characterized in that: The sound field data acquisition module collects the time difference between the sound wave being sent and being received to calculate the length of the sound wave propagation path; Collect the sound pressure amplitude on the incident and reflected paths to calculate the spatial reflectivity; Perform fast Fourier transform on the sound wave signal to obtain the spectrum structure, and extract the maximum frequency in the spectrum structure as the sound field vibration frequency; The energy consumption fluctuation value is the variation range of the power of the sound field, and the power value of the sound field is collected to calculate the energy consumption fluctuation value; The sound field tilt angle is the angle at which the actual propagation direction of the sound field energy is offset from the preset propagation direction. The sound field tilt angle is obtained by collecting the sound pressure value, calculating the sound energy density distribution, and deducing the actual propagation direction of the sound field energy and comparing it with the preset propagation direction.

3. The adaptive sound field optimization control system based on reinforcement learning according to claim 2, characterized in that: In the parameter control module, the sound wave propagation path length and spatial reflectivity are used to construct a sound field characteristic model by adopting the support vector machine algorithm: Set input parameters: collect the sound wave propagation path lengths corresponding to N sound wave receivers at the same time, and the spatial reflectivity of the sound wave receivers. Construct a two-dimensional vector and merge it into a feature vector set as input data. Set kernel function: Select radial basis function as the kernel function to perform feature transformation on the two-dimensional vectors in the feature vector set; Output sound field characteristics: Calculate the kernel function value and support vector coefficient to construct a prediction function and obtain the predicted value of the sound field characteristics.

4. The adaptive sound field optimization control system based on reinforcement learning according to claim 2, characterized in that: The hierarchical clustering method is used to classify the sound field status categories based on the comprehensive sound field vibration frequency and energy consumption fluctuation value: Data preparation: Within the preset period, the data are evenly divided into M acquisition moments to obtain the sound field vibration frequency and energy consumption fluctuation values ​​and normalize them. The same acquisition moment is merged into a sample vector and recorded as z i , each sample vector is regarded as an independent cluster, and the preset target number of clusters is recorded as K; Calculate the inter-cluster distance: Calculate the Euclidean distance between any two clusters as the inter-cluster distance; Cluster merging: In each iteration, two clusters with the smallest distance are selected for merging, and the mean of their sample vectors is used to update the new cluster. This process is repeated until the number of clusters is equal to K. Output sound field status: The sound field status is divided into K clusters, and the sound field status index is calculated based on the standard deviation of the sound field vibration frequency and the mean of the energy consumption fluctuation value in each cluster.

5. The adaptive sound field optimization control system based on reinforcement learning according to claim 4, characterized in that: The predicted sound field characteristics are combined with the sound field state characteristics through a linear weighting function to calculate the sensitivity of the current sound field to the tilt angle. The sensitivity is calculated using a nonlinear activation function to obtain the tilt adjustment threshold.

6. The adaptive sound field optimization control system based on reinforcement learning according to claim 5, characterized in that: The abnormal state detection module compares the sound field tilt angle with the tilt adjustment threshold: if the sound field tilt angle is greater than the tilt adjustment threshold, it is determined that the sound field is tilted; otherwise, it is determined that the sound field is not tilted.

7. The adaptive sound field optimization control system based on reinforcement learning according to claim 1, characterized in that: The energy loss ratio is the attenuation ratio of the sound energy density in the sound field. The energy loss ratio is obtained by calculating the sound energy density at the current moment and comparing it with the historical average sound energy density.

8. The adaptive sound field optimization control system based on reinforcement learning according to claim 7, characterized in that: Detect the air pressure values ​​at different heights in the sound field and calculate the air pressure variation amplitude as the airflow disturbance intensity corresponding to different heights; The energy loss ratio and the airflow disturbance intensity at different heights are combined to evaluate the abnormal level of the sound field using fuzzy reasoning. The energy loss ratio and the airflow disturbance intensity corresponding to different heights are defined as input variables and divided into different fuzzy sets. The sound field abnormality level is defined as the output variable and divided into fuzzy sets; A set of fuzzy rules is formulated to describe the influence of different input variables on output variables, and the abnormality level of the sound field is evaluated based on fuzzy reasoning; Fuzzy reasoning is performed based on fuzzy rules to divide the abnormal level of the sound field into three levels: low, medium and high.

9. The adaptive sound field optimization control system based on reinforcement learning according to claim 8, characterized in that: The feedback execution module adjusts the target sound field according to whether the sound field is tilted and the level of sound field abnormality: If the sound field is tilted and the sound field abnormality level is high, the target sound field is adjusted, the sound energy density is injected into the low sound energy density area, and the actual propagation direction of the sound field energy is adjusted to the preset propagation direction; If the sound field is not tilted and the sound field abnormality level is high, the target sound field is adjusted to inject sound energy density into the low sound energy density area; If the sound field does not tilt and the sound field abnormality level is medium or low, no adjustment is made to the target sound field and the sound field state is stable; If the sound field is tilted and the sound field abnormality level is medium or low, the target sound field is adjusted and the actual propagation direction of the sound field energy is adjusted to the preset propagation direction.

Images