Multilevel active wave interference noise reduction optimization method and system for expressway
By combining a dual-stream Transformer encoder with a secretary vulture optimization algorithm, multi-level active wave interference noise reduction is achieved in highway environments, solving the problem of poor noise reduction effect in existing technologies and improving the adaptive ability and noise reduction effect of the noise reduction system.
Patent Information
- Application Number
- CN202511008636.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-22
- Publication Date
- 2025-10-03
AI Technical Summary
Existing active noise reduction systems find it difficult to achieve deep feature fusion and dynamic adaptive adjustment of complex spatial-temporal sound fields in highway environments, resulting in poor noise reduction effects. Traditional passive noise reduction methods also have limited noise reduction effects in changing and complex environments.
A dual-stream Transformer encoder is used to perform deep modeling of spatial-temporal sound field characteristics. Combined with the secretary vulture optimization algorithm, acoustic data is collected and processed in real time to generate active wave interference control signals. Multi-level, wide-band noise reduction control is achieved through closed-loop optimization.
It achieves precise response and dynamic adaptation to highway traffic noise, improves noise reduction performance, reduces noise levels, and ensures environmental comfort and stability.
Smart Images
Figure CN120748359A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of road traffic environment noise control, and in particular to a multi-level active wave interference noise reduction optimization method and system for highways. Background Art
[0002] In today's highway traffic environment, traffic noise caused by high-speed vehicles and dense traffic is becoming an increasingly prominent problem. Traditional noise reduction methods, mainly passive sound barriers and sound-absorbing materials, have reduced noise transmission to some extent, but their effectiveness is limited in the complex and changing highway environment, and they struggle to address the wide-band and multi-layered noise characteristics.
[0003] In recent years, active noise control technology has become a research hotspot. By deploying acoustic sensors and active noise reduction devices in target areas, antiphase sound waves are generated in real time for interference, enabling more flexible adaptation to different noise scenarios. However, existing active noise reduction systems often use a single time-series modeling or spatial point-to-point control strategy, lacking the ability to deeply integrate features and dynamically adapt to complex spatial-temporal sound fields, making it difficult to achieve precise active optimization of the global noise environment on highways. Active noise reduction parameters often rely on empirical settings or simple optimization algorithms, making it impossible to achieve global optimal adjustment of parameters based on real-time feedback. They are prone to local optimality or insufficient adaptability, resulting in large fluctuations in noise reduction effects and slow system response.
[0004] Therefore, how to provide a multi-level active wave interference noise reduction optimization method and system for highways is a problem that technical personnel in this field urgently need to solve. Summary of the Invention
[0005] One objective of the present invention is to propose a multi-level active wave interference noise reduction optimization method and system for highways. This method integrates deep modeling of spatial and temporal multidimensional sound field features, a dual-stream Transformer encoder, and an improved secretary vulture optimization algorithm. The method describes in detail the technical process for dynamically adaptively optimizing active wave interference noise reduction parameters using intelligent optimization and a deep fusion algorithm. By collecting and processing complex highway sound field data in real time, the method efficiently encodes and fuses temporal and spatial features, and combines noise reduction effect feedback to achieve closed-loop parameter optimization. The method offers excellent noise reduction performance, wide frequency band coverage, strong adaptability, and high control precision.
[0006] The multi-level active wave interference noise reduction optimization method for highways according to an embodiment of the present invention includes the following steps:
[0007] Multiple groups of acoustic sensors are deployed along the highway to collect spatially distributed acoustic signals and time-series acoustic data in real time. The acoustic signals and time-series acoustic data are pre-processed to form a spatial-temporal sound field dataset.
[0008] The spatial-temporal sound field dataset is input into a dual-stream Transformer encoder respectively. The first-stream Transformer encoder encodes the temporal features in the spatial-temporal sound field dataset, and the second-stream Transformer encoder encodes the spatial features in the spatial-temporal sound field dataset. The encoding results of the two streams are fused to generate a joint sound field feature vector.
[0009] Based on the joint sound field eigenvector, active wave interference control signals are calculated and generated, and distributed to active wave interference devices deployed along the highway to drive the active wave interference devices to perform noise reduction control operations;
[0010] The noise reduction effect feedback data after the active wave interferometer is collected. Using the secretary vulture optimization algorithm, the dual-stream Transformer encoder structure parameters and the active wave interferometer control signal parameters are globally optimized based on the noise reduction effect feedback data.
[0011] The optimized dual-stream Transformer encoder and active wave interference control signal are applied to the noise reduction process, continuously optimizing the multi-level and wide-band active noise reduction performance of highways in a closed loop.
[0012] Optionally, the acoustic signals and time-series acoustic data specifically include noise sound pressure level, spectrum distribution, instantaneous amplitude, phase information and real-time sound field data related to traffic flow and environmental parameters collected synchronously at different locations along the highway.
[0013] Optionally, the preprocessing of the acoustic signal and time series acoustic data specifically includes data denoising, normalization, time synchronization, feature extraction and missing data filling.
[0014] Optionally, generating a joint sound field feature vector includes:
[0015] Receive a spatial-temporal sound field dataset, extract a temporal feature set and a spatial feature set, and use them as temporal stream input and spatial stream input respectively;
[0016] For the acoustic signal feature data collected over time at each set of sampling points in the time series input, time series subsequences of different time spans are sequentially intercepted from the time series according to multiple preset sliding windows of different time lengths, with each window sliding a fixed step size. For the spatial feature data collected by acoustic sensors at different locations along the highway at each moment in the spatial stream input, all spatial features within the corresponding spatial radius from each sampling point are selected according to multiple preset spatial neighborhood ranges, forming spatial sub-region feature sets of different spatial scales.
[0017] All generated temporal subsequences are input into the first-stream Transformer encoder, and multi-scale parallel feature encoding is performed on each temporal subsequence respectively; all spatial sub-region feature sets are input into the second-stream Transformer encoder, and multi-scale parallel feature encoding is performed on each spatial sub-region feature set respectively;
[0018] In the first-stream Transformer encoder, multi-layer self-attention encoding is performed on each temporal subsequence in sequence according to the time window to obtain deep temporal feature outputs at different scales. In the second-stream Transformer encoder, multi-layer self-attention encoding is performed on the feature set of each spatial subregion to extract spatial feature outputs at different spatial scales.
[0019] In each encoding layer structure, the temporal feature outputs of the current layer and the spatial feature outputs are input into the cross-stream interactive self-attention mechanism within the layer, and the dynamic interaction and information fusion of temporal features and spatial features are achieved through the cross-stream self-attention head;
[0020] The temporal feature outputs obtained by all coding layers at each time scale and the spatial feature outputs obtained by all coding layers at each spatial scale are input into the fusion layer of the dual-stream Transformer encoder. The fusion layer is an integral part of the dual-stream Transformer encoder structure. The fusion layer adopts a dynamic weight adaptation mechanism to assign fusion weights to the temporal feature outputs of each time scale and the spatial feature outputs of each spatial scale, thereby achieving adaptive adjustment of the importance of the temporal feature outputs of all time scales and the spatial feature outputs of all spatial scales.
[0021] The fusion layer performs weighted fusion on the temporal feature outputs of each time scale and the spatial feature outputs of each spatial scale according to the assigned fusion weights of the temporal feature outputs of each time scale and the fusion weights of the spatial feature outputs of each spatial scale, and outputs a joint sound field feature vector. The joint sound field feature vector comprehensively expresses the characteristics of space-time coupling noise, and serves as the feature basis for the calculation and generation of multi-point active wave interference control signals.
[0022] Optionally, the step of calculating and generating a multi-point active wave interference control signal based on the joint sound field eigenvector, distributing the signal to active wave interference devices arranged along the highway, and driving the active wave interference devices to perform noise reduction control operations includes:
[0023] receiving the output joint sound field feature vector, associating the position coordinates of each active wave interference device in the highway coordinate system with the joint sound field feature vector, extracting feature information corresponding to the spatial position of each device, and forming a control input data set indexed by the device position;
[0024] For each active wave interference device, the control input data set indexed by the device position is combined with the global spatial and temporal features in the joint sound field feature vector. The control feature vectors of all devices are arranged in rows according to the position order of the devices in the highway coordinate system to form a spatial-temporal coupling control matrix, in which each row corresponds to the combined expression of the spatial position, temporal features, and spatial features of an active wave interference device.
[0025] For the space-time coupling control matrix, each row of control eigenvectors is used as the query vector, and all control eigenvectors in the space-time coupling control matrix are used as keys and values. The similarity between the query vector and all key vectors is calculated using the Euclidean distance. The similarity result is processed by the normalization function, and the importance weight of each active wave interference device is obtained using the weighted normalization method. The attention weight of all active wave interference devices in the current sound field environment is output, and the sum of the importance weights of each device is equal to 1.
[0026] Using the calculated importance weights, a weighted summation of the space-time coupling control matrix is performed to obtain a final control parameter set for each active wave interferometer. The control parameter set includes the phase, amplitude, and delay parameters of the anti-phase waveform corresponding to each active wave interferometer.
[0027] generating active wave interference control signals for each active wave interference device according to the obtained control parameter set;
[0028] The generated active wave interference control signal is distributed to each active wave interference device arranged along the highway, driving each active wave interference device to synchronously output anti-phase sound waves, completing multi-point collaborative active wave interference noise reduction control.
[0029] Optionally, the step of collecting noise reduction effect feedback data after the active wave interferometer is used, and using a secretary vulture optimization algorithm to globally optimize the dual-stream Transformer encoder structure parameters and the active wave interferometer control signal parameters according to the noise reduction effect feedback data includes:
[0030] At each noise measurement point along the highway, real-time noise reduction effect feedback data after the active wave interferometer outputs the anti-phase sound wave is collected to form a data set containing noise sound pressure level, noise reduction decibel value, and sound pressure balance index. The noise reduction effect feedback data includes the noise sound pressure level, noise reduction decibel value, and sound pressure balance index collected in real time at different measurement points along the highway;
[0031] The dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters are integrated into the parameter vector to be optimized, the value range of the parameter vector to be optimized is set, and the secretary vulture optimization algorithm population is randomly initialized.
[0032] For each parameter vector to be optimized, based on the noise reduction effect feedback data, the corresponding noise reduction decibel improvement and sound pressure balance index are counted and extracted. The noise reduction decibel improvement and sound pressure balance index are linearly combined according to the preset weighting coefficient, and the noise reduction performance index corresponding to the parameter vector to be optimized is calculated. Based on the noise reduction performance index, the fitness function value of the parameter vector to be optimized is calculated;
[0033] The fitness function values of all the parameter vectors to be optimized are collected. Based on the changing trends of the fitness distribution and noise reduction performance indicators, a snake-like exploration strategy or a hawk-like exploitation strategy is dynamically assigned to each parameter vector to be optimized. Through real-time analysis of fitness fluctuations and convergence status, the usage ratio and switching threshold of snake-like exploration and hawk-like exploitation in the search process are adjusted.
[0034] For each parameter vector to be optimized, the step size and perturbation direction are adaptively adjusted according to the distance from the current global optimal parameter group and the historical evolution memory parameter group, and the parameter vector to be optimized is updated;
[0035] After completing each round of updating the parameter vectors to be optimized, all new parameter vectors to be optimized are used in turn to configure the dual-stream Transformer encoder structural parameters and the active wave interference control signal parameters. Based on the noise reduction effect feedback data actually collected under each set of parameter vector configurations, the noise reduction performance index corresponding to each parameter vector is calculated, and the fitness function values of all parameter vectors to be optimized are re-evaluated.
[0036] An evolutionary memory unit is set up to record the optimal parameter vector to be optimized and the fitness function value in each generation of iteration. In each round of optimization iteration, the historical optimal parameter vector to be optimized stored in the evolutionary memory unit is used to guide the update direction of all parameter vectors to be optimized. After each round of iteration, it is determined whether the maximum number of iterations or the fitness function convergence condition has been reached.
[0037] When the termination condition is met, the optimal parameter vector recorded in the evolutionary memory unit is output and used as the final configuration of the dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters, achieving parameter adaptive optimization and continuous improvement of noise reduction performance.
[0038] Optionally, the optimized dual-stream Transformer encoder and active wave interference control signal are applied to the noise reduction process to continuously optimize the multi-level and wide-band active noise reduction performance of the highway in a closed loop, including:
[0039] The optimized dual-stream Transformer encoder structure parameters and active wave interference control signal parameters are configured to the corresponding modules in the noise reduction process;
[0040] Based on the configuration parameters, the system performs multi-point and multi-level active noise reduction control operations on the acoustic sensors and active wave interferometers deployed along the highway;
[0041] During the noise reduction control operation, the noise reduction effect feedback data of each monitoring point is collected in real time. The feedback data includes noise sound pressure level, noise reduction decibel value and sound pressure balance index;
[0042] Based on the collected noise reduction effect feedback data, the dual-stream Transformer encoder structure parameters and active wave interference control signal parameters are dynamically adjusted and adaptively optimized;
[0043] The adjusted and optimized parameters will continue to be applied to the noise reduction process, forming a continuous closed-loop process of parameter optimization, noise reduction control and performance feedback, and achieving continuous improvement in the multi-level and wide-band active noise reduction performance of highways.
[0044] The multi-level active wave interference noise reduction optimization system for highways according to an embodiment of the present invention includes the following modules:
[0045] The acoustic data acquisition and preprocessing module is used to deploy acoustic sensors along the highway to collect spatially distributed acoustic signals and time-series acoustic data in real time, complete data preprocessing, and generate a spatial-temporal sound field dataset;
[0046] The dual-stream Transformer encoder module receives a spatial-temporal sound field dataset, extracts and encodes temporal and spatial features respectively, fuses the encoded results, and outputs a joint sound field feature vector.
[0047] An active wave interference control signal generation module is used to calculate the active wave interference control signal based on the joint sound field eigenvector and distribute it to the active wave interference device to drive the device to perform noise reduction control;
[0048] The noise reduction effect feedback collection module is used to collect the noise reduction effect feedback data after the active wave interference device acts;
[0049] The secretary vulture optimization algorithm module is used to optimize the dual-stream Transformer encoder structure parameters and active wave interferometer control signal parameters using the secretary vulture optimization algorithm based on the noise reduction effect feedback data;
[0050] The closed-loop optimization execution module is used to continuously apply the optimized parameters to achieve closed-loop optimization of multi-level, wide-band active noise reduction.
[0051] The beneficial effects of the present invention are:
[0052] The multi-level active wave interference noise reduction optimization method for highways proposed in the present invention significantly improves the overall performance of the active noise reduction system in actual complex traffic environments. Compared with the existing technology, the present invention achieves a deep fusion of spatial and temporal characteristics, so that the noise reduction control signal can accurately respond to the dynamic changes and spatial distribution of the noise source, breaking through the bottleneck of the traditional method's lack of adaptability to multi-level, wide-band noise fields. Using an improved secretary vulture optimization algorithm, the noise reduction system parameters are globally adaptively optimized based on real-time noise reduction effect feedback, effectively avoiding the parameters from falling into local optimality, and achieving continuous improvement in noise reduction performance and stable output. While ensuring the high efficiency of active noise reduction, the present invention takes into account the challenges of wide bandwidth, complex noise fields, and variable parameters in actual environments, thereby improving the intelligence, precision, and dynamic adaptability of highway traffic noise control. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0054] Figure 1 This is a flow chart of the multi-level active wave interference noise reduction optimization method for highways proposed by the present invention;
[0055] Figure 2 This is a schematic diagram of the structure of the multi-level active wave interference noise reduction optimization system for highways proposed by the present invention. DETAILED DESCRIPTION
[0056] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.
[0057] refer to Figure 1 The multi-level active wave interference noise reduction optimization method for highways includes the following steps:
[0058] Multiple groups of acoustic sensors are deployed along the highway to collect spatially distributed acoustic signals and time-series acoustic data in real time. The acoustic signals and time-series acoustic data are pre-processed to form a spatial-temporal sound field dataset.
[0059] The spatial-temporal sound field dataset is input into a dual-stream Transformer encoder respectively. The first-stream Transformer encoder encodes the temporal features in the spatial-temporal sound field dataset, and the second-stream Transformer encoder encodes the spatial features in the spatial-temporal sound field dataset. The encoding results of the two streams are fused to generate a joint sound field feature vector.
[0060] Based on the joint sound field eigenvector, active wave interference control signals are calculated and generated, and distributed to active wave interference devices deployed along the highway to drive the active wave interference devices to perform noise reduction control operations;
[0061] The noise reduction effect feedback data after the active wave interferometer is collected. Using the secretary vulture optimization algorithm, the dual-stream Transformer encoder structure parameters and the active wave interferometer control signal parameters are globally optimized based on the noise reduction effect feedback data.
[0062] The optimized dual-stream Transformer encoder and active wave interference control signal are applied to the noise reduction process, continuously optimizing the multi-level and wide-band active noise reduction performance of highways in a closed loop.
[0063] In this embodiment, the acoustic signal and time-series acoustic data specifically include noise sound pressure level, spectrum distribution, instantaneous amplitude, phase information and real-time sound field data related to traffic flow and environmental parameters collected synchronously at different locations along the highway.
[0064] In this embodiment, the preprocessing of the acoustic signal and the time series acoustic data specifically includes data denoising, normalization, time synchronization, feature extraction and missing data filling.
[0065] In this embodiment, generating a joint sound field feature vector includes:
[0066] Receive a spatial-temporal sound field dataset, extract a temporal feature set and a spatial feature set, and use them as temporal stream input and spatial stream input respectively;
[0067] For the acoustic signal feature data collected over time at each set of sampling points in the time series input, time series subsequences of different time spans are sequentially intercepted from the time series according to multiple preset sliding windows of different time lengths, with each window sliding a fixed step size. For the spatial feature data collected by acoustic sensors at different locations along the highway at each moment in the spatial stream input, all spatial features within the corresponding spatial radius from each sampling point are selected according to multiple preset spatial neighborhood ranges, forming spatial sub-region feature sets of different spatial scales.
[0068] All generated temporal subsequences are input into the first-stream Transformer encoder, and multi-scale parallel feature encoding is performed on each temporal subsequence respectively; all spatial sub-region feature sets are input into the second-stream Transformer encoder, and multi-scale parallel feature encoding is performed on each spatial sub-region feature set respectively;
[0069] In the first-stream Transformer encoder, multi-layer self-attention encoding is performed on each temporal subsequence in sequence according to the time window to obtain deep temporal feature outputs at different scales. In the second-stream Transformer encoder, multi-layer self-attention encoding is performed on the feature set of each spatial subregion to extract spatial feature outputs at different spatial scales.
[0070] In each encoding layer structure, the temporal feature outputs of the current layer and the spatial feature outputs are input into the cross-stream interactive self-attention mechanism within the layer, and the dynamic interaction and information fusion of temporal features and spatial features are achieved through the cross-stream self-attention head;
[0071] The temporal feature outputs obtained by all coding layers at each time scale and the spatial feature outputs obtained by all coding layers at each spatial scale are input into the fusion layer of the dual-stream Transformer encoder. The fusion layer is an integral part of the dual-stream Transformer encoder structure. The fusion layer adopts a dynamic weight adaptation mechanism to assign fusion weights to the temporal feature outputs of each time scale and the spatial feature outputs of each spatial scale, thereby achieving adaptive adjustment of the importance of the temporal feature outputs of all time scales and the spatial feature outputs of all spatial scales.
[0072] The fusion layer performs weighted fusion on the temporal feature outputs of each time scale and the spatial feature outputs of each spatial scale according to the assigned fusion weights of the temporal feature outputs of each time scale and the fusion weights of the spatial feature outputs of each spatial scale, and outputs a joint sound field feature vector. The joint sound field feature vector comprehensively expresses the characteristics of space-time coupling noise, and serves as the feature basis for the calculation and generation of multi-point active wave interference control signals.
[0073] In this embodiment, the method of calculating and generating a multi-point active wave interference control signal based on the joint sound field eigenvector, distributing the signal to active wave interference devices arranged along the highway, and driving the active wave interference devices to perform noise reduction control operations includes:
[0074] receiving the output joint sound field feature vector, associating the position coordinates of each active wave interference device in the highway coordinate system with the joint sound field feature vector, extracting feature information corresponding to the spatial position of each device, and forming a control input data set indexed by the device position;
[0075] For each active wave interference device, the control input data set indexed by the device position is combined with the global spatial and temporal features in the joint sound field feature vector. The control feature vectors of all devices are arranged in rows according to the position order of the devices in the highway coordinate system to form a spatial-temporal coupling control matrix, in which each row corresponds to the combined expression of the spatial position, temporal features, and spatial features of an active wave interference device.
[0076] For the space-time coupling control matrix, each row of control eigenvectors is used as the query vector, and all control eigenvectors in the space-time coupling control matrix are used as keys and values. The similarity between the query vector and all key vectors is calculated using the Euclidean distance. The similarity result is processed by the normalization function, and the importance weight of each active wave interference device is obtained using the weighted normalization method. The attention weight of all active wave interference devices in the current sound field environment is output, and the sum of the importance weights of each device is equal to 1.
[0077] Using the calculated importance weights, a weighted summation of the space-time coupling control matrix is performed to obtain a final control parameter set for each active wave interferometer. The control parameter set includes the phase, amplitude, and delay parameters of the anti-phase waveform corresponding to each active wave interferometer.
[0078] According to the obtained control parameter set, the active wave interference control signal of each active wave interference device is generated:
[0079] s m (t) = A m sin(ωt+φ m )·u(t-τ m );
[0080] Among them, s m (t) is the active wave interference control signal of the mth active wave interference device at time t, A m is the amplitude of the output control signal of the mth device, φ m is the phase of the output control signal of the mth device, τ m is the delay time of the control signal output by the mth device, ω is the frequency of the anti-phase sound wave generated by the active wave interference device, u(t-τ m ) is the delay function, which represents the signal delay τ m After output, sin is the sine function;
[0081] The generated active wave interference control signal is distributed to each active wave interference device arranged along the highway, driving each active wave interference device to synchronously output anti-phase sound waves, completing multi-point collaborative active wave interference noise reduction control.
[0082] In this embodiment, the noise reduction effect feedback data after the active wave interference device is used is collected, and the secretary vulture optimization algorithm is used to globally optimize the dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters according to the noise reduction effect feedback data, including:
[0083] At each noise measurement point along the highway, real-time noise reduction effect feedback data after the active wave interferometer outputs the anti-phase sound wave is collected to form a data set containing noise sound pressure level, noise reduction decibel value, and sound pressure balance index. The noise reduction effect feedback data includes the noise sound pressure level, noise reduction decibel value, and sound pressure balance index collected in real time at different measurement points along the highway;
[0084] The dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters are integrated into the parameter vector to be optimized, the value range of the parameter vector to be optimized is set, and the secretary vulture optimization algorithm population is randomly initialized.
[0085] For each parameter vector to be optimized, based on the noise reduction effect feedback data, the corresponding noise reduction decibel improvement and sound pressure balance index are counted and extracted. The noise reduction decibel improvement and sound pressure balance index are linearly combined according to the preset weighting coefficient, and the noise reduction performance index corresponding to the parameter vector to be optimized is calculated. Based on the noise reduction performance index, the fitness function value of the parameter vector to be optimized is calculated:
[0086]
[0087] Where, ΔL dB E is the noise reduction decibel increase. eq is the sound pressure balance index, α, β are weight coefficients, is the fitness function value, represents the i-th parameter vector to be optimized;
[0088] The fitness function values of all the parameter vectors to be optimized are collected. Based on the changing trends of the fitness distribution and noise reduction performance indicators, a snake-like exploration strategy or a hawk-like exploitation strategy is dynamically assigned to each parameter vector to be optimized. Through real-time analysis of fitness fluctuations and convergence status, the usage ratio and switching threshold of snake-like exploration and hawk-like exploitation in the search process are adjusted.
[0089] For each parameter vector to be optimized, the step size and perturbation direction are adaptively adjusted according to the distance from the current global optimal parameter group and the historical evolution memory parameter group, and the parameter vector to be optimized is updated:
[0090] Snake Exploration:
[0091]
[0092] Eagle-style development:
[0093]
[0094] Among them, λ1, λ2, γ are step factors, r1, r2 are random numbers in the interval [0,1], is the current optimal parameter, is the historical optimal parameter recorded by the evolutionary memory unit, randn() is the normal distribution disturbance, is the value of the i-th parameter vector to be optimized at the t+1th iteration;
[0095] After completing each round of updating the parameter vectors to be optimized, all new parameter vectors to be optimized are used in turn to configure the dual-stream Transformer encoder structural parameters and the active wave interference control signal parameters. Based on the noise reduction effect feedback data actually collected under each set of parameter vector configurations, the noise reduction performance index corresponding to each parameter vector is calculated, and the fitness function values of all parameter vectors to be optimized are re-evaluated.
[0096] An evolutionary memory unit is set up to record the optimal parameter vector to be optimized and the fitness function value in each generation of iteration. In each round of optimization iteration, the historical optimal parameter vector to be optimized stored in the evolutionary memory unit is used to guide the update direction of all parameter vectors to be optimized. After each round of iteration, it is determined whether the maximum number of iterations or the fitness function convergence condition has been reached.
[0097] When the termination condition is met, the optimal parameter vector recorded in the evolutionary memory unit is output and used as the final configuration of the dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters, achieving parameter adaptive optimization and continuous improvement of noise reduction performance.
[0098] In this embodiment, the optimized dual-stream Transformer encoder and active wave interference control signal are applied to the noise reduction process to continuously optimize the multi-level and wide-band active noise reduction performance of highways in a closed loop, including:
[0099] The optimized dual-stream Transformer encoder structure parameters and active wave interference control signal parameters are configured to the corresponding modules in the noise reduction process;
[0100] Based on the configuration parameters, the system performs multi-point and multi-level active noise reduction control operations on the acoustic sensors and active wave interferometers deployed along the highway;
[0101] During the noise reduction control operation, the noise reduction effect feedback data of each monitoring point is collected in real time. The feedback data includes noise sound pressure level, noise reduction decibel value and sound pressure balance index;
[0102] Based on the collected noise reduction effect feedback data, the dual-stream Transformer encoder structure parameters and active wave interference control signal parameters are dynamically adjusted and adaptively optimized;
[0103] The adjusted and optimized parameters will continue to be applied to the noise reduction process, forming a continuous closed-loop process of parameter optimization, noise reduction control and performance feedback, and achieving continuous improvement in the multi-level and wide-band active noise reduction performance of highways.
[0104] refer to Figure 2 , a multi-level active wave interference noise reduction optimization system for highways, including the following modules:
[0105] The acoustic data acquisition and preprocessing module is used to deploy acoustic sensors along the highway to collect spatially distributed acoustic signals and time-series acoustic data in real time, complete data preprocessing, and generate a spatial-temporal sound field dataset;
[0106] The dual-stream Transformer encoder module receives a spatial-temporal sound field dataset, extracts and encodes temporal and spatial features respectively, fuses the encoded results, and outputs a joint sound field feature vector.
[0107] An active wave interference control signal generation module is used to calculate the active wave interference control signal based on the joint sound field eigenvector and distribute it to the active wave interference device to drive the device to perform noise reduction control;
[0108] The noise reduction effect feedback collection module is used to collect the noise reduction effect feedback data after the active wave interference device acts;
[0109] The secretary vulture optimization algorithm module is used to optimize the dual-stream Transformer encoder structure parameters and active wave interferometer control signal parameters using the secretary vulture optimization algorithm based on the noise reduction effect feedback data;
[0110] The closed-loop optimization execution module is used to continuously apply the optimized parameters to achieve closed-loop optimization of multi-level, wide-band active noise reduction.
[0111] Example 1:
[0112] To verify the feasibility of the present invention, it was applied to a multi-lane, densely trafficked highway section near residential areas, hospitals, and schools. For a long time, the average roadside noise level during peak hours on this section has been as high as 77.5dB(A), exceeding national environmental standards and significantly impacting the lives of residents and the work of surrounding businesses. Traditional concrete sound barriers have limited noise reduction effectiveness. Their noise suppression capabilities decrease with frequent heavy vehicle traffic and wind direction changes, and they are less adaptable to broadband noise and multi-point noise fields.
[0113] The system of the present invention deploys a group of high-sensitivity acoustic sensors and active wave interference devices at intervals of 100 meters on this road section, totaling 30 groups. The sensors collect acoustic data including A-weighted sound pressure level, time-frequency characteristics, etc. in real time, and the active wave interference device has a wide-band response (20Hz~8kHz) and fast adjustment capabilities. After all the data are pre-processed by the front end, they are uniformly uploaded to the roadside central control unit. The dual-stream Transformer encoder extracts spatial features and temporal features respectively, and then outputs the joint sound field feature vector through adaptive fusion. Based on real-time sound field feedback, the system uses the peregrine falcon optimization algorithm to perform global dynamic optimization of the encoder structure parameters and the control parameters of each active wave interference device, so that the phase, amplitude and delay of the noise reduction signal are continuously adaptively optimized to form a closed-loop noise reduction control process.
[0114] During the continuous testing period from May 10 to June 10, the system covered a variety of working conditions, including weekdays, holidays, and extreme weather conditions. Before all noise reduction devices were activated, the average noise level at 30 typical monitoring points was 77.5dB(A), with a maximum of 82.4dB(A). The average noise reduction was only 4.9dB(A), and the noise distribution was extremely uneven. Some points near entrances and exits and bends had a maximum value of up to 84dB(A), and the night-time fluctuation was as high as 6.1dB. After adopting the method of the present invention, the average value of the measuring points dropped to 67.4dB(A), the maximum value dropped to 72.9dB(A), the minimum value dropped to 63.2dB(A), and the average noise reduction amplitude reached 10.1dB(A). The average response delay of the system was less than 9ms within 10 days, and the standard deviation of the noise between monitoring points was reduced from 6.1dB to 2.7dB after the system's adaptive adjustment. The noise reduction effect is stable during daytime peaks, nighttime troughs and extreme heavy vehicle traffic conditions, with noise reduction fluctuations less than 1.5dB, far superior to traditional sound barrier solutions.
[0115] Table 1 Comparative data of typical operation of multi-point active noise reduction system on highway
[0116]
[0117]
[0118] As can be seen from Table 1, the active wave interference noise reduction optimization method described in the present invention has significant noise reduction effects and technical advantages in the actual multi-point environment of highways. During the continuous field measurement process from May to June 2025, the average environmental noise level at each monitoring point before noise reduction was generally between 76.4 and 79.3 dB (A), which is much higher than the recommended standards for environmental protection and resident health. After using traditional sound barriers for noise reduction, the average noise level decreased to a limited extent, and most monitoring points were still between 71.7 and 73.8 dB (A) after noise reduction, indicating that traditional methods are insufficient in their ability to control complex noise fields and have poor adaptability to broadband and multi-source noise.
[0119] After the active noise reduction system of the present invention was activated, the noise mean values at all monitoring points dropped significantly. After noise reduction, the noise mean values were generally lower than 68.7dB(A), the optimal point reached 66.3dB(A), and the noise reduction range reached 9.8 to 10.6dB(A). This noise reduction range not only far exceeds that of traditional sound barriers (with an improvement of approximately 3.5 to 4dB(A)), but also performs stably at multiple key nodes and different time periods. The noise reduction fluctuation of the active noise reduction system is only 1.0 to 1.5dB, which is significantly better than traditional methods and ensures the balance of environmental noise. The range between the maximum and minimum noise values is also significantly narrowed, and the distribution of environmental noise is smoother, reflecting the system's ability to accurately suppress spatial-temporal multidimensional noise.
[0120] The active noise reduction system has a response delay of less than 9 milliseconds, enabling rapid response and adaptive adjustment to sudden noise and traffic flow changes, ensuring that the system can maintain a high level of noise reduction performance under complex working conditions and changeable weather conditions, and improving the comfort of the environment along the highway and residents' satisfaction.
[0121] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.
Claims
1. A multi-level active wave interference noise reduction optimization method for highways, characterized by: include: Multiple groups of acoustic sensors are deployed along the highway to collect spatially distributed acoustic signals and time-series acoustic data in real time. The acoustic signals and time-series acoustic data are pre-processed to form a spatial-temporal sound field dataset. The spatial-temporal sound field dataset is input into a dual-stream Transformer encoder respectively. The first-stream Transformer encoder encodes the temporal features in the spatial-temporal sound field dataset, and the second-stream Transformer encoder encodes the spatial features in the spatial-temporal sound field dataset. The encoding results of the two streams are fused to generate a joint sound field feature vector. Based on the joint sound field eigenvector, active wave interference control signals are calculated and generated, and distributed to active wave interference devices deployed along the highway to drive the active wave interference devices to perform noise reduction control operations; The noise reduction effect feedback data after the active wave interferometer is collected. Using the secretary vulture optimization algorithm, the dual-stream Transformer encoder structure parameters and the active wave interferometer control signal parameters are globally optimized based on the noise reduction effect feedback data. The optimized dual-stream Transformer encoder and active wave interference control signal are applied to the noise reduction process, continuously optimizing the multi-level and wide-band active noise reduction performance of highways in a closed loop.
2. The multi-level active wave interference noise reduction optimization method for highways according to claim 1 is characterized in that: The acoustic signals and time-series acoustic data specifically include noise sound pressure level, spectrum distribution, instantaneous amplitude, phase information, and real-time sound field data related to traffic flow and environmental parameters collected synchronously at different locations along the highway.
3. The multi-level active wave interference noise reduction optimization method for highways according to claim 1 is characterized in that: The preprocessing of acoustic signals and time series acoustic data specifically includes data denoising, normalization, time synchronization, feature extraction and missing data filling.
4. The multi-level active wave interference noise reduction optimization method for highways according to claim 1 is characterized in that: The generating of the joint sound field feature vector includes: Receive a spatial-temporal sound field dataset, extract a temporal feature set and a spatial feature set, and use them as temporal stream input and spatial stream input respectively; For the acoustic signal feature data collected over time at each set of sampling points in the time series input, time series subsequences of different time spans are sequentially intercepted from the time series according to multiple preset sliding windows of different time lengths, with each window sliding a fixed step size. For the spatial feature data collected by acoustic sensors at different locations along the highway at each moment in the spatial stream input, all spatial features within the corresponding spatial radius from each sampling point are selected according to multiple preset spatial neighborhood ranges, forming spatial sub-region feature sets of different spatial scales. All generated temporal subsequences are input into the first-stream Transformer encoder, and multi-scale parallel feature encoding is performed on each temporal subsequence respectively; all spatial sub-region feature sets are input into the second-stream Transformer encoder, and multi-scale parallel feature encoding is performed on each spatial sub-region feature set respectively; In the first-stream Transformer encoder, multi-layer self-attention encoding is performed on each temporal subsequence in sequence according to the time window to obtain deep temporal feature outputs at different scales. In the second-stream Transformer encoder, multi-layer self-attention encoding is performed on the feature set of each spatial subregion to extract spatial feature outputs at different spatial scales. In each encoding layer structure, the temporal feature outputs of the current layer and the spatial feature outputs are input into the cross-stream interactive self-attention mechanism within the layer, and the dynamic interaction and information fusion of temporal features and spatial features are achieved through the cross-stream self-attention head; The temporal feature outputs obtained by all coding layers at each time scale and the spatial feature outputs obtained by all coding layers at each spatial scale are input into the fusion layer of the dual-stream Transformer encoder. The fusion layer is an integral part of the dual-stream Transformer encoder structure. The fusion layer adopts a dynamic weight adaptation mechanism to assign fusion weights to the temporal feature outputs of each time scale and the spatial feature outputs of each spatial scale, thereby achieving adaptive adjustment of the importance of the temporal feature outputs of all time scales and the spatial feature outputs of all spatial scales. The fusion layer performs weighted fusion on the temporal feature outputs of each time scale and the spatial feature outputs of each spatial scale according to the assigned fusion weights of the temporal feature outputs of each time scale and the fusion weights of the spatial feature outputs of each spatial scale, and outputs a joint sound field feature vector. The joint sound field feature vector comprehensively expresses the characteristics of space-time coupling noise, and serves as the feature basis for the calculation and generation of multi-point active wave interference control signals.
5. The multi-level active wave interference noise reduction optimization method for highways according to claim 1 is characterized in that: The method calculates and generates a multi-point active wave interference control signal based on the joint sound field eigenvector, distributes the signal to active wave interference devices arranged along the highway, and drives the active wave interference devices to perform noise reduction control operations, including: receiving the output joint sound field feature vector, associating the position coordinates of each active wave interference device in the highway coordinate system with the joint sound field feature vector, extracting feature information corresponding to the spatial position of each device, and forming a control input data set indexed by the device position; For each active wave interference device, the control input data set indexed by the device position is combined with the global spatial and temporal features in the joint sound field feature vector. The control feature vectors of all devices are arranged in rows according to the position order of the devices in the highway coordinate system to form a spatial-temporal coupling control matrix, in which each row corresponds to the combined expression of the spatial position, temporal features, and spatial features of an active wave interference device. For the space-time coupling control matrix, each row of control eigenvectors is used as the query vector, and all control eigenvectors in the space-time coupling control matrix are used as keys and values. The similarity between the query vector and all key vectors is calculated using the Euclidean distance. The similarity result is processed by the normalization function, and the importance weight of each active wave interference device is obtained using the weighted normalization method. The attention weight of all active wave interference devices in the current sound field environment is output, and the sum of the importance weights of each device is equal to 1. Using the calculated importance weights, a weighted summation of the space-time coupling control matrix is performed to obtain a final control parameter set for each active wave interferometer. The control parameter set includes the phase, amplitude, and delay parameters of the anti-phase waveform corresponding to each active wave interferometer. generating active wave interference control signals for each active wave interference device according to the obtained control parameter set; The generated active wave interference control signal is distributed to each active wave interference device arranged along the highway, driving each active wave interference device to synchronously output anti-phase sound waves, completing multi-point collaborative active wave interference noise reduction control.
6. The multi-level active wave interference noise reduction optimization method for highways according to claim 1 is characterized in that: The method collects noise reduction effect feedback data after the active wave interference device acts, and uses the secretary vulture optimization algorithm to globally optimize the dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters according to the noise reduction effect feedback data, including: At each noise measurement point along the highway, real-time noise reduction effect feedback data after the active wave interferometer outputs the anti-phase sound wave is collected to form a data set containing noise sound pressure level, noise reduction decibel value, and sound pressure balance index. The noise reduction effect feedback data includes the noise sound pressure level, noise reduction decibel value, and sound pressure balance index collected in real time at different measurement points along the highway; The dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters are integrated into the parameter vector to be optimized, the value range of the parameter vector to be optimized is set, and the secretary vulture optimization algorithm population is randomly initialized. For each parameter vector to be optimized, based on the noise reduction effect feedback data, the corresponding noise reduction decibel improvement and sound pressure balance index are counted and extracted. The noise reduction decibel improvement and sound pressure balance index are linearly combined according to the preset weighting coefficient, and the noise reduction performance index corresponding to the parameter vector to be optimized is calculated. Based on the noise reduction performance index, the fitness function value of the parameter vector to be optimized is calculated; The fitness function values of all the parameter vectors to be optimized are collected. Based on the changing trends of the fitness distribution and noise reduction performance indicators, a snake-like exploration strategy or a hawk-like exploitation strategy is dynamically assigned to each parameter vector to be optimized. Through real-time analysis of fitness fluctuations and convergence status, the usage ratio and switching threshold of snake-like exploration and hawk-like exploitation in the search process are adjusted. For each parameter vector to be optimized, the step size and perturbation direction are adaptively adjusted according to the distance from the current global optimal parameter group and the historical evolution memory parameter group, and the parameter vector to be optimized is updated; After completing each round of updating the parameter vectors to be optimized, all new parameter vectors to be optimized are used in turn to configure the dual-stream Transformer encoder structural parameters and the active wave interference control signal parameters. Based on the noise reduction effect feedback data actually collected under each set of parameter vector configurations, the noise reduction performance index corresponding to each parameter vector is calculated, and the fitness function values of all parameter vectors to be optimized are re-evaluated. An evolutionary memory unit is set up to record the optimal parameter vector to be optimized and the fitness function value in each generation of iteration. In each round of optimization iteration, the historical optimal parameter vector to be optimized stored in the evolutionary memory unit is used to guide the update direction of all parameter vectors to be optimized. After each round of iteration, it is determined whether the maximum number of iterations or the fitness function convergence condition has been reached. When the termination condition is met, the optimal parameter vector recorded in the evolutionary memory unit is output and used as the final configuration of the dual-stream Transformer encoder structure parameters and the active wave interference control signal parameters, achieving parameter adaptive optimization and continuous improvement of noise reduction performance.
7. The multi-level active wave interference noise reduction optimization method for highways according to claim 1 is characterized in that: The optimized dual-stream Transformer encoder and active wave interferometer control signal are applied to the noise reduction process to continuously optimize the multi-level and wide-band active noise reduction performance of highways in a closed loop, including: The optimized dual-stream Transformer encoder structure parameters and active wave interference control signal parameters are configured to the corresponding modules in the noise reduction process; Based on the configuration parameters, the system performs multi-point and multi-level active noise reduction control operations on the acoustic sensors and active wave interferometers deployed along the highway; During the noise reduction control operation, the noise reduction effect feedback data of each monitoring point is collected in real time. The feedback data includes noise sound pressure level, noise reduction decibel value and sound pressure balance index; Based on the collected noise reduction effect feedback data, the dual-stream Transformer encoder structure parameters and active wave interference control signal parameters are dynamically adjusted and adaptively optimized; The adjusted and optimized parameters will continue to be applied to the noise reduction process, forming a continuous closed-loop process of parameter optimization, noise reduction control and performance feedback, and achieving continuous improvement in the multi-level and wide-band active noise reduction performance of highways.
8. A multi-level active wave interference noise reduction optimization system for highways, which implements the multi-level active wave interference noise reduction optimization method for highways according to any one of claims 1 to 7, characterized in that: Includes the following modules: The acoustic data acquisition and preprocessing module is used to deploy acoustic sensors along the highway to collect spatially distributed acoustic signals and time-series acoustic data in real time, complete data preprocessing, and generate a spatial-temporal sound field dataset; The dual-stream Transformer encoder module receives a spatial-temporal sound field dataset, extracts and encodes temporal and spatial features respectively, fuses the encoded results, and outputs a joint sound field feature vector. An active wave interference control signal generation module is used to calculate the active wave interference control signal based on the joint sound field eigenvector and distribute it to the active wave interference device to drive the device to perform noise reduction control; The noise reduction effect feedback collection module is used to collect the noise reduction effect feedback data after the active wave interference device acts; The secretary vulture optimization algorithm module is used to optimize the dual-stream Transformer encoder structure parameters and active wave interferometer control signal parameters using the secretary vulture optimization algorithm based on the noise reduction effect feedback data; The closed-loop optimization execution module is used to continuously apply the optimized parameters to achieve closed-loop optimization of multi-level, wide-band active noise reduction.