Tunnel construction environmental noise suppression method and system using deep learning algorithm

Abstract

The embodiments of the present application disclose a method and system for suppressing noise in a tunnel construction environment using a deep learning algorithm. The method fully acquires the complex and diverse noise information in a tunnel construction environment by collecting a set of original noise signals covering multiple time-domain noise signals and the noise components of construction equipment contained therein. The noise feature distribution obtained by noise feature extraction and processing can then be used to comprehensively characterize the noise characteristics from multiple dimensions, including noise spectrum characteristics, noise source orientation characteristics, and noise energy propagation characteristics. The above-mentioned features are then suppressed using a deep learning noise suppression model, which can intelligently analyze and generate a targeted suppression control feature set. The noise suppression device is further driven based on the set to perform a noise cancellation operation, which can accurately generate a cancellation sound wave signal with a phase opposite to that of each time-domain noise signal.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of noise data processing, and specifically to a method and system for suppressing noise in a tunnel construction environment using a deep learning algorithm. Background Art

[0002] During tunnel construction, environmental noise has always been a critical issue that needs to be addressed. Tunnel construction involves the simultaneous operation of multiple large-scale equipment. The noise generated not only harms the health of construction workers, but also can lead to occupational diseases such as hearing loss due to long-term exposure to high-intensity noise. It also affects construction efficiency and quality, interfering with communication between construction workers and increasing the risk of operational errors.

[0003] However, existing technologies for addressing noise in tunnel construction environments have numerous shortcomings. Some traditional methods simply wrap certain equipment with soundproofing materials. This approach can only partially attenuate the noise emitted by the equipment itself, but fails to comprehensively and effectively address the complex noise levels of the entire construction environment. Other technologies, while attempting to monitor noise, lack in-depth analysis of noise characteristics and cannot accurately grasp key information such as noise sources and propagation characteristics, making it difficult to develop targeted noise reduction strategies. Summary of the Invention

[0004] The embodiments of the present application provide a method and system for suppressing tunnel construction environmental noise using a deep learning algorithm, which are used to effectively and specifically reduce the intensity of environmental noise.

[0005] In a first aspect, an embodiment of the present application provides a method for suppressing tunnel construction environmental noise using a deep learning algorithm, the method comprising: collecting a set of original noise signals from a tunnel construction site, the set of original noise signals comprising multiple time-domain noise signals, each time-domain noise signal containing a noise component generated by at least one construction equipment; performing noise feature extraction processing on the set of original noise signals to obtain a noise feature distribution for each time-domain noise signal, the noise feature distribution comprising noise spectrum features, noise source orientation features, and noise energy propagation features; calling a deep learning noise suppression model to suppress the noise feature distribution to generate a suppression control feature set corresponding to each time-domain noise signal; and driving a noise suppression device to perform a noise cancellation operation based on the suppression control feature set to generate a cancellation sound wave signal having a phase opposite to that of the time-domain noise signal to reduce the intensity of the environmental noise.

[0006] In a second aspect, an embodiment of the present application provides a tunnel construction environment noise suppression system using a deep learning algorithm, comprising:

[0007] processor;

[0008] a storage device having a computer program stored thereon,

[0009] When the computer program is executed by the processor, the processor implements any of the methods for suppressing tunnel construction environment noise using a deep learning algorithm.

[0010] An embodiment of the present application provides a readable storage medium, which stores a program or instruction. When the program or instruction is executed by a processor, the steps of the method for suppressing tunnel construction environment noise using a deep learning algorithm are implemented.

[0011] It can be seen that the embodiments of the present application have the following beneficial effects: by collecting a set of original noise signals covering multiple time-domain noise signals and the construction equipment noise components contained therein, the complex and diverse noise information in the tunnel construction environment is fully acquired; then, based on the noise feature extraction and processing, the noise feature distribution obtained can comprehensively characterize the noise characteristics from multiple dimensions such as noise spectrum characteristics, noise source orientation characteristics and noise energy propagation characteristics; then, the deep learning noise suppression model is used to suppress the above features, which can intelligently analyze and generate a targeted suppression control feature set; further, based on the set, the noise suppression device is driven to perform noise cancellation operations, which can accurately generate cancellation sound wave signals with opposite phases to each time-domain noise signal.

[0012] This design effectively and specifically reduces the intensity of ambient noise and minimizes the potential hazards caused by noise. It also improves the accuracy and efficiency of construction operations, avoids miscommunication and operational errors caused by noise interference, and overall improves the quality and safety of the tunnel construction process. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a flowchart of a method for suppressing noise in a tunnel construction environment using a deep learning algorithm provided in an embodiment of the present application.

[0014] Figure 2 This is a schematic diagram of the basic structure of a tunnel construction environment noise suppression system using a deep learning algorithm provided in an embodiment of the present application. DETAILED DESCRIPTION

[0015] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the embodiments of the present application are further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0016] See also Figure 1 As shown in FIG, this figure is a flow chart of a method for suppressing noise in a tunnel construction environment using a deep learning algorithm provided by an embodiment of the present application. The method can be implemented by a system for suppressing noise in a tunnel construction environment using a deep learning algorithm. Figure 1As shown, the method may include steps 110 to 140.

[0017] Step 110: Collecting an original noise signal set at the tunnel construction site, wherein the original noise signal set includes a plurality of time-domain noise signals, each of which includes a noise component generated by at least one construction equipment.

[0018] In an embodiment of the present application, in the scenario of tunnel construction, there are multiple construction equipment operating at the same time at the construction site. For example, when an excavator is excavating earth, the operation of its engine and the excavation action will generate noise in the corresponding frequency range; for another example, when a loader is transporting materials, the friction of its mechanical parts and the operation of its power system will also generate noise; in addition, a concrete mixer will also generate corresponding noise during the mixing process. At this time, the relevant audio acquisition equipment is used to collect the noise of the entire construction site according to the preset sampling frequency and duration settings. In the set of original noise signals collected, each time domain noise signal is mixed with noise components generated by different construction equipment. For example, one of the collected time domain noise signals may contain the low-frequency roar of the excavator, the medium-frequency mechanical friction sound of the loader, and the high-frequency stirring sound of the concrete mixer.

[0019] Step 120: performing noise feature extraction processing on the original noise signal set to obtain a noise feature distribution of each time-domain noise signal, wherein the noise feature distribution includes noise spectrum features, noise source orientation features, and noise energy propagation features.

[0020] Next, noise feature extraction is performed on the collected original noise signal set. This process aims to fully understand the characteristics of the noise and provide an accurate basis for subsequent suppression processing.

[0021] Optionally, performing noise feature extraction processing on the original noise signal set to obtain a noise feature distribution of each time-domain noise signal includes:

[0022] Step 121: performing multi-scale decomposition processing on the time domain noise signal to generate multiple noise sub-signals, each noise sub-signal corresponding to a noise component in a different frequency band.

[0023] Furthermore, the performing multi-scale decomposition processing on the time domain noise signal to generate multiple noise sub-signals includes:

[0024] Step 1210: Acquire multiple adaptive filter groups of different scales, each adaptive filter group including at least one bandpass filter, wherein the scale parameters of the adaptive filter groups are adjusted according to the average spectral density of the time-domain noise signal; input the time-domain noise signal into the multiple adaptive filter groups of different scales for parallel filtering processing to obtain initial noise sub-signals divided into multiple frequency bands; perform reconstruction error correction processing on the initial noise sub-signals in the frequency band overlapping region to eliminate the signal aliasing effect between adjacent frequency bands, thereby generating the multiple noise sub-signals.

[0025] In an embodiment of the present application, for one of the time-domain noise signals, the scale parameters of the adaptive filter bank are first determined based on its average spectral density. If the average spectral density of the time-domain noise signal indicates that its energy is primarily concentrated within a wide frequency range, an adaptive filter bank of the corresponding scale is selected. Each filter bank contains a bandpass filter with an appropriate center frequency and bandwidth. For example, there are three adaptive filter banks of different scales: A, B, and C. The bandpass filter of filter bank A has a lower center frequency and narrower bandwidth, making it suitable for capturing low-frequency noise components; filter bank B has moderate parameters for processing mid-frequency noise; and filter bank C targets high-frequency noise. The time-domain noise signal is simultaneously input into these three adaptive filter banks for parallel filtering, resulting in initial noise sub-signals corresponding to different frequency bands. However, due to the characteristics of the filters, signal aliasing may occur between adjacent frequency bands. Therefore, reconstruction error correction can be performed on the initial noise sub-signals in the overlapping frequency band regions using a commonly used algorithm (such as the Wiener filter algorithm). Errors in the overlapping regions can be determined and adjusted, ultimately resulting in multiple noise sub-signals corresponding to different frequency bands.

[0026] Step 122: Perform spectrum analysis on the multiple noise sub-signals, and extract the frequency domain energy distribution feature of each noise sub-signal as the noise spectrum feature.

[0027] In a preferred embodiment, performing spectrum analysis on the multiple noise sub-signals to extract frequency domain energy distribution characteristics of each noise sub-signal includes:

[0028] Step 1220: Perform short-time Fourier transform processing on each noise sub-signal to generate a corresponding time-frequency spectrum; perform peak detection processing on the time-frequency spectrum to identify the set of energy peak points in the time-frequency spectrum; construct an energy distribution feature set based on the distribution density, peak amplitude and frequency interval of the energy peak point set, and extract the statistical characteristics of the energy distribution feature set as the frequency domain energy distribution characteristics.

[0029] For each noise sub-signal obtained through multi-scale decomposition, a short-time Fourier transform (SFT) is performed on one of the noise sub-signals, for example. By selecting an appropriate window function and window length, the noise sub-signal is segmented in the time domain. Each segment is then Fourier transformed to generate a corresponding time-spectrogram. The energy distribution at different time and frequency points can be determined on this time-spectrogram. Next, peak detection is performed on the time-spectrogram, using the generalized cross-correlation (GCC) algorithm to identify the set of energy peaks in the graph. For example, detection reveals five energy peaks located at different frequency positions and with different amplitudes. An energy distribution feature set is then constructed based on the distribution density, peak amplitude, and frequency spacing of this energy peak set. For example, the frequency spacing between adjacent peaks is calculated, the number of peaks within a unit frequency range is counted as the distribution density, and the amplitude of each peak is recorded. Finally, statistical features such as the mean and variance are extracted from this energy distribution feature set to serve as the frequency-domain energy distribution feature of the noise sub-signal, which is a component of the noise spectrum feature of the entire time-domain noise signal.

[0030] Step 123: Perform spatial sound field modeling on the time domain noise signal to determine the sound source azimuth and sound wave propagation path characteristics of the time domain noise signal in three-dimensional space, and fuse the sound source azimuth and the sound wave propagation path characteristics into the noise source azimuth characteristics.

[0031] In another optional embodiment, performing spatial sound field modeling on the time domain noise signal to determine the sound source azimuth and sound wave propagation path characteristics of the time domain noise signal in three-dimensional space includes:

[0032] Step 1230: Based on the deployed multiple spatial microphone array nodes, collect the sound pressure signals of the time domain noise signal at different spatial positions; calculate the spatial propagation vector of the time domain noise signal based on the arrival time difference and phase difference of the sound pressure signal; perform wave arrival direction estimation processing based on the spatial propagation vector to determine the azimuth of the sound source; construct a three-dimensional sound field propagation model based on the position coordinates of the spatial microphone array nodes and the azimuth of the sound source, and extract the path attenuation coefficient and reflection path characteristics of the three-dimensional sound field propagation model as the sound wave propagation path characteristics.

[0033] At a tunnel construction site, multiple spatial microphone array nodes can be pre-deployed at different locations. When a time-domain noise signal is collected, each microphone array node receives it simultaneously. However, due to their different distances from the sound source, the signal's arrival time and phase may differ. For example, there are three microphone array nodes, M1, M2, and M3. M1 is closer to the sound source, while M2 and M3 are farther away and in different directions. M1 receives the signal first, while M2 and M3 receive it later, with a phase difference between them. Based on the arrival time and phase differences of the sound pressure signals, a multiple signal classification algorithm is used to calculate the spatial propagation vector of the time-domain noise signal. This vector represents information such as the propagation direction and speed of the sound wave in space. Based on this spatial propagation vector, the direction of arrival (DOA) estimation method is used to ultimately determine the azimuth of the sound source in three-dimensional space. For example, the sound source can be determined to be located on one side of the tunnel at a certain angle to the horizontal. Next, a three-dimensional sound field propagation model is constructed by combining the position coordinates of each spatial microphone array node and the determined azimuth of the sound source. This model considers the attenuation and reflection of sound waves during propagation, extracting path attenuation coefficients and reflection path characteristics. For example, it is found that during sound wave propagation, energy gradually attenuates due to absorption and reflection by tunnel walls, and multiple reflection paths exist. This information constitutes the sound wave propagation path characteristics. Finally, the sound source azimuth angle is combined with the sound wave propagation path characteristics to form the noise source azimuth characteristics of the time-domain noise signal.

[0034] Step 124: Perform energy attenuation analysis on the time domain noise signal, calculate the energy attenuation curve of the time domain noise signal according to the properties of the sound wave propagation medium and the propagation distance, and extract the attenuation gradient characteristics of the energy attenuation curve as the noise energy propagation characteristics.

[0035] In an embodiment of the present application, for the collected time-domain noise signal, the properties of the sound wave propagation medium are taken into consideration. The medium in the tunnel may include air, rock, etc. Different media have different degrees of absorption and scattering of sound wave energy. At the same time, the propagation distance from the sound source to each receiving point is measured. Based on the above information, the energy attenuation curve of the time-domain noise signal during the propagation process is calculated using relevant acoustic principles and algorithms. For example, as the propagation distance increases, the energy gradually decreases, forming a curve. Then, this energy attenuation curve is analyzed to extract its attenuation gradient characteristics. For example, the slope of the curve at different points is calculated to obtain a series of values ​​representing the attenuation rate, which constitute the noise energy propagation characteristics.

[0036] Step 125: jointly encode the noise spectrum characteristics, the noise source orientation characteristics, and the noise energy propagation characteristics to generate a noise characteristic distribution of the time domain noise signal.

[0037] Optionally, the noise spectrum features, noise source azimuth features, and noise energy propagation features extracted previously are jointly encoded. For example, a Huffman coding algorithm is employed to integrate and encode the statistical features of the noise spectrum features, the azimuth angle and propagation path information of the noise source azimuth features, and the attenuation gradient value of the noise energy propagation features. Through this encoding method, the different types of feature information are converted into a unified distribution representation that fully represents the characteristics of the time-domain noise signal, generating a noise feature distribution for the time-domain noise signal.

[0038] Step 130: Call the deep learning noise suppression model to suppress the noise feature distribution and generate a suppression control feature set corresponding to each time domain noise signal.

[0039] As an optional technical solution, calling a deep learning noise suppression model to suppress the noise feature distribution and generate a suppression control feature set corresponding to each time domain noise signal includes:

[0040] Step 131: Input the noise feature distribution into the encoder module of the deep learning noise suppression model for feature compression processing to generate a noise coding feature vector.

[0041] Optionally, the generated noise feature distribution is input into an encoder module of a deep learning noise suppression model. The encoder module compresses and abstracts the input features to extract key information. For example, the encoder module can be composed of multiple convolutional and pooling layers. The noise feature distribution is input into these layers in a specified format. After convolution operations are performed to extract and transform features of different feature dimensions, pooling operations are then used to reduce the dimensionality of the data, ultimately generating a noise encoded feature vector that contains the important information from the noise feature distribution and has a reduced dimensionality compared to the original feature distribution.

[0042] Step 132: Input the noise coding feature vector into the decoder module of the deep learning noise suppression model to perform suppression parameter reconstruction processing to generate an initial suppression parameter sequence.

[0043] The generated noise-encoded feature vector is input to the decoder module of the deep learning noise suppression model. The decoder module's task is to reconstruct the suppression parameters from the encoded feature vector. For example, the decoder module progressively decompresses and transforms the noise-encoded feature vector through a series of deconvolutional and fully connected layers. In this process, the model learns how to recover the noise suppression-related parameter information from the encoded feature vector, ultimately generating an initial suppression parameter sequence containing preliminary parameter values ​​for subsequent noise suppression operations.

[0044] Step 133: performing time domain alignment processing on the initial suppression parameter sequence, adjusting the phase and amplitude of the initial suppression parameter sequence according to the timestamp of the time domain noise signal, and generating the suppression control feature set.

[0045] As another optional technical solution, performing time domain alignment processing on the initial suppression parameter sequence, adjusting the phase and amplitude of the initial suppression parameter sequence according to the timestamp of the time domain noise signal, and generating the suppression control feature set includes:

[0046] Step 1331: Acquire a timestamp sequence of the time-domain noise signal, where the timestamp sequence includes a timing identifier of each sampling point of the time-domain noise signal.

[0047] Before processing the initial suppression parameter sequence, first obtain a timestamp sequence for the time-domain noise signal. This timestamp sequence records the temporal order of each sampling point in the time-domain noise signal. For example, the timestamp sequence can be an array of time values ​​arranged in sampling order, with each value corresponding to a sampling point, accurately representing the position of that sampling point on the time axis.

[0048] Step 1332: Perform timestamp synchronization processing on the initial suppression parameter sequence, perform interpolation and reconstruction on the initial suppression parameter sequence according to the sampling interval of the timestamp sequence, and generate a timestamp-synchronized suppression parameter sequence.

[0049] Optionally, the initial suppression parameter sequence is reconstructed by interpolation based on the sampling interval of the acquired timestamp sequence. For example, if the sampling interval of the timestamp sequence is fixed, but the sampling interval of the initial suppression parameter sequence is inconsistent with that of the timestamp sequence, then an interpolation algorithm is used to insert or delete some parameter values ​​at appropriate positions so that the time points of the initial suppression parameter sequence correspond to the time points of the timestamp sequence, thereby generating a timestamp-synchronized suppression parameter sequence.

[0050] Step 1333: extract the phase reference curve of the time domain noise signal, and calculate the phase offset set of the timestamp-synchronized suppression parameter sequence, where the phase offset set includes the absolute value of the phase difference of each sampling point.

[0051] A phase reference curve is extracted from the time-domain noise signal, reflecting its phase variation. The timestamp-synchronized suppression parameter sequence is then compared with the phase reference curve. The absolute phase difference between the suppression parameter sequence and the phase reference curve at each sampling point is calculated. These absolute phase differences constitute the phase offset set.

[0052] Step 1334: Perform phase compensation processing on the timestamp-synchronized suppression parameter sequence according to the phase offset set to generate a phase-aligned suppression parameter sequence.

[0053] The timestamp-synchronized suppression parameter sequence is phase-compensated based on the calculated phase offset set. For example, for each sampling point, if the phase offset is positive, the phase of the suppression parameter sequence is increased by the corresponding value; if the phase offset is negative, the phase of the suppression parameter sequence is decreased by the corresponding value. This aligns the phase of the suppression parameter sequence with the phase reference curve of the time-domain noise signal, generating a phase-aligned suppression parameter sequence.

[0054] Step 1335: Extract the amplitude envelope curve of the time domain noise signal, and calculate the amplitude scaling factor set of the phase-aligned suppression parameter sequence, where the amplitude scaling factor set includes the amplitude proportional coefficient of each sampling point.

[0055] The amplitude envelope curve is extracted from the time-domain noise signal, describing how the amplitude of the time-domain noise signal changes over time. The phase-aligned suppression parameter sequence and the amplitude envelope curve are then compared, and the amplitude ratio coefficient between the suppression parameter sequence and the amplitude envelope curve is calculated at each sampling point. These coefficients constitute the set of amplitude scaling factors.

[0056] Step 1336: Perform amplitude adjustment processing on the phase-aligned suppression parameter sequence according to the amplitude scaling factor set to generate an amplitude-matched suppression parameter sequence.

[0057] Optionally, the amplitude of the phase-aligned suppression parameter sequence is adjusted based on a set of amplitude scaling factors. For example, if the amplitude scaling factor is greater than 1, the amplitude of the suppression parameter sequence at the corresponding sampling point is increased by a corresponding multiple; if the amplitude scaling factor is less than 1, the amplitude scaling factor is reduced by a corresponding multiple, ultimately generating an amplitude-matched suppression parameter sequence.

[0058] Step 1337: Perform dimensional consistency processing on the amplitude-matched suppression parameter sequence to achieve unit unification of the phase dimension and the amplitude dimension of the suppression parameter sequence.

[0059] It is understood that since phase and amplitude may have different dimensions, normalization is used to unify the units of the phase and amplitude dimensions. For example, if the units of phase are radians and the units of amplitude are volts, they can be converted into a unified and comparable unit form through an adaptive conversion relationship.

[0060] Step 1338: Perform feature dimension splicing processing on the phase parameters and amplitude parameters of the amplitude-matched suppression parameter sequence to generate the suppression control feature set.

[0061] Furthermore, the phase parameters and amplitude parameters of the amplitude-matched inhibition parameter sequence that has undergone dimensional consistency processing are concatenated in terms of feature dimensions. For example, the phase parameters and amplitude parameters are combined in a predetermined order to form a new vector or matrix representation, which is the generated inhibition control feature set.

[0062] In an alternative embodiment, the training process of the deep learning noise suppression model includes:

[0063] Step 210: Obtain a historical noise signal sample set and a corresponding suppression effect label, wherein the suppression effect label includes a residual noise energy value after noise suppression.

[0064] In this step, a large number of historical noise signal sample sets are collected. These samples come from different tunnel construction scenarios and cover the noise conditions generated by various combinations of construction equipment. For each noise signal sample, the corresponding suppression effect label is also recorded. The residual noise energy value in the label is measured after the actual noise suppression operation. For example, there is a historical noise signal sample that is generated by the simultaneous operation of an excavator, loader, and concrete mixer during the construction of a tunnel. After being processed by the noise suppression equipment, the residual noise energy value is measured as a quantized value and used as the suppression effect label for the sample.

[0065] Step 211: Obtain an initial neural network model, where the initial neural network model includes a convolutional coding layer, an attention fusion layer, and a fully connected decoding layer.

[0066] Optionally, the convolutional coding layer consists of multiple parallel convolution channels, each convolution channel corresponds to a convolution kernel of a different scale, which is used to extract and compress the input noise features; the attention fusion layer can automatically focus on the importance of different features and perform fusion processing; the fully connected decoding layer is responsible for reconstructing the suppression parameters based on the encoded features.

[0067] Step 212: Input the historical noise signal sample set into the initial neural network model for forward propagation processing to generate a prediction suppression parameter sequence.

[0068] Optionally, the historical noise signal sample set is sequentially input into the initial neural network model. In the convolutional coding layer, the noise feature distribution is input into multiple parallel convolution channels for feature mapping processing to generate intermediate feature maps of multiple scales. For example, different convolution channels perform different degrees of feature extraction on noise spectrum features, noise source orientation features, etc., to obtain multiple intermediate feature maps of different scales and feature representations. Then, cross-channel feature fusion processing is performed on the above intermediate feature maps to generate a fused feature map. Next, the fused feature map is subjected to maximum pooling processing to generate a noise coding feature vector. This vector is processed by the attention fusion layer to highlight important features, and then input into the fully connected decoding layer to finally generate a predicted suppression parameter sequence.

[0069] Step 213: Calculate the mean square error loss based on the predicted suppression parameter sequence and the suppression effect label, and update the weight parameters of the initial neural network model based on the mean square error loss until convergence, to obtain the deep learning noise suppression model.

[0070] In one possible embodiment, the convolutional coding layer includes multiple parallel convolution channels, each convolution channel corresponds to a convolution kernel of a different scale, and the processing process of the convolutional coding layer includes:

[0071] Step 310: Input the noise feature distribution into the multiple parallel convolution channels for feature mapping processing to generate intermediate feature maps of multiple scales; perform cross-channel feature fusion processing on the intermediate feature maps of the multiple scales to generate a fused feature map; perform maximum pooling processing on the fused feature map to generate the noise encoding feature vector.

[0072] In an embodiment of the present application, the predicted suppression parameter sequence is compared with the suppression effect label, and the mean square error loss is calculated. For example, for each sample's predicted suppression parameter sequence and the residual noise energy value in the corresponding suppression effect label, the average of the sum of squared errors between them is calculated. Based on the mean square error loss, the back propagation algorithm is used to update the weight parameters of the initial neural network model. In each iteration, the connection weights between each neuron in the convolutional coding layer, the attention fusion layer, and the fully connected decoding layer are adjusted so that the mean square error loss gradually decreases until the model converges, and a deep learning noise suppression model is finally obtained.

[0073] Step 140: driving the noise suppression device to perform a noise cancellation operation based on the suppression control feature set, generating a cancellation sound wave signal with a phase opposite to the time domain noise signal to reduce the intensity of the ambient noise.

[0074] In a preferred embodiment, the step of driving the noise suppression device to perform a noise cancellation operation based on the suppression control feature set to generate a cancellation sound wave signal having a phase opposite to that of the time domain noise signal comprises:

[0075] Step 141: generating a target amplitude curve according to the amplitude features in the suppression control feature set; generating a target phase offset according to the phase features in the suppression control feature set.

[0076] In an embodiment of the present application, amplitude features and phase features are extracted from a suppression control feature set. For example, the suppression control feature set is a vector set containing multiple data dimensions, wherein some dimensions correspond to amplitude features and other dimensions correspond to phase features. For the amplitude feature, the amplitude feature data is smoothed, fitted with curves, and other operations are performed through relevant algorithms and data processing methods (polynomial fitting algorithm) to generate a target amplitude curve. This curve describes the amplitude that the cancellation sound wave signal should have at different time points. For the phase feature, the target phase offset is also determined through statistical analysis methods. For example, by performing statistical analysis on the phase feature data, its variation pattern is found, thereby obtaining a target phase offset that can make the cancellation sound wave signal have a phase opposite to that of the original time domain noise signal.

[0077] Step 142: Modulate the reference acoustic wave signal based on the target amplitude curve and the target phase offset to generate the cancellation acoustic wave signal.

[0078] It can be understood that after obtaining the target amplitude curve and the target phase offset, a reference sound wave signal is selected. The reference sound wave signal can be a pre-set simple sound wave signal with a standard frequency, amplitude, and phase. Then, the reference sound wave signal is modulated according to the target amplitude curve and the target phase offset. For example, at each time point, the amplitude of the reference sound wave signal is adjusted according to the target amplitude curve, and the phase of the reference sound wave signal is changed according to the target phase offset. Through the above modulation method, the amplitude and phase of the reference sound wave signal change in the expected manner, and ultimately a cancellation sound wave signal with a phase opposite to that of the original time domain noise signal is generated.

[0079] Step 143: Directly emitting the cancelling sound wave signal through the speaker array of the noise suppression device, so that the cancelling sound wave signal and the time domain noise signal generate destructive interference in the spatial superposition area to reduce the intensity of the ambient noise.

[0080] At the tunnel construction site, the noise suppression device is equipped with a speaker array, which is precisely arranged in an appropriate position to ensure that the canceling sound wave signal can be effectively transmitted. After the canceling sound wave signal is generated, it is emitted by the speaker array in a set direction and angle. For example, based on the previously determined azimuth characteristics of the noise source, the speaker array is adjusted to a position that allows the canceling sound wave signal to propagate directly in the direction of the noise source. The canceling sound wave signal propagates in space and encounters the original time-domain noise signal in a set superposition area. Because the canceling sound wave signal and the original time-domain noise signal have opposite phases, they undergo destructive interference in the superposition area. During the interference process, the amplitudes of the two sound waves cancel each other, reducing the ambient noise intensity in the area. For example, the noise intensity in the area was originally at a certain high value, but after destructive interference, the noise intensity is reduced to an acceptably low level.

[0081] As a non-limiting embodiment, after generating a cancellation sound wave signal having a phase opposite to that of the time domain noise signal to reduce the intensity of the ambient noise, the method further includes:

[0082] Step 400: real-time acquisition of a residual noise signal set after cancellation, wherein the residual noise signal set includes time domain waveforms and spectral distributions of multiple residual noise components; performing energy threshold detection processing on the residual noise signal set, identifying residual noise components exceeding a preset energy threshold as target optimized noise components; extracting optimized characteristic parameters of the target optimized noise components, wherein the optimized characteristic parameters include residual energy gradient, spectral offset, and phase residual deviation; performing correction processing on the suppression control feature set according to the optimized characteristic parameters, generating a phase compensation amount and an amplitude compensation coefficient; and updating the waveform parameters of the canceled acoustic wave signal based on the phase compensation amount and the amplitude compensation coefficient to achieve iterative suppression of the target optimized noise components.

[0083] After reducing the ambient noise intensity, a preset acquisition device is used to collect the offset residual noise signal set in real time. For example, acquisition devices are arranged at multiple locations in the tunnel to ensure that the residual noise signals in each area can be fully collected. The above residual noise signal set contains multiple residual noise components, each of which has its corresponding time domain waveform and spectral distribution. The collected residual noise signal set is subjected to energy threshold detection processing, and an energy threshold is pre-set. For example, the threshold is determined based on the actual noise control requirements and the ambient background noise level. The energy of each residual noise component is compared with the threshold, and the residual noise component with energy exceeding the threshold is identified as the target optimization noise component.

[0084] For the above-mentioned target optimized noise component, its optimized characteristic parameters are further extracted. For example, the residual energy gradient is calculated, that is, the rate of change of the residual noise energy over time; the spectrum offset is analyzed, that is, the movement of the residual noise spectrum on the frequency axis compared with the original noise spectrum; the phase residual deviation is measured, that is, the phase difference between the residual noise and the ideal cancellation state. According to the above-mentioned optimized characteristic parameters, the suppression control feature set is corrected. For example, the phase compensation amount and amplitude compensation coefficient can be calculated based on the residual energy gradient, spectrum offset and phase residual deviation. Finally, the above-mentioned phase compensation amount and amplitude compensation coefficient are used to update the waveform parameters of the cancellation sound wave signal, such as changing the amplitude, adjusting the phase, etc., so that the cancellation sound wave signal is re-emitted to iteratively suppress the target optimized noise component and further reduce the intensity of the ambient noise.

[0085] As a non-limiting embodiment, after generating a cancellation sound wave signal having a phase opposite to that of the time domain noise signal to reduce the intensity of the ambient noise, the method further includes:

[0086] Step 500: Obtain a set of operating status parameters of the construction equipment, the operating status parameter set including the equipment type, location coordinates and operating power; determine a priority weight coefficient according to the equipment type, and calculate the noise radiation influence range of each construction equipment in combination with the location coordinates; generate a multi-device noise coupling feature set based on the operating power and the noise radiation influence range, the coupling feature set characterizing the energy interference distribution of different equipment noises in the spatial superposition area; perform collaborative optimization processing on the suppression control feature set according to the coupling feature set, and adjust the emission timing and spatial pointing angle of the canceling sound wave signal to reduce the superposition interference intensity of multi-device noise.

[0087] During tunnel construction, the equipment monitoring system acquires a set of operating status parameters for construction equipment. For example, construction equipment includes excavators, loaders, and concrete mixers. The system records each device's type, location coordinates within the tunnel, and current operating power in real time. Priority weighting coefficients are determined based on the device type. Different types of equipment are assigned different weights due to their noise characteristics and impact on the construction environment. For example, excavators, which are noisy and have a significant impact on construction progress, may be assigned a higher weighting coefficient; while smaller auxiliary equipment may have relatively lower weighting coefficients. Based on the device's location coordinates, acoustic models and algorithms are used to calculate the noise radiation impact range of each piece of construction equipment. For example, the spatial region affected by each piece of equipment's noise is determined by considering factors such as the propagation distance of sound waves, the geometry of the tunnel, and the medium properties. Based on the device's operating power and noise radiation impact range, a multi-device noise coupling feature set is generated. This coupling feature set describes the energy interference distribution of the noise from different devices in the spatial overlap region. For example, the overlapping effect of the noise from an excavator and a loader within a certain area is analyzed to determine the location and degree of energy enhancement or reduction. Based on this coupling feature set, a collaborative optimization process is performed on the suppression control feature set. For example, by adjusting the emission timing of the cancelling sound wave signal, the cancelling sound wave signal can play a more effective role during the critical time period when the noise of different devices overlaps; at the same time, by adjusting the spatial pointing angle of the speaker array, the cancelling sound wave signal can be emitted more accurately to the noise overlapping area, thereby reducing the superposition interference intensity of multiple device noises and further improving the noise conditions in the tunnel construction environment.

[0088] As a non-limiting embodiment, after generating a cancellation sound wave signal having a phase opposite to that of the time domain noise signal to reduce the intensity of the ambient noise, the method further includes:

[0089] Step 600: Perform sound field sampling processing on the overlapping area of ​​the canceling sound wave signal and the time domain noise signal to generate a sound pressure distribution map; extract the local sound pressure area coordinates and energy fluctuation frequency in the sound pressure distribution map; adjust the spatial directivity angle of the speaker array of the noise suppression device according to the local sound pressure area coordinates, and generate frequency domain equalization parameters based on the energy fluctuation frequency; perform frequency band matching processing on the frequency domain equalization parameters and the suppression control feature set, and reconstruct the frequency spectrum profile of the canceling sound wave signal to cover the full frequency band of the energy fluctuation frequency.

[0090] Optionally, multiple sound field sampling points are arranged in the overlapping area of ​​the canceling sound wave signal and the time-domain noise signal, and sound pressure data is collected using relevant sound field sampling equipment at preset time intervals and spatial distributions. For example, sampling points are set at regular intervals within the overlapping area to continuously collect sound pressure signals over a period of time. The collected sound pressure data is processed and analyzed to generate a sound pressure distribution map that intuitively displays the spatial distribution of sound pressure within the overlapping area. From the sound pressure distribution map, local sound pressure regions (i.e., areas with relatively high sound pressure values) are identified, and the coordinates of these regions are recorded. Simultaneously, the energy fluctuations of the sound pressure signal are analyzed to determine the energy fluctuation frequency. For example, it may be observed that the energy of the sound pressure signal exhibits fluctuations at corresponding frequencies within certain time periods. Based on the coordinates of the local sound pressure regions, the spatial directivity angle of the noise suppression device's speaker array is adjusted to align the speakers with the high-energy regions to enhance the cancellation effect. Furthermore, frequency domain equalization parameters can be generated based on the energy fluctuation frequency. These parameters are used to adjust the amplitude and phase of the canceling sound wave signal at different frequencies to achieve a better frequency response. Frequency-domain equalization parameters are matched to the suppression control feature set through frequency band matching. For example, the frequency-related parameters in the suppression control feature set are adjusted and optimized based on the frequency-domain equalization parameters. This method reconstructs the spectral profile of the canceling acoustic signal to cover the entire frequency range of energy fluctuations, effectively suppressing noise in the area and further improving the acoustic quality of the tunnel construction environment.

[0091] Furthermore, when implementing the above technical solution, those skilled in the art can perform systematic optimization processing based on the adaptive filter group design, short-time Fourier transform algorithm, direction of arrival estimation method, deep learning model architecture and acoustic interference principle in the existing technology, so as to achieve logical connection optimization, spatial positioning and dimensional unification of the filter group and spectrum analysis.

[0092] Furthermore, for multiscale decomposition and spectral analysis, an improved filter bank joint analysis framework can be employed. During the adaptive filter bank design phase, a transition bandwidth matching the short-time Fourier transform window function is pre-set, and an overlap-preservation method is introduced to ensure sub-signal boundary alignment. For direction-of-arrival estimation in spatial sound field modeling, a azimuth angle calculation model based on characteristic subspace decomposition can be constructed by combining generalized cross-correlation time delay estimation with the MUSIC algorithm. Furthermore, a tunnel wall reflection coefficient compensation mechanism can be introduced to correct propagation path analysis.

[0093] In terms of dimensional consistency processing, the normalization method in the IEEE 1451 smart sensor standard can be adopted. The phase radian value and amplitude volt number are mapped to the [-1, 1] interval through Z-score normalization, and a phase-amplitude synchronization matching mechanism based on dynamic time warping is constructed.

[0094] In detail, for the deep learning model architecture, we can also refer to the U-Net encoder-decoder structure, using parallel multi-scale convolution kernels (such as 32×1, 64×3, 128×5) in the encoder part with the channel attention mechanism, and the decoder part uses deconvolution and jump connection structure, and combines the introduction of relevant activation functions (such as PReLU) and loss functions (weighted mean square error).

[0095] By introducing adaptive filtering algorithms (such as FxLMS) in active noise control, adding a propagation delay compensation module during the sound wave transmission stage, and using the cross-correlation method to calculate the propagation delay from the sound source to the speaker array in real time, it is possible to ensure the temporal and spatial synchronization of the sound wave and noise signals, thereby eliminating the destructive interference failure problem caused by not considering the sound wave propagation speed.

[0096] In summary, the embodiment of the present application fully obtains the complex and diverse noise information in the tunnel construction environment by collecting a set of original noise signals covering multiple time-domain noise signals and the construction equipment noise components contained therein; then, based on the noise feature extraction and processing, the noise feature distribution obtained can comprehensively characterize the noise characteristics from multiple dimensions such as noise spectrum characteristics, noise source orientation characteristics and noise energy propagation characteristics; then, the deep learning noise suppression model is used to suppress the above features, which can intelligently analyze and generate a targeted suppression control feature set; further, based on the set, the noise suppression device is driven to perform noise cancellation operations, which can accurately generate cancellation sound wave signals with opposite phases to each time-domain noise signal.

[0097] This design effectively and specifically reduces the intensity of ambient noise and minimizes the potential hazards caused by noise. It also improves the accuracy and efficiency of construction operations, avoids miscommunication and operational errors caused by noise interference, and overall improves the quality and safety of the tunnel construction process.

[0098] See also Figure 2 As shown, this figure is a schematic diagram of the basic structure of a tunnel construction environment noise suppression system 200 using a deep learning algorithm provided in an embodiment of the present application. The tunnel construction environment noise suppression system 200 using a deep learning algorithm includes:

[0099] Processor 201;

[0100] a storage device 202 having a computer program 2020 stored thereon;

[0101] When the computer program 2020 is executed by the processor 201, the processor 201 implements any of the methods for suppressing tunnel construction environment noise using a deep learning algorithm.

[0102] Based on the above, a readable storage medium is provided, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the above method are implemented.

[0103] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems or devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

Claims

1. A method for suppressing noise in tunnel construction environment using a deep learning algorithm, characterized in that: include: Collecting an original noise signal set at a tunnel construction site, the original noise signal set including a plurality of time-domain noise signals, each of which includes a noise component generated by at least one construction equipment; Performing noise feature extraction processing on the original noise signal set to obtain a noise feature distribution of each time-domain noise signal, wherein the noise feature distribution includes a noise spectrum feature, a noise source orientation feature, and a noise energy propagation feature; Calling a deep learning noise suppression model to suppress the noise feature distribution and generate a suppression control feature set corresponding to each time-domain noise signal: inputting the noise feature distribution into an encoder module of the deep learning noise suppression model for feature compression processing to generate a noise coding feature vector; inputting the noise coding feature vector into a decoder module of the deep learning noise suppression model for suppression parameter reconstruction processing to generate an initial suppression parameter sequence; performing time-domain alignment processing on the initial suppression parameter sequence, adjusting the phase and amplitude of the initial suppression parameter sequence according to the timestamp of the time-domain noise signal, and generating the suppression control feature set; The method comprises the following steps: performing time domain alignment processing on the initial suppression parameter sequence, adjusting the phase and amplitude of the initial suppression parameter sequence according to the timestamp of the time domain noise signal, and generating the suppression control feature set, including: obtaining a timestamp sequence of the time domain noise signal, wherein the timestamp sequence includes a timing identifier of each sampling point of the time domain noise signal; performing timestamp synchronization processing on the initial suppression parameter sequence, interpolating and reconstructing the initial suppression parameter sequence according to the sampling interval of the timestamp sequence, and generating a timestamp-synchronized suppression parameter sequence; extracting a phase reference curve of the time domain noise signal, and calculating a phase offset set of the timestamp-synchronized suppression parameter sequence, wherein the phase offset set includes an absolute value of a phase difference of each sampling point; and reconstructing the suppression parameter sequence according to the phase offset. The shift set performs phase compensation processing on the timestamp-synchronized suppression parameter sequence to generate a phase-aligned suppression parameter sequence; extracts the amplitude envelope curve of the time-domain noise signal, calculates the amplitude scaling factor set of the phase-aligned suppression parameter sequence, and the amplitude scaling factor set includes the amplitude proportional coefficient of each sampling point; performs amplitude adjustment processing on the phase-aligned suppression parameter sequence according to the amplitude scaling factor set to generate an amplitude-matched suppression parameter sequence; performs dimensional consistency processing on the amplitude-matched suppression parameter sequence to achieve unit unification of the phase dimension and the amplitude dimension of the suppression parameter sequence; performs feature dimension splicing processing on the phase parameters and the amplitude parameters of the amplitude-matched suppression parameter sequence to generate the suppression control feature set; The noise suppression device is driven to perform a noise cancellation operation based on the suppression control feature set, generating a cancellation sound wave signal with a phase opposite to that of the time domain noise signal to reduce the intensity of the ambient noise.

2. The method according to claim 1, characterized in that The performing noise feature extraction processing on the original noise signal set to obtain the noise feature distribution of each time domain noise signal includes: Performing multi-scale decomposition processing on the time domain noise signal to generate multiple noise sub-signals, each noise sub-signal corresponding to a noise component in a different frequency band; Performing spectrum analysis on the multiple noise sub-signals, and extracting frequency domain energy distribution characteristics of each noise sub-signal as the noise spectrum characteristics; Performing spatial sound field modeling processing on the time domain noise signal to determine the sound source azimuth and sound wave propagation path characteristics of the time domain noise signal in three-dimensional space, and fusing the sound source azimuth and the sound wave propagation path characteristics into the noise source azimuth characteristics; performing energy attenuation analysis on the time domain noise signal, calculating an energy attenuation curve of the time domain noise signal according to properties of the sound wave propagation medium and the propagation distance, and extracting an attenuation gradient feature of the energy attenuation curve as the noise energy propagation feature; The noise spectrum characteristics, the noise source orientation characteristics and the noise energy propagation characteristics are jointly coded to generate a noise characteristic distribution of the time domain noise signal.

3. The method according to claim 2, characterized in that The performing multi-scale decomposition processing on the time domain noise signal to generate a plurality of noise sub-signals includes: Acquire a plurality of adaptive filter banks of different scales, each adaptive filter bank comprising at least one bandpass filter, wherein a scale parameter of the adaptive filter bank is adjusted according to an average spectral density of the time-domain noise signal; Inputting the time domain noise signal into the multiple adaptive filter banks of different scales for parallel filtering processing to obtain initial noise sub-signals divided into multiple frequency bands; The initial noise sub-signal is subjected to reconstruction error correction processing in the frequency band overlapping region to eliminate the signal aliasing effect between adjacent frequency bands, thereby generating the multiple noise sub-signals.

4. The method according to claim 2, characterized in that The performing spectrum analysis on the multiple noise sub-signals to extract the frequency domain energy distribution characteristics of each noise sub-signal includes: Perform short-time Fourier transform processing on each noise sub-signal to generate a corresponding time-frequency spectrum; Performing peak detection processing on the time-frequency spectrum graph to identify a set of energy peak points in the time-frequency spectrum graph; An energy distribution feature set is constructed according to the distribution density, peak amplitude and frequency interval of the energy peak point set, and statistical features of the energy distribution feature set are extracted as the frequency domain energy distribution features.

5. The method according to claim 2, characterized in that The performing spatial sound field modeling processing on the time domain noise signal to determine the sound source azimuth and sound wave propagation path characteristics of the time domain noise signal in three-dimensional space includes: collecting sound pressure signals of the time domain noise signal at different spatial positions according to the deployed multiple spatial microphone array nodes; Calculating the spatial propagation vector of the time domain noise signal based on the arrival time difference and phase difference of the sound pressure signal; Performing direction of arrival estimation processing based on the spatial propagation vector to determine the azimuth of the sound source; A three-dimensional sound field propagation model is constructed according to the position coordinates of the spatial microphone array nodes and the azimuth angle of the sound source, and the path attenuation coefficient and reflection path characteristics of the three-dimensional sound field propagation model are extracted as the sound wave propagation path characteristics.

6. The method according to claim 1, wherein The training process of the deep learning noise suppression model includes: Acquire a historical noise signal sample set and a corresponding suppression effect label, wherein the suppression effect label includes a residual noise energy value after noise suppression; Obtaining an initial neural network model, wherein the initial neural network model includes a convolutional coding layer, an attention fusion layer, and a fully connected decoding layer; Inputting the historical noise signal sample set into the initial neural network model for forward propagation processing to generate a prediction suppression parameter sequence; Calculating a mean square error loss according to the predicted suppression parameter sequence and the suppression effect label, and updating weight parameters of the initial neural network model based on the mean square error loss until convergence, to obtain the deep learning noise suppression model; Among them, the convolutional coding layer includes multiple parallel convolution channels, each convolution channel corresponds to a convolution kernel of a different scale, and the processing process of the convolutional coding layer includes: inputting the noise feature distribution into the multiple parallel convolution channels for feature mapping processing to generate intermediate feature maps of multiple scales; performing cross-channel feature fusion processing on the intermediate feature maps of multiple scales to generate a fused feature map; performing maximum pooling processing on the fused feature map to generate the noise coding feature vector.

7. The method according to claim 1, characterized in that The step of driving the noise suppression device to perform a noise cancellation operation based on the suppression control feature set to generate a cancellation sound wave signal having a phase opposite to that of the time domain noise signal comprises: generating a target amplitude curve according to the amplitude features in the inhibition control feature set; generating a target phase offset according to a phase feature in the inhibition control feature set; modulating a reference acoustic wave signal based on the target amplitude curve and the target phase offset to generate the cancellation acoustic wave signal; The cancelling sound wave signal is directionally emitted through the speaker array of the noise suppression device, so that the cancelling sound wave signal and the time domain noise signal undergo destructive interference in a spatial superposition area to reduce the intensity of the ambient noise.

8. A tunnel construction environment noise suppression system using a deep learning algorithm, characterized in that: include: processor; A storage device having a computer program stored thereon, wherein when the computer program is executed by the processor, the processor implements the method for suppressing noise in a tunnel construction environment using a deep learning algorithm as described in any one of claims 1 to 7.

Images