Multi-source heterogeneous noise control method and system

Abstract

The present application discloses a multi-source heterogeneous noise control method and system, which relates to the field of construction environment noise control. It collects the noise-affected positioning signal transmitted by the fiber optic inertial navigation system, and uses the empirical mode decomposition technology to decompose the noise-affected positioning signal into a series of basic signals. These signals constitute the various oscillation modes of the original signal as intrinsic mode functions. Then, an artificial intelligence-based signal processing and analysis algorithm is introduced at the back end to analyze and correlate these intrinsic mode functions, so as to identify and filter out the noise signals in a series of basic signals, thereby realizing the reconstruction of the noise-reduced positioning signal. In this way, the multi-source heterogeneous noise control of the positioning signal affected by noise can be effectively realized to ensure the safety and reliability of the construction.

Description

Technical Field

[0001] The present application relates to the field of noise control, and more specifically, to a multi-source heterogeneous noise control method and system. Background Art

[0002] In the construction field, especially in building construction, tunnel construction, bridge construction, road construction and other environments, accurate positioning and posture detection are crucial to improving operational safety and production efficiency. As a high-precision positioning technology that determines the specific position and posture of construction equipment in a construction environment, the fiber optic inertial navigation system has important application prospects in the construction environment.

[0003] However, the working environment in the construction environment is complex and harsh, and is full of various noise sources. Construction equipment will be disturbed by various noises during operation. These noises come from vibration, electromagnetic interference, etc. These noises are multi-source and heterogeneous, that is, they may differ in frequency, amplitude, phase and other aspects. If these noises are not processed, the positioning accuracy of the inertial navigation system will be seriously affected, resulting in large errors in the posture detection of the construction equipment, thus affecting the safety and reliability of the construction.

[0004] Therefore, a multi-source heterogeneous noise control solution is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a multi-source heterogeneous noise control method and system, which collects the noise-affected positioning signal transmitted by the fiber optic inertial navigation system, and uses the empirical mode decomposition technology to decompose the noise-affected positioning signal into a series of basic signals, which constitute the various oscillation modes of the original signal as intrinsic mode functions. Then, an artificial intelligence-based signal processing and analysis algorithm is introduced at the back end to analyze and correlate these intrinsic mode functions, so as to identify and filter out the noise signals in a series of basic signals, thereby realizing the reconstruction of the noise-reduced positioning signal. In this way, the multi-source heterogeneous noise control of the positioning signal affected by noise can be effectively realized, and the positioning accuracy and stability of the fiber optic inertial navigation system in harsh environments are improved, thereby realizing more accurate construction equipment positioning to ensure the safety and reliability of construction.

[0006] According to one aspect of the present application, a method for controlling multi-source heterogeneous noise is provided, which includes:

[0007] Acquire the positioning signal affected by noise transmitted by the fiber optic inertial navigation system;

[0008] Performing empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions;

[0009] Extracting signal features from the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors;

[0010] Identifying noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector; and

[0011] The semantic feature vector of the noise reduction signal is passed through a signal reconstructor based on a generative adversarial network to obtain a reconstructed noise reduction positioning signal.

[0012] According to another aspect of the present application, a multi-source heterogeneous noise control system is provided, comprising:

[0013] A signal acquisition module, used to obtain the positioning signal affected by noise transmitted by the fiber optic inertial navigation system;

[0014] An empirical mode decomposition module, used for performing empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions;

[0015] A signal feature extraction module, used for performing signal feature extraction on the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors;

[0016] A noise identification module, used for identifying noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector; and

[0017] The signal reconstruction module is used to pass the semantic feature vector of the noise reduction signal through a signal reconstructor based on a generative adversarial network to obtain a reconstructed noise reduction positioning signal.

[0018] Compared with the prior art, the present application provides a multi-source heterogeneous noise control method and system, which collects the noise-affected positioning signal transmitted by the fiber optic inertial navigation system, and uses the empirical mode decomposition technology to decompose the noise-affected positioning signal into a series of basic signals, which constitute the various oscillation modes of the original signal as intrinsic mode functions. Then, an artificial intelligence-based signal processing and analysis algorithm is introduced at the back end to analyze and correlate these intrinsic mode functions, so as to identify and filter out the noise signals in a series of basic signals, thereby realizing the reconstruction of the noise-reduced positioning signal. In this way, it is possible to effectively achieve multi-source heterogeneous noise control of the positioning signal affected by noise, improve the positioning accuracy and stability of the fiber optic inertial navigation system in harsh environments, thereby achieving more accurate construction equipment positioning to ensure the safety and reliability of construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0020] Figure 1 is a flow chart of a method for controlling multi-source heterogeneous noise according to an embodiment of the present application;

[0021] Figure 2 A system architecture diagram of a multi-source heterogeneous noise control method according to an embodiment of the present application;

[0022] Figure 3 Flow chart of sub-step S4 of the multi-source heterogeneous noise control method according to an embodiment of the present application;

[0023] Figure 4 It is a block diagram of a multi-source heterogeneous noise control system according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] The working environment in the construction environment is complex and harsh, filled with various noise sources. Construction equipment will be disturbed by various noises during operation. These noises come from vibration, electromagnetic interference, etc. These noises are multi-source and heterogeneous, that is, they may differ in frequency, amplitude, phase and other aspects. If these noises are not processed, the positioning accuracy of the inertial navigation system will be seriously affected, resulting in large errors in the posture detection of the construction equipment, thus affecting the safety and reliability of the construction. Therefore, a multi-source heterogeneous noise control solution is desired.

[0030] In the technical solution of the present application, a method for controlling multi-source heterogeneous noise is proposed. Figure 1 It is a flow chart of a method for controlling multi-source heterogeneous noise according to an embodiment of the present application. Figure 2 : is a system architecture diagram of a multi-source heterogeneous noise control method according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the multi-source heterogeneous noise control method of the embodiment of the present application, the steps include: S1, acquiring a positioning signal affected by noise transmitted by a fiber optic inertial navigation system; S2, performing empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions; S3, performing signal feature extraction on the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors; S4, identifying noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector; and, S5, passing the noise reduction signal semantic feature vector through a signal reconstructor based on an adversarial generative network to obtain a reconstructed noise reduction positioning signal.

[0031] In particular, the S1, obtains a positioning signal affected by noise transmitted by a fiber optic inertial navigation system. It is worth mentioning that the positioning signal transmitted by the fiber optic inertial navigation system (FOG) may be affected by noise, which may come from various sources, such as imperfections of optical components, non-idealities of optical fibers, changes in environmental conditions, etc.

[0032] In particular, S2 performs empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions. Considering that the positioning signal affected by noise is a complex nonlinear and non-stationary signal, there are signal characteristics of various signal sources. Therefore, in order to better understand the frequency domain characteristics and oscillation structure in the positioning signal affected by noise, so as to better identify and distinguish the noise component and the effective signal component, in the technical solution of the present application, it is necessary to perform empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions. It should be understood that empirical mode decomposition is an adaptive signal processing method that allows nonlinear and non-stationary time series to be decomposed into a series of basic signals, called intrinsic mode functions, which represent different oscillation modes in the signal, thereby revealing the intrinsic oscillation characteristics and change laws of the signal. By performing empirical mode decomposition on the positioning signal affected by noise, the complex signal can be effectively decomposed into oscillation modes of different frequencies, providing a basis for subsequent signal processing and noise removal, and helping to improve the accuracy and stability of the positioning system.

[0033] In particular, the S3 extracts signal features from the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors. In order to be able to perform feature analysis on the intrinsic mode functions of different oscillation modes in the noise-affected positioning signal so as to more accurately identify the signal source represented by the oscillation mode of each frequency, and provide a more reliable feature representation for subsequent signal processing and noise removal, in the technical solution of the present application, the series of intrinsic mode functions are further subjected to feature mining in a signal feature extractor based on a convolutional neural network model, so as to extract the semantic features of each intrinsic mode function in the series of intrinsic mode functions, thereby obtaining a series of intrinsic mode signal semantic feature vectors. In particular, the series of intrinsic mode signal semantic feature vectors can better characterize the important features in each intrinsic mode signal, help to distinguish between effective signals and noise components, and further improve the accuracy and stability of the positioning system. Specifically, each layer of the signal feature extractor based on the convolutional neural network model is used to perform the following operations on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolutional feature map; pooling based on the local feature matrix on the convolutional feature map to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activation feature map; wherein the output of the last layer of the signal feature extractor based on the convolutional neural network model is the series of intrinsic mode signal semantic feature vectors, and the input of the first layer of the signal feature extractor based on the convolutional neural network model is the series of intrinsic mode functions.

[0034] Convolutional Neural Network (CNN) is a deep learning model that is specifically designed to process data with a grid structure, such as images and speech. The core idea of ​​CNN is to extract the features of the input data through convolution operations, and to perform high-level representation and abstraction of the features by stacking them layer by layer. The following are the basic components and working principles of CNN: Convolution layer: The convolution layer is the core component of CNN and is used to extract the features of the input data. It performs convolution operations on the input data by applying a set of learnable convolution kernels (filters). The convolution operation can capture local patterns and features in the input data and generate a series of feature maps; Activation function: After the convolution layer, a nonlinear activation function such as ReLU is usually applied. The activation function introduces nonlinear features, allowing the network to learn more complex patterns and representations; Pooling layer: The pooling layer is used to reduce the size and number of parameters of the feature map and extract the most important features. Common pooling operations include maximum pooling and average pooling; Fully connected layer: After a series of convolutional and pooling layers, some fully connected layers are usually added. The fully connected layer converts the feature map of the previous layer into an output result, such as classification or regression; Dropout: In order to prevent overfitting, Dropout technology is often used in CNN. Dropout randomly discards a part of neurons during the training process to reduce the dependency between neurons and improve the generalization ability of the model. Through the back propagation algorithm, CNN can automatically learn to extract features from the input data and optimize according to the training objectives. During the training process, CNN adjusts the network parameters by minimizing the loss function so that the output results are as close to the true labels as possible.

[0035] In particular, the step S4 is to identify the noise component from the series of intrinsic mode signal semantic feature vectors to obtain the noise reduction signal semantic feature vector. In particular, in a specific example of the present application, Figure 3 As shown, S4 includes: S41, calculating the inter-signal autocorrelation semantic measurement weight of each intrinsic modal signal semantic feature vector in the series of intrinsic modal signal semantic feature vectors to obtain a series of inter-signal autocorrelation semantic measurement weights; S42, based on a predetermined threshold, masking the series of inter-signal autocorrelation semantic measurement weights to obtain a sequence of masked inter-signal autocorrelation semantic measurement weights; and, S43, using the sequence of masked inter-signal autocorrelation semantic measurement weights as weights, calculating the position-weighted sum of the series of intrinsic modal signal semantic feature vectors to obtain a denoised signal semantic feature vector.

[0036] Specifically, the S41 calculates the signal autocorrelation semantic measurement weights of each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors to obtain a series of signal autocorrelation semantic measurement weights. It should be understood that in order to explore the correlation between each intrinsic mode signal semantic feature, so as to understand the intrinsic relationship between different intrinsic mode signal features, dig out the patterns and laws between these signal features, and more accurately distinguish the noise component and the effective signal component, in the technical solution of the present application, the signal autocorrelation semantic measurement weights of each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors are further calculated to obtain a series of signal autocorrelation semantic measurement weights. By calculating the signal autocorrelation semantic measurement weights of each intrinsic mode signal semantic feature vector, it is possible to perform autocorrelation correlation analysis between these intrinsic mode signal semantic features, so as to better understand the intrinsic relationship between the semantic features of the intrinsic mode signal, so as to better distinguish the signals of different signal sources, retain the effective components in the positioning signal, and provide a basis for the noise reduction of the positioning signal. In particular, in a specific example of the present application, the signal-to-signal autocorrelation semantic measurement weights of each intrinsic modal signal semantic feature vector in the series of intrinsic modal signal semantic feature vectors are calculated to obtain a series of signal-to-signal autocorrelation semantic measurement weights, including: respectively calculating the matrix products between each weight matrix and each intrinsic modal signal semantic feature vector in the series of intrinsic modal signal semantic feature vectors to obtain a series of coded intrinsic modal signal semantic feature vectors; adding a bias weight vector to each coded intrinsic modal signal semantic feature vector in the series of coded intrinsic modal signal semantic feature vectors to obtain a series of biased intrinsic modal signal semantic feature vectors; activating the series of biased intrinsic modal signal semantic feature vectors to obtain a series of activated intrinsic modal signal semantic feature vectors; and calculating the product between each activated intrinsic modal signal semantic feature vector in the series of activated intrinsic modal signal semantic feature vectors and the transposed vector of a preset reference feature vector to obtain the series of signal-to-signal autocorrelation semantic measurement weights. Wherein, activating the series of biased intrinsic modal signal semantic feature vectors to obtain a series of activated intrinsic modal signal semantic feature vectors includes: using the Selu activation function to activate the series of biased intrinsic modal signal semantic feature vectors. More specifically, the inter-signal autocorrelation semantic metric weights of each intrinsic modal signal semantic feature vector in the series of intrinsic modal signal semantic feature vectors are calculated using the following autocorrelation attention weight formula to obtain the series of inter-signal autocorrelation semantic metric weights; wherein the formula of the autocorrelation attention weight is:

[0037]

[0038] Among them, h i is the i-th intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors, and W i Represent the weight coefficient vector and weight coefficient matrix respectively, B i is the offset vector, Selu(·) represents the Selu function, λ and α are both hyperparameters, e i is the ith inter-signal autocorrelation semantic metric weight in the series of inter-signal autocorrelation semantic metric weights.

[0039] Specifically, the S42, based on a predetermined threshold, masks the series of inter-signal autocorrelation semantic metric weights to obtain a sequence of masked inter-signal autocorrelation semantic metric weights. In particular, in a specific example of the present application, the inter-signal autocorrelation semantic metric weights less than or equal to the predetermined threshold in the series of inter-signal autocorrelation semantic metric weights are set to zero to obtain a sequence of masked inter-signal autocorrelation semantic metric weights. It should be understood that in order to remove the noise signal in the noise-affected positioning signal and retain the effective components of the positioning information therein, in the technical solution of the present application, it is necessary to further mask the series of inter-signal autocorrelation semantic metric weights based on a predetermined threshold to obtain a sequence of masked inter-signal autocorrelation semantic metric weights. By setting a threshold, it can help to screen out inter-signal autocorrelation metric weights with important positioning semantic information and significant correlation, highlight those important features that exceed the threshold, and treat those weights below the threshold as noise or unimportant information for masking, which helps to reduce the impact of noise and improve the accuracy and reliability of positioning signal processing. In particular, in a specific example of the present application, more specifically, based on a predetermined threshold, the series of inter-signal autocorrelation semantic metric weights are masked using the following masking formula to obtain a sequence of masked inter-signal autocorrelation semantic metric weights; wherein the masking formula is:

[0040]

[0041] Among them, e i is the i-th inter-signal autocorrelation semantic metric weight in the series of inter-signal autocorrelation semantic metric weights, mask(·) is a masking operation, is the i-th masked signal autocorrelation semantic metric weight in the sequence of masked signal autocorrelation semantic metric weights.

[0042] Specifically, the S43 uses the sequence of autocorrelation semantic measurement weights between the masked signals as weights to calculate the weighted sum by position of the series of intrinsic mode signal semantic feature vectors to obtain the denoised signal semantic feature vector. It should be understood that the signal with a higher weight after masking represents the effective component related to positioning in the positioning signal affected by noise, while the noise component is effectively identified and eliminated after the masking operation. In addition, although the influence of noise is removed after the masking operation, since different intrinsic mode signals contribute differently to positioning, it is necessary to give higher weights to the semantic features of signals with stronger contributions. Based on this, in the technical solution of the present application, the sequence of autocorrelation semantic measurement weights between the masked signals is further used as weights to calculate the weighted sum by position of the series of intrinsic mode signal semantic feature vectors to obtain the denoised signal semantic feature vector. By using the masked sequence of semantic metric weights of autocorrelation between signals as weights to perform weighted summation between semantic features of different intrinsic mode signals, the effective components related to positioning in the positioning signal under the influence of noise can be highlighted, which helps to focus on the expression of effective information related to positioning, thereby improving the performance and stability of the positioning system. In particular, in a specific example of the present application, the sequence of the semantic metric weights of autocorrelation between masked signals is used as the weight, and the position-weighted sum of the series of semantic feature vectors of intrinsic mode signals is calculated using the following signal global semantic representation formula to obtain the semantic feature vector of the noise reduction signal; wherein the signal global semantic representation is:

[0043]

[0044] in, is the i-th masked signal autocorrelation semantic metric weight in the sequence of masked signal autocorrelation semantic metric weights, h i is the i-th intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors, softmax(·) represents the softmax function, t is the number of vectors in the series of intrinsic mode signal semantic feature vectors, and V is the denoised signal semantic feature vector.

[0045] It is worth mentioning that in other specific examples of the present application, noise components can be identified from the series of intrinsic mode signal semantic feature vectors in other ways to obtain the denoised signal semantic feature vector, for example: input the series of intrinsic mode signal semantic feature vectors; identify the noise components in each intrinsic mode function; perform denoising on each intrinsic mode function according to the identified noise components. Common denoising methods include wavelet denoising, convolutional neural network (CNN) denoising, mean filtering, etc.; recombine the denoised intrinsic mode functions to obtain the denoised signal; perform feature extraction on the denoised intrinsic mode function to obtain the denoised signal semantic feature vector.

[0046] In particular, the S5 passes the semantic feature vector of the noise reduction signal through a signal reconstructor based on a generative adversarial network to obtain a reconstructed noise reduction positioning signal. That is, in the technical solution of the present application, the semantic features of the noise reduction signal after noise reduction and autocorrelation integration with different weights are used to reconstruct the signal, so as to generate a reconstructed noise reduction positioning signal. In this way, it is possible to effectively control the multi-source heterogeneous noise of the positioning signal affected by noise, improve the positioning accuracy and stability of the fiber optic inertial navigation system in harsh environments, thereby achieving more accurate construction equipment positioning to ensure the safety and reliability of construction. It is worth mentioning that Generative Adversarial Networks (GANs) is a deep learning model architecture consisting of a generator and a discriminator, which generates realistic data samples through adversarial training.

[0047] In the technical solution described above, each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors respectively represents the waveform image semantic features of each intrinsic mode function obtained by empirical mode decomposition of the positioning signal affected by noise. Considering that there are obvious source domain signal differences between the intrinsic mode functions obtained by empirical mode decomposition of the positioning signal affected by noise, there are relatively obvious feature distribution differences between the series of intrinsic mode signal semantic feature vectors. Furthermore, in the process of calculating the signal autocorrelation semantic metric weights of each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors to obtain a series of signal autocorrelation semantic metric weights, and based on a predetermined threshold, masking the series of signal autocorrelation semantic metric weights to obtain a sequence of masked signal autocorrelation semantic metric weights, and then using the sequence of masked signal autocorrelation semantic metric weights as weights to calculate the position-weighted sum of the series of intrinsic mode signal semantic feature vectors to obtain the denoised signal semantic feature vector, a masking mechanism is actually applied based on the signal autocorrelation semantic association strength between each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors. In this way, important components are highlighted and unimportant components are suppressed, so that the obtained denoised signal semantic feature vector has the complexity of fusion semantic feature expression of single sample-to-sample waveform semantic features. Therefore, it is expected to improve the overall distribution regression constraint of the denoised signal semantic feature vector based on the predetermined category of the signal reconstructor of the generative adversarial network.

[0048] Based on this, in a preferred embodiment of the present application, the denoised signal semantic feature vector is passed through a signal reconstructor based on a generative adversarial network to obtain a reconstructed denoised positioning signal, specifically including: calculating the point sum of the denoised signal semantic feature vector and the square root of the length of the denoised signal semantic feature vector and the inverse of the square root of the second norm of the denoised signal semantic feature vector to obtain a denoised signal semantic offset feature vector; calculating the exponential function of the denoised signal semantic offset feature vector with a natural constant as the base to obtain a denoised signal semantic offset prediction feature vector; calculating the point product of the denoised signal semantic feature vector and the first norm of the denoised signal semantic feature vector and a weight hyperparameter to obtain a denoised signal semantic constraint feature vector; calculating the point sum of the denoised signal semantic offset prediction feature vector and the denoised signal semantic constraint feature vector to obtain an optimized denoised signal semantic feature vector; and passing the optimized denoised signal semantic feature vector through the signal reconstructor based on a generative adversarial network to obtain the reconstructed denoised positioning signal.

[0049] Specifically, considering the temporal propagation aggregation distribution of the semantic feature vector of the denoised signal for the data embedding coding temporal context of the transaction data based on the overall distribution regression constraint of the classifier, in the above preferred example, the structured norm of the semantic feature vector of the denoised signal is used as the local canonical coordinates for each eigenvalue of the semantic feature vector of the denoised signal to determine the vector overall distribution representation of the semantic feature vector of the denoised signal relative to the class rotation offset of the eigenvalue for the class offset prediction direction of each eigenvalue of the semantic feature vector of the denoised signal as the center, and the bounding box of the vector eigenvalue distribution of the semantic feature vector of the denoised signal is used for eigenvalue constraint to improve the constraint of the semantic feature vector of the denoised signal under the overall regression distribution, thereby improving the training speed of the model and the signal quality of the semantic feature vector of the denoised signal through the signal reconstructor based on the adversarial generation network to obtain the reconstructed denoised positioning signal. In this way, it is possible to effectively achieve multi-source heterogeneous noise control of the positioning signal affected by noise, improve the positioning accuracy and stability of the fiber-optic inertial navigation system in harsh environments, and achieve more accurate construction equipment positioning to ensure the safety and reliability of construction.

[0050] Furthermore, a multi-source heterogeneous noise control system is also provided.

[0051] Figure 4 FIG. 1 is a block diagram of a multi-source heterogeneous noise control system according to an embodiment of the present application. Figure 4 As shown, according to the multi-source heterogeneous noise control system 300 of the embodiment of the present application, it includes: a signal acquisition module 310, which is used to obtain the positioning signal affected by the noise transmitted by the fiber optic inertial navigation system; an empirical mode decomposition module 320, which is used to perform empirical mode decomposition on the positioning signal affected by the noise to obtain a series of intrinsic mode functions; a signal feature extraction module 330, which is used to extract signal features from the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors; a noise identification module 340, which is used to identify noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector; and a signal reconstruction module 350, which is used to pass the noise reduction signal semantic feature vector through a signal reconstructor based on an adversarial generative network to obtain a reconstructed noise reduction positioning signal.

[0052] As described above, the multi-source heterogeneous noise control system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a multi-source heterogeneous noise control algorithm. In a possible implementation, the multi-source heterogeneous noise control system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the multi-source heterogeneous noise control system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the multi-source heterogeneous noise control system 300 can also be one of the many hardware modules of the wireless terminal.

[0053] Alternatively, in another example, the multi-source heterogeneous noise control system 300 and the wireless terminal may also be separate devices, and the multi-source heterogeneous noise control system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0054] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for controlling multi-source heterogeneous noise, characterized in that: include: Acquire the positioning signal affected by noise transmitted by the fiber optic inertial navigation system; Performing empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions; Extracting signal features from the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors; Identifying noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector; as well as The semantic feature vector of the noise reduction signal is passed through a signal reconstructor based on a generative adversarial network to obtain a reconstructed noise reduction positioning signal; The series of intrinsic mode functions are subjected to signal feature extraction to obtain a series of intrinsic mode signal semantic feature vectors, including: the series of intrinsic mode functions are respectively subjected to a signal feature extractor based on a convolutional neural network model to obtain the series of intrinsic mode signal semantic feature vectors; The series of intrinsic mode functions are respectively passed through a signal feature extractor based on a convolutional neural network model to obtain the series of intrinsic mode signal semantic feature vectors, including: Each layer of the signal feature extractor based on the convolutional neural network model is used to perform the following operations on the input data in the forward pass of the layer: Perform convolution on the input data to obtain a convolution feature map; Performing pooling on the convolutional feature map based on a local feature matrix to obtain a pooled feature map; and Performing nonlinear activation on the pooled feature map to obtain an activated feature map; The output of the last layer of the signal feature extractor based on the convolutional neural network model is the series of intrinsic mode signal semantic feature vectors, and the input of the first layer of the signal feature extractor based on the convolutional neural network model is the series of intrinsic mode functions; Identifying noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector, comprising: Calculating the inter-signal autocorrelation semantic metric weight of each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors to obtain a series of inter-signal autocorrelation semantic metric weights; Based on a predetermined threshold, masking the series of inter-signal autocorrelation semantic metric weights to obtain a sequence of masked inter-signal autocorrelation semantic metric weights; and The sequence of the autocorrelation semantic metric weights between the masked signals is used as the weight, and the position-weighted sum of the series of intrinsic mode signal semantic feature vectors is calculated to obtain the denoised signal semantic feature vector.

2. The multi-source heterogeneous noise control method according to claim 1 is characterized in that: Calculating the inter-signal autocorrelation semantic metric weight of each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors to obtain a series of inter-signal autocorrelation semantic metric weights, including: Calculating matrix products between each weight matrix and each intrinsic mode signal semantic feature vector in the series of intrinsic mode signal semantic feature vectors respectively to obtain a series of encoded intrinsic mode signal semantic feature vectors; Adding a bias weight vector to each coded intrinsic modal signal semantic feature vector in the series of coded intrinsic modal signal semantic feature vectors to obtain a series of biased intrinsic modal signal semantic feature vectors; activating the series of biased intrinsic modal signal semantic feature vectors to obtain a series of activated intrinsic modal signal semantic feature vectors; and The product between each activated intrinsic mode signal semantic feature vector in the series of activated intrinsic mode signal semantic feature vectors and the transposed vector of the preset reference feature vector is calculated to obtain the series of inter-signal autocorrelation semantic metric weights.

3. The multi-source heterogeneous noise control method according to claim 2 is characterized in that: Activating the series of biased intrinsic modal signal semantic feature vectors to obtain a series of activated intrinsic modal signal semantic feature vectors includes: using a Selu activation function to activate the series of biased intrinsic modal signal semantic feature vectors.

4. The multi-source heterogeneous noise control method according to claim 3 is characterized in that: Based on a predetermined threshold, masking is performed on the series of inter-signal autocorrelation semantic metric weights to obtain a sequence of masked inter-signal autocorrelation semantic metric weights, including: The inter-signal autocorrelation semantic metric weights that are less than or equal to the predetermined threshold in the series of inter-signal autocorrelation semantic metric weights are set to zero to obtain the sequence of masked inter-signal autocorrelation semantic metric weights.

5. A multi-source heterogeneous noise control system for executing the multi-source heterogeneous noise control method according to any one of claims 1 to 4, characterized in that: include: A signal acquisition module, used to obtain the positioning signal affected by noise transmitted by the fiber optic inertial navigation system; An empirical mode decomposition module, used for performing empirical mode decomposition on the positioning signal affected by noise to obtain a series of intrinsic mode functions; A signal feature extraction module, used for performing signal feature extraction on the series of intrinsic mode functions to obtain a series of intrinsic mode signal semantic feature vectors; A noise identification module, used for identifying noise components from the series of intrinsic mode signal semantic feature vectors to obtain a noise reduction signal semantic feature vector; as well as The signal reconstruction module is used to pass the semantic feature vector of the noise reduction signal through a signal reconstructor based on a generative adversarial network to obtain a reconstructed noise reduction positioning signal.

Images