Method, device and equipment for suppressing interference signal of train

Abstract

The invention relates to the technical field of train control, and discloses a train interference signal suppression method, device and equipment and a storage medium, and the method comprises the steps: obtaining a to-be-processed train control signal; inputting the train control signal into a diffusion model, and inhibiting an interference signal in the train control signal based on the diffusion model to obtain a reconstructed signal; wherein the diffusion model comprises a forward diffusion process and a reverse denoising process; a forward diffusion process: gradually adding a preset noise signal into the original train control signal until the original train control signal is converted into a random noise signal; and a reverse de-noising process: de-noising the random noise signal step by step to obtain a reconstructed signal. By gradually diffusing and denoising and gradually removing noise signals, reconstruction of original control signals is achieved, the influence of interference signals on the original control signals is reduced, the method can adapt to different situations of interference on train control signals, and the method has great practical significance on safe and efficient operation of trains.

Description

Technical Field

[0001] The present invention relates to the field of train control technology, and in particular to a method, a device, a equipment and a storage medium for suppressing train interference signals. Background Art

[0002] With the development of modern communication technology and power electronics technology, the degree of railway electrification and equipment electronicization is getting higher and higher, which means that the electromagnetic influence on each device will become greater and greater. At the same time, the integration of train onboard equipment is getting higher and higher. In addition, in order to effectively utilize the train space and facilitate management, the wiring of onboard equipment signal lines is becoming more and more centralized, which increases the possibility of electromagnetic interference on the signal transmission lines of train onboard equipment. If the train digital baseband control signal is affected by electromagnetic interference, it will cause a malfunction in the train operation. Summary of the invention

[0003] In view of this, the present invention provides a method, device, equipment and storage medium for suppressing train interference signals to solve the problem that the digital baseband control signal of the train is erroneous due to electromagnetic interference, which may cause failure in train operation.

[0004] In the first aspect, the present invention provides a method for suppressing train interference signals, the method comprising: obtaining a train control signal to be processed; inputting the train control signal to be processed into a diffusion model, suppressing the interference signal in the train control signal to be processed based on the diffusion model, and obtaining a reconstructed signal; wherein the diffusion model comprises: a forward diffusion process and a reverse denoising process; the forward diffusion process: gradually adding a preset noise signal to the original train control signal until the original train control signal is converted into a random noise signal; the reverse denoising process: gradually denoising the random noise signal to obtain the reconstructed signal.

[0005] In an optional embodiment, a preset noise signal is gradually added to an original train control signal until the original train control signal is converted into a random noise signal, comprising: gradually adding the preset noise signal to the original control signal until the original control signal is converted into the random noise signal, obtaining time step information for converting the original control signal into the random noise signal and the random noise signal.

[0006] In some optional embodiments, the aforementioned method for suppressing train interference signals further includes: training the denoising process is obtained based on the following steps: inputting the training noise signal, the random noise signal and the time step information into a diffusion denoising network; the diffusion denoising network obtains a reconstructed signal based on the prediction of the training noise signal, calculates the loss between the reconstructed signal and the original control signal, and adjusts the parameters of the diffusion denoising network; repeating the aforementioned process until the loss reaches a preset threshold and the diffusion denoising network training is completed.

[0007] In an optional embodiment, the diffusion denoising network includes: a coding layer, an intermediate layer and a decoding layer; wherein the coding layer is used to extract the time domain features of the train control signal; the intermediate layer is used to extract the connection between the deep features of the train control signal and transmit the connection to the decoding layer; the decoding layer is used to perform feature fusion and feature reconstruction based on the time domain features and the connection; the coding layer and the intermediate layer are jump connected, the coding layer and the decoding layer are jump connected, and the intermediate layer and the decoding layer are jump connected.

[0008] In an optional embodiment, the coding layer includes: at least two layers of residual connection networks, the residual connection networks include a one-dimensional convolution layer and a linear rectification activation function, the time step information is added between at least two layers of the residual connection networks, and the time step information is encoded using sinusoidal position coding; and a downsampling layer, which extracts the signal input to the downsampling layer and transmits the extracted signal to the next coding layer.

[0009] In an optional implementation, obtaining the train control signal to be processed includes: obtaining the time domain signal of each sampling channel of the train; and segmenting the time domain signal based on a field programmable gate array to obtain the train control signal to be processed.

[0010] In an optional embodiment, the train control signal to be processed is input into the diffusion model, and the interference signal in the train control signal to be processed is suppressed based on the diffusion model to obtain a reconstructed signal, including: inputting the train control signal to be processed into the diffusion model, and suppressing the interference signal in the train control signal to be processed based on the diffusion model to obtain a reconstructed signal with multi-channel output; based on the preset weight of each channel and each of the reconstructed signals, obtaining a multi-channel reconstruction result of the train control signal to be processed.

[0011] In a second aspect, the present invention provides a device for suppressing train interference signals, the device comprising: an acquisition module, used to acquire a train control signal to be processed; a reconstruction module, used to input the train control signal to be processed into a diffusion model, suppress the interference signal in the train control signal based on the diffusion model, and obtain a reconstructed signal; wherein, the diffusion model gradually adds a preset noise signal to the train control signal until the train control signal is converted into a random noise signal, and gradually denoises the random noise signal to obtain the reconstructed signal.

[0012] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for suppressing train interference signals of the above-mentioned first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0013] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for suppressing train interference signals according to the first aspect or any corresponding embodiment thereof.

[0014] The interference signal in the train control signal is suppressed based on the diffusion model. The diffusion phenomenon in nature is first imitated, and then the noise signal is gradually removed through the step-by-step diffusion denoising method to achieve reconstruction of the original control signal. This can reduce the impact of the interference signal on the original control signal. It has strong adaptability and can adapt to different situations that interfere with the original control signal of the train. It has great practical significance for the safe and efficient operation of the train. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 A schematic flow chart of a method for suppressing train interference signals provided by an embodiment of the present invention is shown;

[0017] Figure 2 A schematic diagram of the training process of the diffusion denoising network is shown;

[0018] Figure 3 A schematic diagram of the test process of the diffusion model is shown;

[0019] Figure 4The overall structural diagram of the diffusion model is shown;

[0020] Figure 5 The structural schematic diagram of the diffusion denoising network is shown;

[0021] Figure 6 A schematic diagram showing the structure of the encoding layer and the decoding layer is shown;

[0022] Figure 7 A schematic diagram of a multi-channel reconstructed signal is shown;

[0023] Figure 8 A schematic diagram showing the effect of the control signal after noise signal and interference suppression is shown;

[0024] Fig. 9 A schematic diagram of the structure of a device for suppressing train interference signals is shown;

[0025] Fig.10 A schematic diagram showing the hardware structure of a computer device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0027] With the increase in railway operating mileage and the expansion of operating scope, trains have been running from plains to high-altitude areas, which requires greater traction force for train operation, and thus requires greater traction current, and the intensity of electromagnetic interference generated by traction current will also increase. During the normal operation of the train, due to the unevenness of the contact network, the unstable contact between the pantograph and the contact network will generate a bow-net offline pulse. Due to the existence of a phase-splitting area in the contact network, a bow-net offline pulse will also be generated when the train passes through the phase-splitting area. The train bow-net offline will generate a strong electromagnetic interference pulse, and this is the main source of electromagnetic interference for the train.

[0028] In addition, during the operation of the train, if there is thunderstorm, it will be affected by lightning. Lightning enters the train through radiation coupling and conduction coupling, which will affect the signal transmission quality of the train's onboard equipment. If there are some illegal man-made electromagnetic attacks, the high-intensity electromagnetic pulses emitted by them will also enter the train through radiation coupling or conduction coupling, affecting the signal quality of the signal line of the train's onboard equipment.

[0029] Trains run in a complex electromagnetic environment, and the signal transmission lines of onboard equipment are also in a complex electromagnetic environment. In a complex electromagnetic environment, different forms or types of electromagnetic interference enter the signal line in a variety of ways, affecting the normal communication between onboard equipment and the normal transmission of train control signals, which has a potential adverse effect on the safety of train operation.

[0030] Train control signals are subject to electromagnetic interference, which causes signal errors and is an important cause of train operation failures. Onboard control signals will be affected by the above electromagnetic interference, which will pose a threat to driving safety. For example, when the speed sensor is subject to electromagnetic interference and causes information reading errors, the train will suddenly stop, seriously affecting the safety and efficiency of train operation.

[0031] Therefore, the correct transmission of train control signals is the premise for ensuring the normal operation of the train, which puts forward higher requirements for the high-quality transmission of train control signals. Most of the train control signals are digital baseband signals and their modulation signals, which are sometimes also called digital baseband control signals. In the related technology, adaptive filters are used to eliminate interference from digital baseband signals. The design of adaptive filters includes time domain methods such as minimum mean square error or linear prediction based on variable step size, but there are performance defects in low signal-to-noise ratio environments.

[0032] In addition, the Empirical Mode Decomposition (EMD) algorithm is also a commonly used time-domain signal denoising method. EMD decomposes signals or composite signals with background noise, and the obtained intrinsic mode functions (IMFs) and residual components are all single-component signals, that is, they only have a single frequency component. Then, the Hilbert transform is performed to obtain the instantaneous frequency and time-frequency spectrum of each IMF component, which eliminates the limitations of the previous signal processing method that uses the Fourier transform theory, and is significantly improved over the previous time-frequency method. However, EMD decomposition has the phenomenon of modal aliasing, which will cause the physical meaning of the intrinsic mode function components of the signal to be unclear, thereby affecting the accuracy of the algorithm decomposition.

[0033] The wavelet threshold algorithm based on time domain and frequency domain has the disadvantage that the selection of wave basis and threshold has a great influence on the result.

[0034] In addition, deep learning models such as Convolutional Neural Networks (CNN), Denoising Convolutional Neural Networks (DCNN), and Generative Adversarial Networks (GAN) based on deep learning can also be applied to the field of signal interference suppression. However, for train control signals, since train signals are relatively complex and there is a large amount of electromagnetic interference, high-frequency noise and broadband noise will appear on the signal lines of on-board equipment. The commonly used signal interference suppression methods cannot meet the needs of high-speed train on-board control signal interference suppression. Therefore, it is necessary to design an algorithm that can effectively remove interference signals from train on-board control signals.

[0035] According to an embodiment of the present invention, a method embodiment for suppressing train interference signals is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] In this embodiment, a method for suppressing train interference signals is provided, which can be used in a mobile terminal, such as a mobile phone, a tablet computer, a desktop computer, or a train control terminal. Figure 1 A schematic diagram of a flow chart of a method for suppressing train interference signals provided by an embodiment of the present invention is shown. Figure 1 As shown, the process includes the following steps:

[0037] Step S101, obtaining a train control signal to be processed.

[0038] In this step, the voltage or current data of each sampling channel is continuously collected through the data acquisition card to obtain the time domain expression of each sampling channel signal. In the embodiment of the present invention, the quadrature phase shift keying signal (Quadrature Phase Shift Keying, referred to as QPSK) is selected, the amplitude of the signal is 1V, and the sampling points of a single code element are 400.

[0039] Step S103, input the train control signal into the diffusion model, suppress the interference signal in the train control signal based on the diffusion model, and obtain a reconstructed signal; wherein, the diffusion model includes: a forward diffusion process and a reverse denoising process; the forward diffusion process, gradually adds a preset noise signal to the original train control signal until the original train control signal is converted into a random noise signal; the reverse denoising process, gradually denoises the random noise signal to obtain a reconstructed signal.

[0040] In this step, the diffusion model data set includes multiple types of noise, including pulse noise, additive white noise and bow net offline noise. The diffusion model data set is used for training and testing the diffusion model. The diffusion model includes: a forward diffusion process and a reverse denoising process. The train control signal in step S103 is a useful signal. In the forward diffusion process, preset Gaussian noise is continuously added to the useful signal to diffuse the useful signal into a random noise signal. In the reverse denoising process, the result of reconstructing the train control signal is obtained by estimating and removing the added noise. The reverse denoising process is difficult to process directly, and the real reverse process can be approximated by training the diffusion denoising network model.

[0041] The method for suppressing train interference signals provided in this embodiment, when suppressing interference signals in train control signals based on a diffusion model, first imitates the diffusion phenomenon in nature, and then gradually removes the noise signal through a step-by-step diffusion denoising method to achieve reconstruction of the original control signal. This can reduce the impact of interference signals on the original control signal, has strong adaptability, and can adapt to different situations that interfere with the original control signal of the train, and has great practical significance for the safe and efficient operation of the train.

[0042] In some optional embodiments, a preset noise signal is gradually added to the original train control signal until the original train control signal is converted into a random noise signal, including: gradually adding a preset noise signal to the original control signal until the original control signal is converted into a random noise signal, and obtaining time step information for converting the original control signal into the random noise signal.

[0043] In some optional embodiments, the aforementioned method for suppressing train interference signals further includes: training the denoising process based on the following steps: inputting the training noise signal, the random noise signal and the time step information into the diffusion denoising network; the diffusion denoising network obtains the reconstructed signal based on the prediction of the training noise signal, calculates the loss between the reconstructed signal and the original control signal, and adjusts the parameters of the diffusion denoising network; repeats the aforementioned process until the loss reaches a preset threshold and the diffusion denoising network training is completed. Wherein, the training noise signal comes from the diffusion model data set.

[0044] Figure 2 Figure 2 shows a schematic diagram of the training process of the diffusion denoising network. Figure 2 As shown, the preset noise signal is continuously added to the useful signal x0 to form a Markov chain, and the random noise signal x is obtained. t , random noise signal can also be called diffuse noise signal, the specific addition formula is as follows:

[0045]

[0046] in, represents a1*a2*…*a n , a n The preset value represents the degree of noise added to each diffusion. is a gradually decreasing parameter. It decreases as the number of diffusion steps increases and is used to control the intensity of the noise. represents the noise added in the forward diffusion process. The training noise signal y includes: a useful signal and a noise signal. Step S201: Diffusion noise signal x t , time step t and noise signal y, input into the diffusion denoising network; Step S202: The diffusion denoising network obtains the reconstructed signal based on the prediction of the interference signal Step S203: The process of calculating the root mean square error RMSE between the reconstructed signal and the original control signal x(t) is as follows:

[0047]

[0048] By calculating the root mean square error between the reconstructed signal and the original control signal, the loss between the reconstructed signal and the original control information signal can be obtained, and the parameters of the diffusion denoising network can be adjusted based on the back propagation algorithm.

[0049] From the diffusion model data set, a noise signal, a diffusion noise signal obtained based on the forward diffusion process, and corresponding diffusion time step information are selected, input into the diffusion denoising network, and steps S201 to S203 are repeated, and the data in the diffusion model data set are iterated N times until the root mean square error reaches the minimum value or the expected value, and the optimal diffusion denoising network is obtained. In this way, the train control signal can be reconstructed through an efficient and simple loss function, which can improve the accuracy of the diffusion denoising network and reduce the amount of calculation of the diffusion denoising network.

[0050] Figure 3 Figure 2 shows a schematic diagram of the test process of the diffusion model. Figure 3 As shown, in some optional implementations, the aforementioned method for suppressing train interference signals further includes: testing the diffusion model. Specifically, the observed noise signal and the generated random noise signal x T And the time step number T of generating random noise signal, input the optimal diffusion denoising network, and get the denoised signal x at each step T-1 , estimate and remove the noise added in each step until the time step t is zero, and obtain the reconstructed signal.

[0051] In this way, based on the test of the diffusion model, the final performance of the diffusion model can be evaluated, the accuracy and reliability of the diffusion model can be guaranteed, and a basis can be provided for the deployment and application of the diffusion model.

[0052] Figure 4 The overall structure of the diffusion model is shown in Figure 2. Figure 4 As shown, the diffusion model includes: a model training part 401 and a model testing part 402. In the model training part 401, the noisy signal y(t) after the useful signal x(t) and the noise signal n(t) are superimposed is input into the training network, and the diffusion denoising network training is performed to obtain the reconstructed signal Calculate the reconstructed signal The root mean square error between the signal and the useful signal x(t) is used to update the diffusion denoising network parameters according to the back propagation algorithm. After N iterations, until the root mean square error reaches the expected value, it is considered that the optimal diffusion denoising network is obtained, and the training process of the diffusion denoising network model is completed.

[0053] In the model testing part 402, the observed noisy signal is input into the optimal diffusion denoising network to obtain a reconstructed signal.

[0054] In some optional embodiments, the diffusion denoising network includes: a coding layer, an intermediate layer and a decoding layer; wherein the coding layer is used to extract the time domain features of the train control signal; the intermediate layer is used to extract the connections between the deep features of the train control signal, and transmit the connections between the deep features of the train control signal to the decoding layer; the decoding layer is used to perform feature fusion and feature reconstruction based on the time domain features and connections; there is a jump connection between the coding layer and the intermediate layer, a jump connection between the coding layer and the decoding layer, and a jump connection between the intermediate layer and the decoding layer.

[0055] Among them, deep features refer to the abstract representation extracted by the middle layer of the neural network. Deep features capture high-level abstract concepts in the data and help the diffusion model better understand and learn data. Deep features can help improve the performance and generalization ability of the diffusion model.

[0056] Skip connection, also known as residual connection, is an important component in the deep neural network architecture. In the encoding module, intermediate layer module and decoding module of the diffusion model, the input is directly connected to the output to allow information to jump between different modules. The main purpose of the skip connection is to solve the gradient disappearance problem in deep neural networks and accelerate the training process. The skip connection allows the gradient to be back-propagated more easily by passing the input information directly to the subsequent layers. It helps to maintain the stability of the gradient and makes the diffusion model easier to train.

[0057] Each residual block in the diffusion model contains a skip connection. In Resnet, the output of each residual block is added to the input and then passed through the activation function ReLU. This allows the diffusion model to learn the residual (difference) instead of directly learning the entire mapping, thereby reducing the difficulty of training.

[0058] Figure 5 Figure 2 shows the structural diagram of the diffusion denoising network. Figure 5 As shown, the neural network part of the diffusion model, that is, the overall structure of the diffusion denoising network can adopt the U-net network structure, in which the encoding layer is used to extract the time domain features of the signal, and the decoding layer is used to restore the signal. The diffusion denoising network includes: three layers of down-sampling encoding layers, one intermediate layer and three layers of up-sampling decoding layers. The input of the diffusion denoising network is a diffusion T-step signal, which is input in batches of 16 at a time, with a size of 1*400. After passing through the first residual connection network, the diffusion T-step signal is superimposed with the noise signal as the conditional input, and after passing through three layers of down-sampling encoding layers (RRD), intermediate layers (R+Att+R) and three layers of up-sampling decoding layers (RRU), the output reconstructed signal is obtained. The intermediate layer includes a residual connection network R connected in series, a self-attention mechanism layer Att and another residual connection network R.

[0059] The use of jump connections between the encoding layer, the intermediate layer, and the decoding layer can retain the details lost during the encoding process and improve the accuracy of the signal recovery results. Downsampling can improve the diffusion model's ability to identify the minority class by selecting a portion of the data from the majority class and recombining it with the minority class into a new data set to balance the number of samples in different categories, which can effectively reduce the amount of data and balance the data set.

[0060] In some optional embodiments, the coding layer includes: at least two layers of residual connection network (Resnet), the residual connection network includes a one-dimensional convolution layer (Conv1d) and a linear rectification activation function (ReLU), time step information (t) is added between at least two layers of residual connection network, and the time step information is encoded by sinusoidal position; and, a downsampling layer (Downsample), which extracts the signal input to the downsampling layer and transmits the extracted signal to the next level coding layer.

[0061] Figure 6 Schematic diagram of the structure of the coding layer and the decoding layer is shown. Figure 6 As shown, in this embodiment, the mathematical description of the coding layer is:

[0062]

[0063] Among them, x l represents input, x l+1 Indicates the output, Represents a residual block. The diffusion time step information t is added between the first residual block 601 and the second residual block 602. The downsampling process (D) can use 1*2 maximum pooling. In this way, the accuracy of the diffusion model can be improved and the calculation amount of the diffusion model can be reduced.

[0064] In some optional embodiments, the decoding layer includes: at least two layers of residual connection network (Resnet), the residual connection network includes a one-dimensional convolution layer (Conv1d) and a linear rectifier activation function (ReLU), time step information (t) is added between the two residual connection networks, and the time step information is encoded by sinusoidal position; and, an upsampling layer (Upsample), using a ConvTranspose1d deconvolution structure with a step size of 2. The input of the decoding layer is jump-connected to the feature map of the corresponding encoding layer.

[0065] Specifically, the structure of the coding layer is RRD, and the structure of the decoding layer is RRU. The structural setting of the coding layer, the decoding layer and the intermediate layer can improve the efficiency of the diffusion model and improve the accuracy of the signal reconstruction result.

[0066] In some optional embodiments, obtaining a train control signal to be processed includes: obtaining a time domain signal of each sampling channel of the train; obtaining the time domain signal based on a field programmable gate array; and the field programmable gate array segments the time domain signal to obtain the train control signal to be processed.

[0067] In this embodiment, based on segmenting the time domain signal, a signal sequence corresponding to a single code element is obtained. The time domain signal and the train control signal to be processed obtained after segmentation are stored. In this way, the train control signal to be processed can be collected in real time, the requirements of the diffusion model can be met, and the timing problem can be solved to ensure that the train control signal to be processed can remain stable within a specific clock cycle.

[0068] In some optional embodiments, the field programmable gate array segments the time domain signal, including: segmenting the time domain signal using a preset sliding window, where the preset sliding window is determined based on the length of a single code element of the train control signal to be processed.

[0069] In this embodiment, the length of a single slide is the length of one symbol. A 1*400 sliding window can be selected, and the sliding interval can be set to 400 to segment the QPSK signal into single symbol segments, and transmit the segmented QPSK signal to the diffusion model.

[0070] In some optional embodiments, the train control signal to be processed is input into a diffusion model, and the interference signal in the train control signal to be processed is suppressed based on the diffusion model to obtain a reconstructed signal, including: inputting the train control signal to be processed into a diffusion model, and suppressing the interference signal in the train control signal to be processed based on the diffusion model to obtain a reconstructed signal with multi-channel output; based on the preset weight of each channel and each reconstructed signal, a multi-channel reconstruction result of the train control signal to be processed is obtained.

[0071] Figure 7 FIG. 1 shows a schematic diagram of a multi-channel reconstructed signal. Figure 7 As shown, in this embodiment, the original control signal can be restored by majority decision, the diffusion model is set to multi-channel output, and an appropriate weight is assigned to the output result of each channel during the training phase of the diffusion model, for example, ω1, ω2…ω n , multiply the output result of each channel by its corresponding preset weight and then sum them up to obtain the multi-channel reconstruction result of the train control signal to be processed.

[0072] During the training and testing process of the diffusion model, the one-dimensional sequence of several segments of single code element signals is input into the base diffusion model for training, so that the root mean square error between the weighted signals reconstructed by the diffusion model and the original control signal reaches the expected value, and the optimal diffusion model is obtained. During the test process, the observed one-dimensional time-domain noisy signal sequence is input into the optimal diffusion model to obtain the useful signal after multi-channel reconstruction. In this way, the original control signal can be effectively restored, and the stability and accuracy of signal reconstruction can be improved.

[0073] The diffusion model in the embodiment of the present invention is trained on a training set with Gaussian noise interference of signal-to-noise ratios of -30dB, -20dB, -10dB and 0dB. The computer used is Windows 11 operating system, 32G memory; CPU: Intel(R) Core(TM) Ultra 5 125H; GPU: NVIDIA GeForce RTX 4060Laptop GPU. The diffusion model application process is to input the observed noise signal into the optimal diffusion denoising network to obtain a reconstructed signal. 10 groups of data with different signal-to-noise ratios are processed by the method for suppressing train interference signals provided by the embodiment of the present invention. The average value of the obtained results is shown in Table 1, where the signal-to-noise ratio unit in Table 1 is dB.

[0074] Table 1

[0075] Observation signal-to-noise ratio Signal-to-noise ratio after diffusion model reconstruction Signal-to-noise ratio improvement Root mean square error -0.204 31.107 31.311 0.01969 -9.695 29.892 39.587 0.02264 -19.315 31.565 50.880 0.01867 -24.660 29.973 54.633 0.02242

[0076] It can be seen from Table 1 that the method for suppressing train interference signals provided by the embodiments of the present invention shows good performance under noise interference above -25dB, and the signal-to-noise ratio of the reconstructed signal is significantly improved. The generative interference suppression method in the embodiments of the present invention can well extract and integrate the waveform features of the train control signal, and then restore the original signal waveform to the greatest extent during the reconstruction process.

[0077] Figure 8 Figure 2 shows a schematic diagram of the effect of the control signal after noise signal and interference suppression. Figure 8 As shown in the figure, SNR represents signal-to-noise ratio. The method for suppressing train interference signals provided by the embodiment of the present invention significantly improves the effect and performance of signal reconstruction compared with the current algorithm based on convolutional denoising neural network and wavelet transform.

[0078] In this embodiment, a device for suppressing train interference signals is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0079] This embodiment provides a device for suppressing train interference signals, such as Fig. 9 As shown, including:

[0080] The acquisition module 901 is used to acquire the train control signal to be processed.

[0081] The reconstruction module 902 is used to input the train control signal to be processed into the diffusion model, suppress the interference signal in the train control signal based on the diffusion model, and obtain the reconstructed signal; wherein, the diffusion model includes: a forward diffusion process and a reverse denoising process; the forward diffusion process: by gradually adding a preset noise signal to the original train control signal until the original train control signal is converted into a random noise signal; the reverse denoising process: gradually denoising the random noise signal to obtain the reconstructed signal.

[0082] In some optional implementations, the reconstruction module 902 includes:

[0083] The first reconstruction unit is used to gradually add a preset noise signal to the original control signal until the original control signal is converted into a random noise signal, and obtain time step information of the original control signal being converted into the random noise signal.

[0084] In some optional implementations, the aforementioned device for suppressing train interference signals further includes:

[0085] The training module is used to input the training noise signal, the random noise signal and the time step information into the diffusion denoising network; the diffusion denoising network obtains the reconstructed signal based on the prediction of the training noise signal, calculates the loss between the reconstructed signal and the original control signal, and adjusts the parameters of the diffusion denoising network; the above process is repeated until the loss reaches a preset threshold, and the diffusion denoising network training is completed.

[0086] In some optional embodiments, the diffusion denoising network includes: a coding layer, an intermediate layer and a decoding layer; wherein the coding layer is used to extract the time domain features of the train control signal; the intermediate layer is used to extract the connection between the deep features of the train control signal and transmit the connection to the decoding layer; the decoding layer is used to perform feature fusion and feature reconstruction based on the time domain features and the connection; there is a jump connection between the coding layer and the intermediate layer, a jump connection between the coding layer and the decoding layer, and a jump connection between the intermediate layer and the decoding layer.

[0087] In some optional embodiments, the coding layer includes: at least two layers of residual connection networks, the residual connection networks include a one-dimensional convolution layer and a linear rectification activation function, time step information is added between the at least two layers of residual connection networks, and the time step information is encoded using sinusoidal position coding; and a downsampling layer, which extracts the signal input to the downsampling layer and transmits the extracted signal to the next level coding layer.

[0088] In some optional implementations, the acquisition module 901 includes:

[0089] The first acquisition unit is used to acquire the time domain signal of each sampling channel of the train; based on the field programmable gate array, the time domain signal is segmented to obtain the train control signal to be processed.

[0090] In some optional implementations, the first acquisition unit includes:

[0091] The first acquisition subunit is used to segment the time domain signal using a preset sliding window, where the preset sliding window is determined based on the length of a single code element of the train control signal to be processed.

[0092] In some optional implementations, the reconstruction module 902 further includes:

[0093] The second reconstruction unit is used to input the train control signal to be processed into a diffusion model, suppress the interference signal in the train control signal to be processed based on the diffusion model, and obtain a reconstructed signal with multi-channel output; based on the preset weight of each channel and each reconstructed signal, obtain a multi-channel reconstruction result of the train control signal to be processed.

[0094] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0095] The device for suppressing train interference signals in this embodiment is presented in the form of a functional unit, where the unit refers to an application specific integrated circuit (ASIC) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0096] The device for suppressing train interference signals provided in the embodiment of the present invention suppresses interference signals in train control signals based on a diffusion model, imitates the diffusion phenomenon in nature, and gradually removes noise signals through a step-by-step diffusion denoising method to achieve reconstruction of the original control signal. It can reduce the impact of interference signals on the original control signal, has strong adaptability, and can adapt to different situations that interfere with the original control signal of the train, and has great practical significance for the safe and efficient operation of the train.

[0097] The embodiment of the present invention also provides a computer device having the above Fig. 9 The device shown is used for suppressing train interference signals.

[0098] See also Fig.10 , Fig.10 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Fig.10 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display graphical information of a graphical user interface on an external input / output device (such as a display device coupled to an interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Fig.10 A processor 10 is taken as an example.

[0099] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0100] The aforementioned memory 20 stores instructions executable by at least one processor 10, so that the aforementioned at least one processor 10 executes the method shown in the above embodiment.

[0101] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0102] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0103] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Fig.10 The example of connecting through bus is taken in the following.

[0104] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator rod, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (such as a light emitting diode) and a tactile feedback device (such as a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0105] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0106] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0107] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for suppressing train interference signals, characterized in that: The method comprises: Obtaining train control signals to be processed; Inputting the train control signal into a diffusion model, suppressing interference signals in the train control signal based on the diffusion model, and obtaining a reconstructed signal; Among them, the diffusion model includes: a forward diffusion process and a reverse denoising process; the forward diffusion process: gradually adding a preset noise signal to the original train control signal until the original train control signal is converted into a random noise signal; the reverse denoising process: gradually denoising the random noise signal to obtain the reconstructed signal.

2. The method according to claim 1, characterized in that The method includes gradually adding a preset noise signal to an original train control signal until the original train control signal is converted into a random noise signal, comprising: The preset noise signal is gradually added to the original control signal until the original control signal is converted into the random noise signal, and the time step number information of the original control signal being converted into the random noise signal and the random noise signal are obtained.

3. The method according to claim 2, characterized in that Also included is that the training of the reverse denoising process is based on the following steps: Inputting the training noise signal, the random noise signal and the time step information into a diffusion denoising network; The diffusion denoising network obtains a reconstructed signal based on the prediction of the training noise signal, calculates the loss between the reconstructed signal and the original control signal, and adjusts the parameters of the diffusion denoising network; The above process is repeated until the loss reaches a preset threshold, and the diffusion denoising network training is completed.

4. The method according to claim 3, characterized in that The diffusion denoising network includes: Encoding layer, intermediate layer and decoding layer; Wherein, the coding layer is used to extract the time domain features of the train control signal; The intermediate layer is used to extract the relationship between the deep features of the train control signal and transmit the relationship to the decoding layer; The decoding layer is used for performing feature fusion and feature reconstruction based on the time domain features and the connection; There is a jump connection between the encoding layer and the intermediate layer, there is a jump connection between the encoding layer and the decoding layer, and there is a jump connection between the intermediate layer and the decoding layer.

5. The method according to claim 4, characterized in that The coding layer comprises: At least two layers of residual connection network, the residual connection network includes a one-dimensional convolution layer and a linear rectification activation function, the time step number information is added between the at least two layers of the residual connection network, and the time step number information is encoded by sinusoidal position; and The down-sampling layer extracts the signal input to the down-sampling layer and transmits the extracted signal to the next encoding layer.

6. The method according to claim 1, characterized in that Acquiring the train control signal to be processed, comprising: Obtain the time domain signal of each sampling channel of the train; Based on a field programmable gate array, the time domain signal is segmented to obtain the train control signal to be processed.

7. The method according to any one of claims 1 to 6, characterized in that Inputting the train control signal to be processed into the diffusion model, suppressing the interference signal in the train control signal to be processed based on the diffusion model, and obtaining a reconstructed signal, including: Inputting the train control signal to be processed into the diffusion model, suppressing interference signals in the train control signal to be processed based on the diffusion model, and obtaining a reconstructed signal output by multiple channels; Based on the preset weights of the channels and the reconstructed signals, a multi-channel reconstruction result of the train control signal to be processed is obtained.

8. A device for suppressing train interference signals, characterized in that: The device comprises: An acquisition module, used for acquiring a train control signal to be processed; A reconstruction module is used to input the train control signal to be processed into a diffusion model, suppress the interference signal in the train control signal based on the diffusion model, and obtain a reconstructed signal; wherein, the diffusion model gradually adds a preset noise signal to the train control signal until the train control signal is converted into a random noise signal, and gradually denoises the random noise signal to obtain the reconstructed signal.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for suppressing train interference signals as described in any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for suppressing train interference signals according to any one of claims 1 to 7.

Images