The Research Of Anti-interference Decoding Algorithm In Cab Signaling Based On Deep Learning

As an important part of Chinese train control system,cab signaling plays an important role in ensuring train operation safety.With the development of high-speed and heavy haul railway,the electromagnetic interference becomes increasingly serious,and the types of locomotives are more diverse.In the specific disturbance environment,the traditional filtering methods cannot filter out the noise in the useful signal band,and the existing cab signaling decoding methods are difficult to meet the demand.In this background,based on the time-domain and frequency-domain characteristics of cab signal and interference coupling mode,combined with deep learning technology,a cab signaling anti-interference decoding method based on optimized denoising convolutional neural network(Dn CNN)and deep belief network(DBN)is proposed.The main contents of this thesis are as follows(1)According to the structure and working principle of cab signaling,the mechanism of cab signaling and the characteristics of ground-onboard information transmission are analyzed;Secondly,considering the disturbance factors and coupling mechanism of cab signal,the coupling ways of unbalanced traction current interference and adjacent line interference are analyzed;through the analysis of the disturbed signals,the anti-interference decoding method of cab signaling is proposed.(2)Based on the characteristics of specific disturbance and FSK(frequency shift keying)signal in time domain and frequency domain,this thesis focuses on the selection of appropriate methods to suppress interference.According to the characteristics of interference,three kinds of noisy signal datasets are constructed.Because the FSK signal as time sequence information has inherent correlation,convolution neural network is used to eliminate the noise and improve the signal to noise ratio(SNR)in an end-to-end manner.To reduce the running time of the neural network on the basis of ensuring its noise reduction performance,the multi-scale depth separable convolution is applied instead of the standard convolution.The simulation results and the actual data verification show that the SNR can be improved up to 15 d B.(3)For the denoised signal,considering the high symmetry of FSK signal spectrum,its feature vector is extracted,and the frequency decoding problem is transformed into a classification problem.The feature of FSK signal is extracted by using deep belief network(DBN),and the decoding framework of DBN is designed and implemented,particle swarm optimization(PSO)algorithm is used to optimize the number of hidden layers and nodes in DBN to raise the accuracy of FSK signal anti-interference decoding.Compared with the original DBN,the optimized DBN improves the anti-interference decoding accuracy of FSK signal.The simulation and experimental results show that the proposed method has strong adaptability,it can extract the characteristic of FSK signal more clearly,and effectively enhance the decoding accuracy of FSK signal.It provides a reference for improving the anti-electromagnetic interference ability of cab signaling under condition of high speed and heavy load,has engineering application value.