Track Circuit Signal Denoising Method Based on Q-Learning Algorithm.

This paper presents the track circuit signal acquisition procedure and the structure of the receiving antenna. A new track circuit signal noise extraction method by adding a center tap of the receiving antenna is proposed. Furthermore, the Q-learning algorithm is applied to process the track circuit signal based on the extracted noise. By employing the Q-learning algorithm, the strategy of the denoising system can be adjusted continuously with the change of environmental noise. Our method automatically learns different types of noise during normal operation. Experimental results indicate the signal-to-noise ratio (SNR) of signal could be improved by 15-30 dB.Compared with the traditional filtering method, this denoising method has a better effect on the noise in the signal working band.