Sparse Representation Denoising of Roadsid Acoustic Signals Based on Wavelet-domain RLS Filters

When the train wayside acoustic signal is used for fault warning and fault diagnosis, Target acoustic signal with fault information and strong background noise are collected together, which affects the subsequent processing. Therefore, for the noise interference problem, this paper proposes a method combining wavelet transform as well as sparse representation, and also creatively adding RLS filter to perform noise reduction on the wayside acoustic signal. In the proposed method. Firstly, the wavelet coefficients after wavelet decomposition are adaptively attenuated using the sparse representation method. Secondly, the RLS filter with the set filtering threshold is then noise reduced. Finally, the signal is reconstructed by two wavelet transforms and inverse transform, which can maximize the sparse signal and recover the signal with higher quality. The simulation results show that this method has better noise reduction effect compared to the traditional wavelet transform and filter noise reduction.