Efficient Data Compression of Acoustic Signals in Rail Using Sparse Decomposition and Kurtosis-Guided Resampling

Abstract

With the rapid advancement of modern railway technology, remote monitoring of rail safety has become increasingly important. The acoustic signal in the rail has become a key method for remote monitoring due to its long propagation distance and high speed. However, these acoustic signals face the challenge of large data volumes before transmission, necessitating effective compression. In this study, an innovative acoustic signal data dimension reduction method is proposed for acoustic emission signals with periodic pulse characteristics and narrow-band frequency domain features generated by wheel damage. It integrates sparse decomposition and kurtosis-guided resampling to compress these signals. The aim is to reduce the training time and dimensionality of the learning dictionary, thereby achieving sparse representation of the acoustic signal in the rail. In this method, the impact interval is determined using sliding window technology, and the data between adjacent impacts are down-sampled to significantly reduce the amount of signal data while retaining key impact characteristics. Furthermore, a Hankel matrix is used to organize the data after dimensionality reduction to optimize the subsequent sparse decomposition process. Using finite element simulation and experimental verification of service lines, this study systematically discusses the influence of various parameters on sparse decomposition and signal reconstruction. The experimental results show that, compared to the discrete cosine transform, wavelet compression algorithm, and piecewise aggregate approximation method, the proposed method not only retains the impact characteristics of the original acoustic signal but also achieves a higher compression ratio, demonstrating excellent performance and broad engineering application prospects. This study provides a novel and efficient signal processing technology for rail safety monitoring, contributing to the further development of railway safety monitoring technology.