针对钢轨踏面电磁超声检测信号幅值小、信噪比低的问题，研究适用于电磁超声的数字信号处理方法，利用小波变换对信号进行消噪处理，通过希尔伯特变换提取信号包络，并对处理后的数据进行超声成像。采用LabWindows/CVI和Matlab混合编程的方式设计上位机数据分析软件。试验表明，该方法显著提高了电磁超声回波信号的信噪比，能有效地检测钢轨踏面缺陷并进行定位。

To increase the amplitude and the SNR(Signal Noise Ratio) of electromagnetic acoustic signals in rail surface detection, proper signal processing methods were studied. Wavelet transform was used to denoise the signals and the signal envelopes were extracted through Hilbert transform algorithm. The ultrasonic images of the rail surface were presented, making the detection results more clearly. A hybrid programming language between LabWindows/CVI and Matlab was used to develop the software. Experiments indicated that the methods could improve the SNR of electromagnetic acoustic signals significantly, thus leading to the effective detection and location of rail surface defects.