Abstract: Coal-rock interface recognition has always been a worldwide difficulty in mining.The echo reflected from coal-rock interface has the problems of low SNR, mode mixing, and being hard to obtain the accurate waveform of the echo from the coal-rock interface.Therefore, it is important to extract the useful component from the original echo.Variational Mode Decomposition (VMD) has been successfully applied to extract the useful component from original signal.However, it relies much on the parameter selection.Inappropriate parameters will severely affect the performance of VMD.Considering that the real-time processing of ultrasonic echo is required in underground mine sites, VMD doesn't meet the requirement.Therefore, Empirical Wavelet Transform (EWT) is proposed to extract the useful echo signal from coal-rock interface.Based on the simulation and experiment data, as an adaptive signal processing method, EWT achieves adaptive decomposition and accurately extracts useful component without any input parameters, which satisfies the requirements of real-time processing and imaging of echo signal.Based on the simulation and experiment data, EWT can effectively extract the useful echo signal.In addition, concerning that the processed echo signal has the problem of wavelet distortion, an optimal iterative method with weighted coefficients is proposed to fit the echo envelope based on asymmetric Gaussian model.Finally, a peak focalization imaging method is proposed to make image with the fitted echo envelope.Processed results indicate that this method can achieve the goal of recognizing the coal-rock interface accurately.