Abstract: A real-time detection method of working face is urgently needed for coal mine safety, and seismic imaging while mining is an effective way. Based on the analysis of characteristics of seismic signals from coal mining machine, we compared the applicability of three interference methods, and used Bayesian inversion to put forward a tomographic imaging method of in-seam wave while mining in working face. The results showed that using interferometry by deconvolution or by cross-coherence can effectively compress the wavelet of source, and the frequency spectrum are wider than that from cross-correlation; in order to get high-accuracy arrival time and high -accuracy CT, two step interference method is effective. The method can realize real-time imaging the working face, and find the velocity change area thereby warning geological disasters in time.