Abstract: With the gradual application of full-face tunnel boring machines (TBMs) to the excavation of rock roadways in coal mines, there is an increasingly urgent need for the accurate and rapid advance prediction of unfavorable geological structures. This study analyzed the characteristics of active-source seismic wave-based advance detection methods and the applicability of the advance detection technology based on TBM seismic while drilling (TSWD) with TBM rockbreaking vibration as a seismic source. Then, by combining the geological and production conditions of coal mine roadways, this study proposed a horizontal seismic profiling (HSP)-based advance detection method applicable to the excavation of roadways in coal mines using TBMs. Then, the detection instrument with designed integrated explosionproof hardware was applied to a TBM-excavated bottom drainage roadway in the Shoushan No.1 Mine, Pingdingshan City, Henan Province. This study detected near-horizontal thin coal seams in the roadway by constructing a spatial observation method and optimized the array arrangement parameters of geophones in the narrow space of a double shield TBM-excavated roadway. Finally, this study processed the original signals and obtained images of detection results through time-frequency analysis, cross-correlation interferometric processing, and the joint inversion of reflection and scattering data. The results show that the spatial observation method can identify thin coal seams that obliquely intersect with the roadway at low angles, with the optimal identification results obtained when the distance between the seismic source and the first geophone was designed at 15 m. Unfavorable geological structures in surrounding rocks can be accurately inferred by extracting effective signals through time-frequency analysis, determining virtual seismic source channels and reflection characteristic curves through cross-correlation interferometry, and plotting the stratigraphic reflection energy distribution maps of the detected area through imaging based on the joint inversion of reflection and scattering data. The results revealed by in-situ excavation were highly consistent with the detection results, indicating that the HSP-based advance detection method can achieve non-destructive advance prediction of geological conditions within a range of 100 m in front of the tunneling face. Therefore, this method assists in increasing TBMs' tunneling speed for rock roadways in coal mines.