Abstract: Aiming at the problem that dangerous rock collapse is highly sudden and traditional deformation monitoring is difficult to achieve effective disaster monitoring and early warning, an in-situ microseismic monitoring test was carried out on the collapse process of toppling-type dangerous rocks in karst areas. On the basis of proposing an intelligent noise reduction method for microseismic signals, the microseismic evolution characteristics during the collapse process of toppling-type dangerous rocks were analyzed. The research shows that according to the evolution process of microseismic signals, the collapse process of toppling-type dangerous rocks can be divided into three stages: "initial cracking, fracture expansion, and collapse approaching". There are significant differences in the fracture behavior of the main control discontinuity of dangerous rocks and microseismic signal characteristics in different stages. The microseismic signal characteristics when collapse is approaching mainly include: the microseismic signal waveform shows a period of "quiet period" before becoming active again; the dominant frequency distribution narrows and concentrates in the low-frequency band; low-frequency signal energy becomes dominant, MFCC θ continues to decrease, the b-value drops significantly before rising again; microseismic signal activity declines during a "quiet period" before increasing once more., and an increase in the proportion of shear - type signals. The application of microseismic technology in dangerous rock collapse monitoring is feasible and can provide an effective way for disaster early warning of dangerous rock collapse.