Abstract: Mine seismic while headinguses the vibration generated by the rock cutting of roadheader as the seismic source for detection, which has the advantages of safety, simplicity and efficiency. However, the seismic records collected by mine seismic while heading are complex in wave composition, weak in effective signal energy, and strong noise interference generated by roadheader. The signal and noise ratio isextremely low, and the denoising is difficult. It is necessary to study the denoising method for mine seismic records obtained while heading. By improving the search strategy of artificial bee colony (IABC) algorithm, the convergence speed and accuracy of artificial bee colony algorithm are improved. The improved artificial bee colony algorithm is used as the optimization algorithm of independent component analysis (ICA) method to improve the signal and noise separation effect of independent component analysis. The geological-geophysical model of the advance detection of the excavation earthquake is established, and the noised variable frequency amplitude continuous signal is given as the source wavelet of the mine seismic while heading. The finite difference method is used for numerical simulation to obtain the seismic records of themine seismic while heading, and the random noise is added into the simulation seismic records. The independent component analysis method based on improved artificial bee colony (IABC-ICA) is used to separate the noise and effective signal. On this basis, the cross-correlation and pulse deconvolution methods are used to process the compressed wavelet, and the wave field characteristics consistent with the seismic records simulated by the Ricker wavelet source are obtained, which verifies the denoising effect of the method. Combined with the field test data, the IABC-ICA denoising, cross correlation, deconvolution and other processing are carried out to extract the effective wave signal from the actual seismic records with low signal and noise ratio. The study shows that the independent component analysis method based on improved artificial bee colony can better separate the effective signal and interference noise signal in the recordswith mine seismic while heading, and improve the extraction quality of effective signal.