Aimed at the problems that mine wireless communications are low-accuracy in channel estimation caused by complicated and random characteristics, a multi-user mine communication system assisted by intelligent reflecting surface (IRS) technique is proposed in this paper to improve the accuracy of channel estimation by optimizing the transmission path and reassigning the wireless transmission environment. First, in combination with IRS technique, a model of the IRS-assisted multi-user information transmission in the mine is established. And then, the parallel factor (PARAFAC) decomposition channel estimation algorithm is derived, and the performances of PARAFAC algorithm in the IRS-assisted mine wireless communications is simulated based on the model. The simulation results show that compared with the traditional least-square (LS) algorithm and orthogonal matching pursuit algorithm, the proposed PARAFAC algorithm can obtain about 8 dB signal-to-noise ratio (SNR) gain at the same normalized mean-square error 10-2 while the execution time is slightly less than that of LS algorithm.