Abstract: In view of problems of complex underground channel environment and increasing of noise and error rate caused by multipath effect, a system model of orthogonal frequency division multiplexing was established, and the least squares error estimation algorithm and the minimum mean square error estimation algorithm were researched. Based on the rectangular pilot assignment, an estimation algorithm was proposed which uses curve fitting algorithm in the time domain to estimate the data signal channel coefficients. The proposed algorithm does not require high accuracy of the channel coefficients of pilot signals, only requires the constructed function as close as possible to the real value compared with the interpolation algorithm. The simulation results show that the proposed estimation algorithm can greatly reduce implementation complexity, and estimation performance has't big change compared with the minimum mean square error estimation algorithm. Simulation results and theoretical analysis are basically identical.