Abstract: In the study, a theoretical maritime multipath channel model was constructed; the model was suitable for the different sea states and frequencies for maritime ship communications. Aimed at SC-FDE (Single Carrier-Frequency Domain Equalization) system over maritime multipath channel, a method for sparse channel estimation based on compressed sensing was proposed. The Zadoff-Chu sequence was used as the pilot sequence to design a circulate Toeplitz observation matrix, which was used as measure matrix of compressed sensing. Based on Sparsity Adaptive Matching Pursuit algorithm, T-SAMP algorithm was proposed. The NMSE (Normalized Mean Square Error) and BER (Bit Error Rate) of T-SAMP, OMP (Orthogonal Matching Pursuit), and LS (Least Squares) algorithms were compared. The simulation results demonstrated that T-SAMP algorithm is more practical than OMP algorithm because of its capability of signal reconstruction without prior information of the sparsity, and the proposed T-SAMP algorithm achieves better performance than the traditional LS algorithm with reduced length of pilot sequence.