Abstract: The pilot pattern problem for sparse channel estimation in orthogonal frequency division multiplexing (OFDM) systems has been examined. To optimize the performance of sparse channel estimation, an optimized algorithm was proposed. The algorithm was based on minimizing mutual coherence of the measurement matrix in CS theory, to group with increasing the generated random number. The mutual coherence value of each group was calculated, and compared to get the maximum mutual coherence value, consequently to optimize the pilot pattern. Simulation results demonstrate that the pilot pattern obtained using the proposed algorithm can get 30%lower mean square error(MSE)compared with the original pilot pattern.