To make full use of the sparse multipath transmission characteristics of the channel when transform domain communication system is in the high-speed wireless transmission status and improve the channel estimation accuracy of TDCS, a compressed sensing sparse channel estimation method is proposed for TDCS.A pilot pattern is designed for TDCS ,the data frame by it maintains the orthogonality between TDCS signals and the main user meanwhile the measurement matrix has weak mutual coherence , then the channel impulse response is reconstructed by the method of Dantzig Selector. Simulations based on COST207 rural area channel model show that the use of new algorithm can effectively reduce the mean square error of the sparse channel estimation and get 1dB performance gain compared with the least square estimation algorithm when the bit error rate is 0.002.