The conventional channel estimation algorithm used in the current TD-SCDMA system can not be expanded to multi-cell joint detection for its low channel estimation accuracy for adjacent cells. Firstly the conventional channel esti- mator was introduced and the reasons for the performance degradation when extended to multi-cell channel estimation were analyzed. Then three high accuracy multi-cell channel estimation algorithms were proposed for multi-cell joint de- tection by changing the structure of the channel estimation matrix

based on matched filter （MF）

minimum mean square error （MMSE）

and a combination of both. Simulation results prove that all of the three newly proposed algorithms have higher accuracy than conventional Steiner channel estimation and thus can improve the performance of the multi-cell joint detection of TD-SCDMA system.