中文摘要：

传统双模式算法中的切换时机一般采用经验确定,在迭代达到一定次数或者均方误差（MSE）降低到某一范围时硬性将算法进行切换。针对这种情况,提出了一种基于聚类技术的软判决双模式均衡算法,在不影响算法的收敛速度和精度前提下,让算法间自动进行切换,更具有实际意义和价值。该算法首先通过分析初始均衡算法输出的星座图信息,再使用减法聚类获得粗略的星座图轮廓,最后采用模糊C-均值（FCM）聚类进行二次处理,以获得精准的星座图信息。若所得星座图符合判断标准则切换至后续算法完成均衡,实现了算法中的软切换。仿真结果验证了该算法的有效性。

英文摘要：

Switching time of traditional dual-mode algorithms generally determined empirically, which is that when through a certain number of iterations or mean square error is reduced to a range, algorithm gets its hard switching. For this situation, this paper proposed a clustering technology based soft decision dual-mode constant modulus blind equalization algorithm. New algorithm was more practical because that it can switch automatically without affecting the convergence speed and accuracy. It firstly analysed the information of initial outputs of constellation equalization algorithm, and then obtained a rough outline of the constellation diagram according to subtractive clustering. Finally it used fuzzy C-means （FCM） clustering for secondary processing to obtain accurate information on the constellation. If the resulting constellation meets criteria, equalizer will switch to the subsequent algorithm to complete equalization, which achieved a soft switching in the algorithm. Simulation results show the effectiveness of the algorithm.