中文摘要：

针对K-best检测算法易将最优路径舍去的特点和K-best检测算法搜索星座图中所有点的特点，提出一种性能改进型K-best检测算法和几种降低复杂度K-best检测算法．性能改进型K-best检测算法在进行QR分解之前对信道矩阵进行最小均方误差（MMSE）滤波，能有效减小最优路径被舍弃的概率，提高算法性能；降低复杂度K-best检测算法采用类似球形译码检测的方法减少搜索星座图中点的个数．仿真结果显示，性能改进型K-best检测算法比基于排序QR分解（SQRD）的K-best检Sjn,4算法有1dB的性能增益．降低复杂度K-best检测算法在K=4时有性能损失；当K=8时，降低复杂度K-best检测算法和原K-best检测算法有同样的性能，同时前者比后者需要更少的计算量．

英文摘要：

Aiming at the optimum path excluding characteristics and the full constellation searching characteristics of the K-best detection algorithm, an improved-performance K-best detection algorithm and several reduced-complexity K-best detection algorithms are proposed. The improved-performance K-best detection algorithm deploys minimum mean square error （MMSE） filtering of a channel matrix before QR decomposition. This algorithm can decrease the probability of excluding the optimum path and achieve better performance. The reducedcomplexity K-best detection algorithms utilize a sphere decoding method to reduce searching constellation points. Simulation results show that the improved performance K-best detection algorithm obtains a 1 dB performance gain compared to the K- best detection algorithm based on sorted QR decomposition （SQRD）. Performance loss occurs when K = 4 in reduced complexity K-best detection algorithms. When K = 8, the reduced complexity K-best detection algorithms require less computational effort compared with traditional K-best detection algorithms and achieve the same performance.