In order to reduce more complexity while maintaining a good bit error rate performance, a new SM-SD algorithm is proposed. The new SM-SD algorithm employs a different real-valued equivalent transformation from existing sphere decoding algorithms, and it has a unique search tree structure, the adjacent two layers of the search tree are independent of each other. The principle and process of the new algorithm are analyzed, and the computation complexity of SM-SD algorithms is compared by the matrix analysis. Then, the bit error rate and computational complexity of SM-SD algorithms are compared by simulation in different SM systems. Theoretical analysis and simulation results show that the new SM-SD algorithm has a very close performance to Maximum-Likelihood optimum detection, with lower computational complexity than other existing SM-SD algorithms. Thus, the new SM-SD algorithm is more suitable for SM signals detection.