Although soft-output sphere decoding algorithm can approach to the bit error performance of Maximum Likelihood（ML）algorithm

it has high computational complexity.This paper presents a complexity-reduced soft-output sphere decoding algorithm

which introduces the reduced noise element to the reverse iterative upper triangular matrix Rof QR decomposition algorithm so as to narrow the initial search radius.Then it uses the minimum distance criterion to effectively trim the search tree and narrow the tree search space.MATLAB simulation results show that this algorithm can reduce the system computational complexity to a great extent while approaching to the bit error performance of traditional soft sphere decoding.