Previous studies unveiled that huge channel capacity can be achieved by employing multi-element antenna arrays at both transmitter and receiver. Foschini et.al. proposed a vertical Bell Laboratories layered space-time (V-BLAST) system with a detection algorithm based on decision feedback. This system is computationally efficient and yet supporting a high data rate. However, the detection algorithm shows unsatisfactory performance when an equal number of antennas are employed at both ends. In this paper, a parallel detection algorithm is proposed to improve the performance in such scenarios. The detector consists of many low complexity sub-detectors, all of them operating in parallel on a sub-channel matrix. By selecting the submatrix appropriately the overall performance can be improved substantially.