This paper presents a signal de-multiplexing scheme in branch metric calculation of soft decision decoding for a spatially multiplexed MIMO system. In a conventional MIMO system, de-multiplexing is carried out over the symbols received by multiple antenna elements and is separated from decoding. Therefore, the de-multiplexing requires multiple uncorrelated antenna elements in order to realize a full rank channel matrix. This leads to the limitation of the system capacity by the number of the receive antenna elements. Instead of splitting de-multiplexing and decoding, in this paper, a de-multiplexing scheme at branch metric calculation in soft decision decoding is proposed. Based on ideal interleaving, independence among the coded symbols is assumed. Thus, the full rank channel matrix with the size of minimum free distance can be realized for signal de-multiplexing. As examples of the proposed system, the performance of repetition codes, block codes, and convolutional codes with MMSE de-multiplexing and soft decision Viterbi decoding on a Rayleigh fading channel is investigated. It is shown through numerical analysis that a lower BER can be achieved with a larger minimum free distance for the same normalized transmission rate that is given as the product of the coding rate and the number of multiplexed signal streams.