Efficient ML Decoder for Quasi-orthogonal Space-Time Codes When Using QAM Constellation

The most attractive class of space-time codes comprises the full-rate orthogonal codes, such as the Alamouti code. These codes allow a simple individual decoding of each symbol of the space-time block, leading to a reduced complexity at the receiver. They exist however only for two transmit antennas. If one intends to increase the transmit diversity order, then it may either use a sub-unitary rate orthogonal code or use a full-rate quasi-orthogonal code in which the decoding is done in pairs of symbols. Unfortunately, the pair-wise decoding leads to a large search space of the Maximum Likelihood (ML) decoder, especially for high-rate constellations, such as 16-QAM or 64- QAM. This paper presents a less complex general ML decoding algorithm, with linear pre-processing phase, applicable for a class of quasi-orthogonal space-time codes with rectangular constellations (e.g. QPSK, 16-QAM, 64-QAM).