Improved quasi-orthogonal codes

Orthogonal codes are a class of space-time block codes that achieve full diversity, though at a loss of code rate. Such codes have simple single-symbol decoders. The code rate can be made higher by constructing quasi-orthogonal codes, which trade a small loss of performance for full rate coding. Typically, quasi-orthogonal codes perform the best when joint maximum likelihood (ML) decoding is used within subsets of the space-time codewords that are mutually orthogonal. We show that for a particular class of quasi-orthogonal codes, the interference which the subset of the symbols to be jointly decoded sees from other symbols is such that the signal to interference ratio remains constant for all channel realizations. Using the same constellation for all symbols in the subset reduces the minimum distance for such codes and results in a loss of performance. As a remedy to this problem, we introduce a rotation-based method that aims at maximizing the minimum distance in the space-time constellation. Numerical simulations show that significant improvements to previously reported results without rotation can be attained.