Space-time codes with controllable ML decoding complexity for any number of transmit antennas

We construct a class of linear space-time block codes for any number of transmit antennas that have controllable ML decoding complexity with a maximum rate of 1 symbol per channel use. The decoding complexity for $M$ transmit antennas can be varied from ML decoding of $2^{\lceil \log_2M \rceil -1}$ symbols together to single symbol ML decoding. For ML decoding of $2^{\lceil \log_2M \rceil - n}$ ($n=1,2,...$) symbols together, a diversity of $\min(M,2^{\lceil \log_2M \rceil-n+1})$ can be achieved. Numerical results show that the performance of the constructed code when $2^{\lceil \log_2M \rceil-1}$ symbols are decoded together is quite close to the performance of ideal rate-1 orthogonal codes (that are non-existent for more than 2 transmit antennas).