Approximate Maximum Likelihood Detection of SDMA OFDM over Unknown Time-Varying Frequency-Selective Channels

An approximate maximum-likelihood (ML) multiuser detector for space-division multiple-access (SDMA) orthogonal frequency division multiplexing (OFDM) systems is presented suitable for implementation in systems with a low number of users and multipath components over unknown time-varying, frequency-selective multipath channels. The system employs an initial pilot slot to estimate the different user channels and then, it employs the expectation maximization (EM) algorithm to simultaneously track the channel changes and optimally demodulate the user signals. We show that when the number of subcarriers is quite high (e.g. 256), the system converges to within 0.5 dB of the ML algorithm with a maximum of three iterations in channels with normalized Doppler of 0.02 and with a very slow pilot repetition rate. On the other hand, when the number of subcarriers is low (e.g. 64), then the system performance gets within 2 dB of the ML one with the same channel parameters. In addition, upon introducing oversampling (OS), the performance of the system in the latter case can be significantly enhanced