Minimum probability of error demodulation for multipath OFDM-SDMA systems

The two major challenges encountered in the design of high speed wireless networks are the impairments caused by the propagation channel and scarcity of the electromagnetic spectrum. Orthogonal frequency division multiplexing (OFDM) simplifies the former problem by transforming a frequency selective channel into a frequency flat one. On the other hand, spectral efficiency can he achieved by space division multiple access. These two techniques can therefore he combined to achieve reliable wireless communication at high data rates. The resulting system, however, suffers from co-channel interference. In an earlier paper we proposed a class of demodulators that effectively combat this interference. These demodulators are based on the principle of minimizing the error probability in linear demodulation of the received symbols. The demodulators show considerable improvement in BER compared to the MMSE and adaptive RLS demodulation techniques. These demodulators however assume a known channel at the receiver. In this paper we dispense with this assumption by developing training based adaptive channel estimator, which is then coupled with these demodulators. The demodulators still show considerable gain in performance over the MMSE and RLS demodulators even with the estimated channel.