Maximum likelihood decoding of QAM signals in Markov modulated fading channels

This paper presents a new decoder for quadrature-amplitude-modulated (QAM) signals transmitted through Markov modulated fading channels. Markov modulated channels arise in city environments when a direct line of sight is blocked intermittently, and for large area coverage mobile cells where the channel has large variations in fading statistics. The channel is modelled as a time varying AR process, with coefficients which depend on a Markov chain. An optimal batch algorithm is presented which provides estimates of the channel and the digital signal. This is achieved by optimally coupling a Viterbi algorithm and a Kalman filter. The optimality is demonstrated via the expectation maximization algorithm. In addition, a sub-optimal on-line algorithm has been derived. By considering the channel to be Markov modulated, benefits can be gained over standard techniques which assume fixed fading statistics.