Blind equalization of fast fading Ricean channels via the EMV algorithm
Hoang T. Nguyen, Bernard C. Levy · 2003
We propose a blind maximum-likelihood equalization method for frequency selective Ricean fading channels with finite coherence time. This method employs the expectation maximization Viterbi algorithm (EMVA) developed for blind channel estimation and signal detection. In our previous work, the Viterbi algorithm (VA) is used to implement the E-phase of an expectation-maximization (EM) iteration for an HMM which describes the noise corrupted output of a deterministic finite state machine (FSM) driven by the transmitted data. To apply the EMVA to Ricean fading channels, we extend the HMM to a class of random FSMs associated with a deterministic finite-state trellis, to which the VA is applicable. As a result, the EMVA can be used to obtain a blind maximum-likelihood estimate of unknown parameters. Simulation results show that the EMVA estimates accurately the channel specular part and exhibits a near-optimum error rate performance, as if all parameters were known.