Joint channel estimation and data detection using a blind Bayesian decision feedback equaliser

A blind adaptive algorithm for channel equalisation is proposed based on a joint channel estimation and data detection approach. A basic unit of the algorithm consists of a bank of least mean square (LMS) adaptive filters and Bayesian symbol-by-symbol decision feedback equalisers (DFEs). To increase reliability, a variety of initial conditions can be covered by including several such units. The performance of each unit is monitored by examining its estimated mean square error (MSE), and those units which perform poorly can then be switched off. The nature of this blind adaptive algorithm leads to simple and efficient parallel implementation. As is demonstrated in the simulation, convergence can be achieved within a few hundred symbols when a 4-level symbol constellation is used