Bayesian blind turbo equaliser with reduced complexity based on channel estimate monitoring

The authors recall the application of fixed-lag particle smoothing (FLPS) to perform blind turbo equalisation under a Bayesian framework, assuming that the noise variance is unknown. This approach has been successfully applied in previous works, at the price of increasing equaliser computational complexity. This study´s main contribution is to propose a strategy to monitor, along the iterations, the quality of channel estimate and soft information about the coded symbols, to switch from FLPS to the BCJR algorithm when performing soft-input soft-output equalisation. This approach allows saving significant computational effort without affecting bit error rate performance. Simulation results and a careful computational complexity analysis show the effectiveness of the techniques herein presented.