Soft-output detector for channels with intersymbol interference and Markov noise memory

The paper presents the forward-backward soft-output detector for channels with intersymbol interference and correlated (and/or signal-dependent) channel noise. We utilize the Markov noise assumption to keep the detector complexity in check. Particular attention is given to channels with Gauss-Markov noise statistics. We demonstrate that the detector can be built with the same hardware used for a maximum likelihood sequence detector (Viterbi detector), i.e., using a bank of finite impulse response (FIR) filters. Possible applications of the detector are in iterative decoding applications (e.g., turbo coding).