Abstract
This paper proposes a bit-selective iterative processing scheme for turbo equalization. We consider convolutionally coded and turbo coded transmissions over inter-symbol interference (ISI) channels. The receiver starts with conventional iterative channel equalization and decoding. After each iteration, bit convergence status is measured via cross-entropy. Short windows are then applied around unconverged bits and the subsequent detection/decoding iterates only within the windows. The rapid decrease in selected bits after each such iteration reduces computational complexity significantly, but performance is well maintained by utilizing the Markov property in the trellis detection/decoding.