Kohonen Maps Based Near Optimum Sequence Detection

This communication presents a near optimum sequence detector based on machine learning, i.a. artificial intelligence, capable of iterative operation. To the best of the authors' knowledge, this is the first maximum a-posteriori probability based detector relying solely and exclusively on Kohonen maps. The online training of this Kohonen detector is facilitated by deploying a tailored probability based neighborhood function. The proposed Kohonen detector is tailored to the generation of soft outputs using approximate log-likelihood ratios. Selected simulation results demonstrate the promising performance capability of the proposed Kohonen detector.