Blind Estimation of Linear and Nonlinear Sparse Channels

This paper presents a Clustering Based Blind Channel Estimator for a special case of sparse channels – the zero pad channels. The proposed algorithm uses an unsupervised clustering technique for the estimation of data clusters. Clusters labelling is performed by a Hidden Markov Model of the observation sequence appropriately modified to exploit channel sparsity. The algorithm achieves a substantial complexity reduction compared to the fully evaluated technique. The proposed algorithm is used in conjunction with a Parallel Trellis Viterbi Algorithm for data detection and simulation results show that the overall scheme exhibits the reduced complexity benefits without performance reduction. Keywords—blind estimation and equalization, clustering techniques, sparse zero pad channels.