A novel cluster-based maximum likelihood blind equalization of ISI impaired channels

A novel clustering-based blind channel equalizer suitable for both linear and nonlinear channels is proposed. The clusters formed by the received data are identified using a new class of unsupervised clustering algorithms known as K-Harmonic Means (KHMp). The KHMp algorithms are insensitive to the initialization of the cluster centers owing to a built-in boosting function, resulting in better performance over algorithms used in the past like ISODAT A. The identified cluster representatives are then mapped to the input signal vectors using a discrete Hidden Markov Model and the mapping is used to compute the branch metrics in a cluster-based maximum likelihood sequence estimator (MLSE) to perform signal detection. Computer simulations showing the equalizer performance with the new clustering algorithm are presented.