Data Clustering by Minimum Difference Tree and PRI Transform

Data clustering is an important problem in communication systems. In this paper a new algorithm for data clustering in a communication system is presented. The algorithm combines two techniques, the minimum difference tree and the PRI transform, to cluster the interleaved pulse trains. Since the method uses matrix and recursive computation, it has inherently the ability of parallel execution. To parallelize the algorithm a systolic array, the best parallel structure for matrix and recursive operation, is designed and the improvement in the total execution time is discussed.