Improving the performance of speech clustering method

Speech clustering is an unsupervised technique that uses the similarity features of acoustic data for creation of clusters. Traditional clustering methods are used in speech clustering and these methods require the knowledge about the number of clusters. Quality of speech clusters are depending on estimation of prior number of clusters. For this reason, this paper addresses the problem of assessment of number of clusters. The proposed method discovers the effective clusters by known number of clusters. Experimental results are conducted for demonstrating the efficiency of proposed methods using real time datasets.