THE APPLICATION OF FUZZY CLUSTERING AFTER IMPROVEMENT ON SPEAKER RECOGNITION

Abstract. This paper presents a new speaker recognition method using the improved initialization of the FCM clustering algorithm and fuzzy vector quantization. It relaxes the FCM membership normalization of conditions. Also, we introduce a new initial clus-tering algorithm, which can have better effectiveness of the cluster in the existence of the opposition point. And this way avoids the result to maintain the local minimum problem. The experimental results show that this new method is feasible and effective