Fuzzy Nearest Prototype Classifier Applied to Spea...
ABSTRACT：In a vector quantisation（VQ）based speaker identification system，a speaker model is created for each speaker from the training speech data by using the k-means clustering algorithm.For an unknown utterance analysed into a sequence of vectors，the nearest prototype classifier is used to identify speaker.To achieve the higher speaker identification accuracy，a fuzzy approach is proposed in this paper.We propose a distortion measure based on the fuzzy k-nearest neighbour rule，which is regarded as an extension of the commonly used distortion based on the nearest neighbour rule.We also propose a fuzzy membership function associated with a sequence of test vectors and use the fuzzy nearest prototype classifier for speaker identification.Experiments show that combining the proposed distortion and the fuzzy nearest prototype classifier gives better results than the nearest prototype classifier.
KEYWORDS：speaker identification；fuzzy nearest prototype classifier；nearest prototype classifier，prototype-based minimum error classifier；distortion measure；vector quantisation；fuzzy c-means clustering；k-means clustering