Automatic speaker recognition with formant trajectory tracking using CART and neural networks
M.M. Tanabian, P. Tierney, Bahram ZahirAzami · 2006
In this paper we propose a method of speaker recognition based on tracking formant frequency trajectories. We show that the formant frequencies are viable features for clustering speakers into classes for speaker identification. We used two different classification methods and compared the results. We show that neural network classifier produces better performance than the classification and regression trees (CART) method and CART's misclassification rate is higher.