A Speaker Independent Arabic Isolated Spoken Digits Recognition System Using Fuzzy Kohonen Clustering Network

Abstract

A Fuzzy Kohonen Network, which is capable of recognizing isolated Arabic spoken number speaker independently is described. The Fuzzy Kohonen Clustering Network (FKCN) algorithm, is based on the integration of Fuzzy C-Means (FCM) and Kohonen Clustering Network (KCN). FKCN is unsupervised, non-sequential, and uses fuzzy membership values from FCM as learning rates. Simulation results clearly indicate the superiority in recognition accuracy performance of FKCN when compared to that obtained for FCM, KCN and the conventional LBG (Linde-Buzo-Gray).