A Hybrid voice identification System with Fuzzy Technique and ART2 Neural Network on BPF Technique

Abstract

Abstract:

In this work, we evaluate the performance of voice identification through the hybrid method using fuzzy and Adaptive Resonance Theory2. The Voice identification is an important task, which shows the active interaction of natural human-machine, for over last important two decades. The objective of this work, it is consists in working out an identification rate of voice identification. The proposed methodology presented allows evaluating the identification process which considers a voice-to-voice matching system using frequency of any words as “Hello”, “Left”, “Right”, “Up”, “Down”, “On”, “Off”, “Turn-on” and “Turn-off”. The above nine words are used for identification. The inputting voice is taken by microphone under the control condition. This approach used the Audacity and PRAAT software for removing the noise and gated the data for calculation the identification. The overall identification rate of all input words is more than 99% and 3% FAR.

Keywords: Voice, FFT, Fuzzy, ART, Signal, Bandwidth.