ELEC 301: Final Project
Text Independent Speaker Recognition

Nils Bagge and Chris Donica
Introduction Introduction

In this project we attempted to identify speakers by comparing a sample of their voice to a database of speakers. There were two phases to out method. In the first phase, we created a "codebook" of the speakers to characterize their vocal characteristics using training sentences. In the second phase, we compared a sample of a speaker's voice to the codebook to determine the identity of the speaker. Our system is at its best roughly 70% accurate in identifing the correct speaker.

The Authors:

Background
Training
System
Architecture
Testing
System
Architecture
Software
Results
Conclusions
Sources