Exploration of perceptually-based features for speaker recognition: An empirical study
Lists
Only you see this
Abstract
Perceptually-based features have been extensively used in speech recognition, but lack intensive study in speaker recognition. In this paper, we empirically explore perceptually-based features for speaker recognition. For the perceptual linear predictive (PLP) analysis[1], we use the adaptive component weighting (ACW) technique[4] to improve its performance for speaker recognition. The F-ratio metric and the DTW based text-dependent speaker identification method are employed to evaluate the effectiveness of the perceptually-based features in speaker recognition. Based on the experimental results, we demonstrate the effectiveness of the ACW based PLP analysis and report the optimal orders of the four features used in speaker recognition.<br/><br/>
BibTeX
@inproceedings{xiang1996exploration,
author = {Beichen Xiang and Ke Chen and Hong Chi},
title = {{Exploration of perceptually-based features for speaker recognition: An empirical study}},
booktitle = {Research},
year = {1996},
}Loading related content...