Speaker Recognition based on 80/20 Genetic Algorithm

C.-T. Chen, S.-Y. Lung, C.-F. Yang, and M.-C. Lee (Taiwan)

Keywords

80/20 genetic algorithms, hard-limited Karhunen-Loeve transform, speaker recognition

Abstract

An 80/20 genetic algorithm (GA) as a suboptimal technique is introduced to text-independent speaker recognition. The proposed 80/20 GA scheme attempts to detect the unknown speaker. Three databases, each with 100 Mandarin speakers, are collected for system evaluation. We will investigate the correct rate performance of the 80/20 GA over a pure GA.

Important Links:



Go Back