Abstract
Over the past few years, artificial neural networks (ANNs) based on learning vector quantisation (LVQ) algorithms have received considerable attention as pattern classifiers. LVQ2 is currently the preferred choice for automatic speaker identification (ASI) applications. The paper investigates 76 ANN ASI learning schedules incorporating the main LVQ variants, for a 21 speaker text independent database. It concludes that a one stage schedule based on LVQ1 with weak or no repulsion is at least as efficient as more complex LVQ2 schedules.

This publication has 10 references indexed in Scilit: