Development of a Text-Dependent Speaker Identification System with the OGI Toolkit
Asterios Toutios, Konstantinos G. Margaritis · 2002
This paper discusses the development a text-dependent speaker identification system. In particular, we develop a neural network, collect the input corpus, train, and measure its efficiency. The development is done with the CSLU-NN environment, a set of programs in Tcl and C, part of the OGI toolkit, originally built for speech recognition tasks, and adapted to our speaker recognition needs. Thus we propose a well-described and quite simple way for building pattern recognizers that deal with .wav files, in general. Our results demonstrate the efficiency of our implementation.