Enhancing automatic speaker identification using phoneme clustering and frame based parameter and frame size selection
Jason W. Pelecanos, Stefan Slomka, Sridha Sridharan · 2003
The aim of this study is to investigate variations in speaker Identification performance for the various American English phonemes for different speech parameterisation types and frame sizes. Results on KING wideband data indicate that various phonemes have different optimal parameterisation types and frame sizes. A system able to utilise these alternatives is proposed. The system is shown to improve SI system performance when optimal parameterisation schemes are substituted for each phoneme.