"A Karaoke style feedback system for an isolated digit speech recogniser"

Supervised by Professor PD Green

Abstract

As technology is advancing, it is trying to model computer systems to be more and more similar to human beings. In Robotics robots are made to move like humans; in A.I. computers "think" like humans. It only seems logical that they should also talk and understand speech like hmans. In the 1930's a Princeton researcher created the first speech synthesizer and sparked a whole sequence of events that led to current Hidden Markov Model-based speech recognising systems like the famous IBM ViaVoice.But the use of voice recognition systems has also expanded to other areas - systems control, teaching, medicine; and it is on these last two that this report will focus on. The purpose of this HMM based speech recogniser to provide feedback on pronunciation of isolated words, for users who want to practice their English pronunciation. Automatic Speech Recogniser creation will be discussed in detail along with all the relevant technologies and techniques - the final challenge will be to produce an interface to provide karaoke - style feedback.