Software Tool for Pronunciation Training of Specific English Terminology

Abstract

This paper describes a learning tool developed in MATLAB environment for training English pronunciation of specific terminology, focused on special vocabulary from signal processing and electronics by default. The tool enables to measure three phonetic parameters, namely accent, intonation and voicing. This is done using various computational methods and algorithms including basic filtering, short-time energy, average magnitude difference function or harmonic-to-noise ratio. Spoken words are compared with reference pronunciation in terms of phonetic parameters. Each parameter can be evaluated separately or all parameters together. The learner gets immediate feedback in two forms – percentage correctness of its pronunciation and verbal recommendation at which points to improve its pronunciation, supported by an indicative visual feedback in form of graphs showing each of the phonetic parameters along the spoken word. Two regimes of practices are possible.