A Pilot Study for Algorithmic Diction Detection for Use by Singers and Vocal Teachers

Abstract

This paper introduces an algorithmic signal processing method to quantify vocal dic-tion using audio files that can potentially assist singers and teachers. Clear diction and pronunciation in singing is important for a variety of reasons and should be ex-ercised alongside the development of voice. In order to convey a clear verbal mes-sage, strong diction is needed. To accomplish this goal of diction detection, the in-terpretation of the consonants is of prime significance. The proposed algorithm works with features such as zero crossing rate, spectral spread, spectral flux and spectral centroid. In this paper, we offer a proposed framework and algorithm of dic-tion detection using modern applicable audio features and extraction techniques. Fu-ture approach for analysis of diction is also defined.