Comparison of methods for determining speech voicing based on tests performed on paired consonants and continuous speech

Abstract Voicing is an important phonetic characteristic of speech. Each phoneme belongs to a group either voiced or unvoiced sounds. We investigated and compared the performance five algorithms widely used estimate speech voicing. All were implemented in Matlab tested on both short consonants continuous Phonetically paired (voiced vs unvoiced) parts read from audio books experiments. The tuned harmonics-to-noise ratio method gave best results situations, ie for Using this method, overall voicing Czech, Polish, Hungarian English was investigated. showed highest proportion parts, approx. 75 %. In other languages, around 70