The Evaluation of Voice Quality via Signal Typing in Voice using Narrowband Spectrograms

Abstract

Traditionally, voice quality is evaluated by auditory-perceptual judgment but objective-acoustic analyses are also valid tools. Spectrography is a part of objective-acoustic analyses and is used since around 50 years. Recently, Sprecher et al. (2010) designed a new classification scheme for signal typing in voice using narrowband spectrograms. The present study aimed to explore the external validation between these 2 methods in a larger dataset. In total 300 voice samples were used based on the vowel [a:], in which n=270 had various degrees of dysphonia and n=30 were vocally normal. All voice samples were evaluated by 7 speech-language therapists using the RBH-scale and 2 judges analyzed the signal typing in voice using narrowband spectrograms. The intra- and inter-rater reliability of all RBH parameters was fair in average (kappa=0.234-0.383) but the rater reliability was moderate between the 2 judges rating signal typing in voice (kappa=0.457). A strong concurrent validity was identified between the signal typing in voice using narrowband spectrograms and auditory-perceptual judgment of hoarseness and breathiness (H: r=0.746, p=0.000; B: r=0.746, p=0.000) but only fair in roughness (r=0.370, p=0.000). The present results confirmed the results of Sprecher et al. (2010) and it might be meaningful to benefit signal typing in voice using narrowband spectrograms on sustained vowels as a valid complement in the evaluation of hoarseness and breathiness for clinical utility and research.