Cross-gender acoustic differences in hypernasal speech and detection of hypernasality

Abstract

Spectrum of hypernasal speech produced by cleft palate (CP) speakers carries acoustic information regarding hypernasality (HP). This study compares the variation of acoustic features of HP in continuous read speech with gradually increasing severity of HP within and across gender and detect HP. Three point vowels /i/, /a/ and /u/ from continuous read speech produced by seven male and seven female speakers are used. In first part of this study, comparative study of variation in acoustic parameters within and across vowel category of males and females are made. Experimental results show considerable variability in spectral features among CP subjects. High back vowel /u/ shows greatest variability for both males and females. Mid low vowel /a/ shows least variation among the vowels for females and high front vowel /i/ shows least variation for males. The inter-speaker variability measurement suggests that high back vowel /u/ is mostly affected and has the highest variability. In the second part, ratio of vowel space area (VSA) of hypernasal and normal speech is used as a threshold for HP detection. Results show that VSA spanned by CP males is shrunken by 3.7 times than CP females. This indicates that VSA spanned by CP females is larger compared to CP males.