Introduction: Hypernasality is a perceptual characteristic of speech produced by individuals with resonance disorders. Because of the poor reliability of perceptual judgements of hypernasality, researchers have attempted to develop more reliable estimates by using acoustic measurements. Traditionally, acoustic studies of hypernasality have focused on vowel production. Unfortunately, vowel-derived measures of hypernasality are time-consuming, highly dependent on talker characteristics, and are affected by formant frequencies and phonetic contexts. The aim of this study was to obtain reliable spectral correlates of hypernasality by performing 1/3 octave band analysis on consonants obtained from samples of fluent speech produced by speakers with cleft palate and maxillectomy.
Methods and Results: The speech materials were two alveolar consonants (/s/ and /n/) produced by 14 adults with either cleft palate (12) or maxillectomy (2), and 20 control speakers. Samples of /n/ were produced within single Cantonese words (/man22/ and /min22/, each repeated five times). Samples of /s/ were obtained from five Cantonese words (/sei33/, /sik2/, /si21/, /sik5/, and /sy55/) produced within oral sentences. Energy levels within 1/3 octave bands were calculated using the procedure by Kataoka et al. (2001). Two metrics were calculated for each speaker: i) ‘High5’, representing the difference in level between /s/ and /n/ spectra at high frequencies, and ii) ‘High-Mid3’, representing the difference in energy level between high and mid frequency bands of /s/ spectra. Both measures were found to be significantly different between hypernasal and control groups, and to significantly correlate with perceptual ratings of hypernasality. These spectral measures also showed high intra- and inter-judge reliability.
Discussion: The metrics based on the analysis of /s/ and /n/ spectra were easy to calculate, highly reliable, and suitable for the analysis of fluent speech samples of speakers with cleft palate and maxillectomy. These spectral measurements were also found to be more highly correlated with perceptual judgments than both i) 1/3-octave-band measurements of vowels produced by, and ii) nasalance scores obtained from the same speakers. These results show that spectral measurements of fricative consonants obtained from connected speech samples could be useful for the clinical assessment of resonance disorders.
Kataoka, R., Warren, D.W., Zajac, D.J., Mayo, R., & Lutz, R.W. (2001). The relationship between spectral characteristics and perceived hypernasality in children. Journal of the Acoustical Society of America, 109, 2181–2189.
Kent, R.D., Weismer, G., Kent, J.F., Vorperian, H.K., & Duffy, J.R. (1999). Acoustic studies of dysarthric speech: Methods, progress, and potential. Journal of Communication Disorders, 32, 141–180.