Comparison of Cantonese word-initial obstruents produced by different types of alaryngeal and laryngeal speakers of Cantonese : a spectral moment analysis

Abstract

This study investigated the use of spectral moment analysis (SMA) in comparing Cantonese word-initial obstruents produced by laryngeal (NL) and alaryngeal speakers: esophageal (SE), tracheoesophageal (TE) and electrolaryngeal (EL) speakers. Four spectral moments were obtained from the speech samples: M1 (centre of gravity), M2 (standard deviation), M3 (skewness) and M4 (kurtosis). A total of 40 native Cantonese laryngeal and alaryngeal speakers read aloud 13 monosyllabic CV Cantonese words formed by a word-initial obstruent /p, pʰ, t, tʰ, k, kʰ, kw, kwʰ, f, s, h, ts, tsʰ/ followed by the vowel /a/. Significant difference in M1 and M3 value was found for EL speakers, as compared to fricatives produced by NL speakers. Also, significant difference in M1 and M3 was found across manners of articulation. However, no spectral moment value differed significantly, when comparing stops with fricatives and affricates produced by EL speakers. In addition, there were significantly difference in M1 and M3 values, when comparing /p/ with /t/ and /k/. Nonetheless, no spectral moment measure could differ /t/ and /k/ significantly for SE and TE speakers. Results showed that M1 and M3 measures could be used to differentiate alaryngeal from laryngeal speakers in the same place of articulation, and distinguish across manner and places of articulation for each speaker group.