Acoustic Variation in Child Speakers across Gender, Age and Ethnicity
Faculty Mentor(s)
Dr. Sonja Trent-Brown, Hope College
Document Type
Poster
Event Date
4-13-2012
Abstract
Listeners can identify adult speakers with respect to gender and ethnicity by attending to perceptual cues in the acoustic signal (Thomas & Reaser, 2004, Trent-Brown, 2004). This study explores the age at which these acoustic parameters begin to vary. Hillenbrand et al. (1995) conducted an analysis of the American English vowel space and published target acoustic descriptions for men, women, and children. Results were presented for "children" (ages 10-12), across gender, with no mention of ethnicity. Although the adult voice is typically achieved at age 14 for girls and age 15 for boys (Berger, 2008), there’s a gradual lowering of the voice beginning with the onset of puberty as early as age 10 (Teen Growth, 2000), indicated by lower fundamental frequency (F0). Studies have provided evidence of both perceptual and acoustic variation with respect to speaker ethnicity (Trent-Brown, et al., 2009). Child speakers were recorded producing neutral context (/h-vowel-d/) words and sentences containing those target words. Speaker productions were screened for dialects. Acoustic variation was measured in terms of differences in temporal and spectral acoustic features such as vowel duration, fundamental frequency, and formant resonance frequencies. A multivariate analysis of variance showed significant differences for both age and gender. For gender, there was a significant fundamental frequency variation, with higher values for girls than for boys. There was also a significant difference for F0 with respect to age, such that as age increased, fundamental frequency decreased. These findings are in the expected direction, mirroring adult patterns, suggesting that gender and age are acoustically important considerations when conducting research with child speakers. Analysis of the data exploring variation across ethnicity is currently underway. Implications are evident for acoustic analysis of child corpora, for awareness of linguistic profiling (Baugh, 2005) and for educating teachers of diverse students.
Recommended Citation
A recommended citation will become available once a downloadable file has been added to this entry.
