Normalizing pitch variability in intoxicated speech

Abstract

Intoxicated speech studies have shown mixed results with respect to prosodic characteristics. Both an overall F0 rise [Hollien et al., J. Acoust. Soc. Am. 110, 3198–3206 (2001)] and F0 lowering [Johnson et al., Phonetica 47, 215–237 (1990)] have been observed under intoxication. These conflicting findings may be due to increased F0 fluctuations at various levels of intoxication and use of inconsistent prosodic domains, such as averaging over entire sentences or utterances. If phonetic analysis takes into account prosodic boundaries, then models will account for appropriate prosodic rises and falls [Taylor, J. Acoust. Soc. Am 107, 1697–1714 (2000)]. This paper reports an analysis of recordings of both intoxicated and control subjects. Speech samples are analyzed at sober and three lev-els of intoxication. Of particular interest are the prosodic patterns at the often unreported descending limb of intoxication, which displays the widest range of variability in behavior. Use of F0 parameter extraction, stylization, and rotation methods over appropriate prosodic domains factors out spurious variation at different intoxication levels. Evaluation of this normalization procedure is reported with respect to both within- and across-subject variation. Implications for the analysis of normal, intoxicated, and disordered speech are discussed.