Intraspeaker/interspeaker variation in selected voice parameters and speaker identification.
Al Yonovitz, Angela Smith · The Journal of the Acoustical Society of America · 1991
Female speakers provided voice exemplars in both a speaking task (reading) and a sustained phonation into a reflectionless tube. Nine serial trials, spaced three days apart were obtained from each subject. The fundamental frequency and coefficient of variation were obtained from the speaking task (32.768 s). Jitter, shimmer, and spectral function were obtained from the sustained phonation and was based on the analysis of 100 consecutive cycles of the glottal wave. These data comprised the ‘‘known’’ exemplars. A tenth, ‘‘unknown’’ exemplar was also obtained. Multivariate linear discriminant analysis was utilized to determine the classification of both the known and unknown exemplars. The analysis correctly classified 69.1% of the known exemplars and 66.7% of the unknown. Speaking fundamental frequency provided the greatest discrimination and spectral function the least.