Cues for the perception of expressive speech

Abstract

In a previous study the contribution of fundamental frequency (F0) to the perception of expressive speech was examined using a selective adaptation technique. Listeners heard either F0-present or F0-removed adaptors (vocoder-processed VCV syllables) from one of two expressive categories: angry or happy. In a two-alternative listening task, contrastive aftereffects (characterized by a tendency to label test stimuli as originating from the non-adapted category) were documented only in the F0-present condition. This suggests that F0 is important for the perception of emotional expressions. However, listeners were still able to identify F0-removed syllables as angry or happy at a rate significantly better than chance (58%). Subsequent analyses revealed systematic differences in formant frequencies as potential cues for categorization (higher F1 and F2 frequencies for happy compared to angry tokens). A discriminant analysis was performed using formant measurements (F1-F3) and measures related to the voicing source. Removing voice source measures (including mean F0) produced a decrease in classification accuracy that closely matched listener response patterns for the F0-removed stimuli. The results suggest that in the absence of F0, formant frequencies may be used for perception of angry and happy speech.