Relations between Prosodic Variables and Emotions in Normal American English Utterances

Abstract

Relations between a speaker's emotional states and some acoustic prosodic variables of his speech were investigated experimentally. The speaker's emotional states, as reflected in 30 utterances, were evaluated by 19 subjects on nine semantic differential scales. The utterances were also evaluated for pitch, loudness, and speed. Durational, fundamental-frequency, and amplitude characteristics of each utterance were measured. Linear correlation coefficients were computed between the acoustic variables and the listeners' judgment of the speaker's emotional states. Significant correlations were found between some of the acoustic variables (e.g., maximum fundamental frequency) and the judgments of some types of emotion. Higher correlations were found between the acoustic variables and judgments of degree of emotion. Correlation coefficients were also computed between judgments of emotion and judgments of prosodic features. Many of these correlations were significant, and higher than the correlations with the acoustic variables. The possibility of regular, but nonlinear, relations between acoustic prosodic variables of speech and perceived emotion was also investigated.