William J. Strong
Dept. of Phys. and Astronomy, Brigham Young Univ., Provo, UT 84602
An extensive set of carefully recorded utterances provided a speech database for investigating acoustic correlates among eight emotional states. Four actors and four actresses simulated the emotional states of anger, joy, neutral, nervousness, hate, fear, sadness, and depression. Many acoustic parameters were extracted from analyses of the 64 (8x8) emotion portrayals. A reduced set of acoustic parameters was obtained by eliminating some highly correlated parameters. The talkers realized ``unique'' emotions, such as anger, with consistent values of the parameters. Different talkers realized ``ambiguous'' emotions, such as neutral and nervousness, with different values of the parameters. Emotion ``pairs,'' such as sadness and depression, had similar parameter values. The acoustic parameters tended to correlate strongly with ``unique'' emotions but less well with ``ambiguous'' emotions.