Automatic Emotion Recognition by the Speech Signal

This paper dis cusses approaches to recognize the emotional user state by analyzing spoken utterances on both, the semantic and the signal level. We classify seven emotions: joy, anger, irritation, fear, disgust, sadness and neutral inner state. The introduced methods analyze the wording, the degree of verbosity, the temporal intention rate as well as the history of user utterances. As prosodic features duration, pitch and energy contribute to a robust recognition. Further more the problem of spotting for emotional phrases in the human-computer-interaction is alluded. User profiling supports the adaptation of different cultural comprehensions of verbally expressed emotions. To legitimate the applied features results of usability studies are introduced. Finally fields of application are shown and results are discussed.