Subjective Evaluation of a Speech Emotion Recognition Interaction Framework

Abstract

In the current work, a conducted subjective evaluation of three basic components of a framework for applied Speech Emotion Recognition (SER) for theatrical performance and social media communication and interaction is presented. The multidisciplinary survey group used for the evaluation is consisted of participants with Theatrical and Performance Arts background, as well as Journalism and Mass Communications Studies. Initially, a publically available database of emotional speech utterances, Acted Emotional Speech Dynamic Database (AESDD) is evaluated. We examine the degree of agreement between the perceived emotion by the participants and the intended expressed emotion in the AESDD recordings. Furthermore, the participants are asked to choose between different coloured lighting of certain scenes captured on video. Correlations between the emotional content of the scenes and selected colors are observed and discussed. Finally, a prototype application for SER and multimodal speech emotion data gathering is evaluated in terms of Usefulness, Ease of Use, Ease of Learning and Satisfaction.