Emotion Recognition of a Speaker Using Facial Expression Intensity of Thermal Image and Utterance Time

Abstract

Herein, we propose a method for recognizing human emotions that utilizes the standardized mean value of facial expression intensity obtained from a thermal image and the standardized mean value of the time at utterance. In this study, the emotions of one subject could be discerned with 76.5% accuracy when speaking 23 kinds of utterances while intentionally displaying the five emotions of "anger," "happiness," "neutrality," "sadness," and "surprise."

Human–Computer Communication Using Recognition and Synthesis of Facial Expression

“…Sensor fusion is a promising way to improve the recognition accuracy of facial expression or emotion recognition. Several studies [2,3] to improve accuracy using sensor fusion have produced promising results.…”