Analyze Multiple Emotional Expressions in a Sentence

In this paper, we present an emotion recognition approach that can analyze multiple emotional expressions form Mandarin Chinese speech. In this proposed approach, 14 basic and derived emotional categories can be recognized and quantified as the relative components. A large emotional corpus containing 19,136 sentences with 14 emotional expressions were recorded by 34 voice actors. And 979,787 labels of emotional categories were annotated by 37 annotators. We then implemented a Mandarin Chinese speech emotion recognition system with a polygon chart based on WD-KNN classifier. This proposed approach shows how emotions can be recognized by speech signals, and in turn how emotional states can be visualized.