Multi-modal Emotion Recognition Based on Speech and Image

For the past two decades emotion recognition has gained great attention because of huge potential in many applications. Most works in this field try to recognize emotion from single modal such as image or speech. Recently, there are some studies investigating emotion recognition from multi-modal, i.e., speech and image. The information fusion strategy is a key point for multi-modal emotion recognition, which can be grouped into two main categories: feature level fusion and decision level fusion. This paper explores the emotion recognition from multi-modal, i.e., speech and image. We make a systemic and detailed comparison among several feature level fusion methods and decision level fusion methods such as PCA based feature fusion, LDA based feature fusion, product rule based decision fusion, mean rule based decision fusion and so on. We test all the compared methods on the Surrey Audio-Visual Expressed Emotion (SAVEE) Database. The experimental results demonstrate that emotion recognition based on fusion of speech and image achieved high recognition accuracy than emotion recognition from single modal, and also the decision level fusion methods show superior to feature level fusion methods in this work.