Affective classification in video based on semi-supervised learning

  • Authors:

  • Shangfei Wang;Huan Lin;Yongjie Hu

  • Affiliations:

  • School of Computer Science and Technology, University of Science and Technology of China, Anhui, P.R. China;School of Computer Science and Technology, University of Science and Technology of China, Anhui, P.R. China;School of Computer Science and Technology, University of Science and Technology of China, Anhui, P.R. China

  • Venue:
  • ISNN'11 Proceedings of the 8th international conference on Advances in neural networks - Volume Part III
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

In the previous work of affective video analysis, supervised learning methods are frequently used as classifiers. However, labeling abundant examples is time consuming and even impossible for it needs annotation from human beings. While unlabeled video clips are easy to be obtained and they are adequate. In this paper, we present a semisupervised approach to recognize emotions from videos. Firstly, visual and audio features are extracted. Then bivariate correlation is used to select sensitive features. After that, low density separation, a semi-supervised learning algorithm, is adopted as the classifier. The comparative experiments on our own constructed database showed that the semi-supervised algorithm performs better than supervised one, illuminating the effectiveness and feasibility of our approach.