A hand gestures recognition approach combined attribute bagging with symmetrical uncertainty

In order to improve the performance of traditional hand gestures recognition method, the idea of ensemble learning is adopted in this paper, and a weighted Attribute Bagging method is proposed based on Attribute Bagging (AB) and the concept of symmetrical uncertainty (SU). Firstly, features of the hand gestures are extracted from the preprocessed pictures. Secondly, different classifiers can be trained on a random of attribute subset from the original attribute space, and then symmetrical uncertainty is applied to calculate and represent the relevance of attributes. As a result, weight for each classifier can be determined. In the end, weighted voting is taken and the final result can be gotten. The proposed method can solve the dependency of classifiers training by AB algorithm. The experimental result shows the proposed method is more effective compared with other traditional algorithms.