Image Set-Based Hand Shape Recognition Using Camera Selection Driven by Multi-class AdaBoosting

We propose a method for image set-based hand shape recognition that uses the multi-class AdaBoost framework. The recognition of hand shape is a difficult problem, as a hand’s appearance depends greatly on view point and individual characteristics. Using multiple images from a video camera or a multiple-camera system is known to be an effective solution to this problem. In our proposed method, a simple linear mutual subspace method is considered as a weak classifier. Finally, strong classifiers are constructed by integrating the weak classifiers. The effectiveness of the proposed method is demonstrated through experiments using a dataset of 27 types of hand shapes. Our method achieves comparable performance to the kernel orthogonal mutual subspace method, but at a smaller computational cost.