Self-supervised learning for object recognition based on kernel discriminant-EM algorithm

It is often tedious and expensive to label large training data sets for learning-based object recognition systems. This problem could be alleviated by self-supervised learning techniques, which take a hybrid of labeled and unlabeled training data to learn classifiers. Discriminant-EM (D-EM) proposed a framework for such tasks and current D-EM algorithm employed linear discriminant analysis. However, the algorithm is limited by its dependence on linear transformations. This paper extends the linear D-EM to nonlinear kernel algorithm, Kernel D-EM, based on kernel multiple discriminant analysis (KMDA). KMDA provides better ability to simplify the probabilistic structures of data distributions in a discrimination space. We propose two novel data-sampling schemes for efficient training of kernel discriminants. Experimental results show that classifiers using KMDA learning compare with SVM performance on standard benchmark tests, and that Kernel D-EM outperforms a variety of supervised and semi-supervised learning algorithms for a hand-gesture recognition task and fingertip tracking task.