A Hand Gesture Recognition Model Based on Semi-supervised Learning

Abstract

The traditional vision based hand gesture recognition technology requires a lot of light environment and backgrounds. Focused on these above problems, this paper presents a new hand gesture recognition model, in which, the unsupervised sparse auto-encoder neural network model is applied to train the image patches, in order to extract the edge feature that is the weight, and the pooled features are used as the input of the classifier for classification. The fine turning for the parameter of the entire net is to improve the classification accuracy finally.