Multi-view Hand Shape Recognition Using Statistical LLE

Abstract

The image-based object recognition problem becomes complicated when the objects of interest are not posed at a fixed view. This study attempts to recognize the multi-view hand shapes in Taiwanese sign language (TSL) based on a novel statistical locally linear embedding (LLE). The original LLE is an unsupervised nonlinear dimensionality reduction approach that utilizes the local linearity to discover the low dimensional manifold embedded in the high dimensional space. This suggests that LLE may preserve neighborhood configuration in the nonlinear structure of the multi-view hand shape data distribution. For better classification performance, this study proposes a statistical LLE that incorporates the class label information statistically to improve the multiclass classification capability posterior to dimensionality reduction. Experimental results show that the statistical LLE gave a classification performance superior to the original LLE and the linear dimensionality reduction methods such as LDA and PCA in multi-view TSL hand shape recognition problem.