PROJECT TITLE :

Silhouette Analysis for Human Action Recognition Based on Supervised Temporal t-SNE and Incremental Learning - 2015

ABSTRACT:

This paper develops a person's action recognition method for human silhouette sequences based on supervised temporal t-stochastic neighbor embedding (ST-tSNE) and incremental learning. Inspired by the SNE and its variants, ST-tSNE is proposed to learn the underlying relationship between action frames in a very manifold, where the class label information and temporal info are introduced to well represent those frames from the identical action class. As to the incremental learning, an vital step for action recognition, we have a tendency to introduce three strategies to perform the low-dimensional embedding of new information. 2 of them are motivated by native methods, domestically linear embedding and locality preserving projection. Those two techniques are proposed to learn specific linear representations following the native neighbor relationship, and their effectiveness is investigated for preserving the intrinsic action structure. The rest one is predicated on manifold-oriented stochastic neighbor projection to search out a linear projection from high-dimensional to low-dimensional area capturing the underlying pattern manifold. Extensive experimental results and comparisons with the state-of-the-art methods demonstrate the effectiveness and robustness of the proposed ST-tSNE and incremental learning strategies in the human action silhouette analysis.