Frame difference history image for gait recognition

In this paper, we propose a simple but effective human identification method based on gait features using frame difference history image (FDHI). Before constructing the FDHI feature, a sequence-based silhouette normalization scheme and an alignment pre-processing step are applied. After that, a post-processing step is devised for getting more representative gait signatures for human identification. Two types of FDHI templates are then extracted and represented more compactly by a dimensionality reduction technique, i.e., Principal Component Analysis (PCA) followed by Linear Discriminant Analysis (LDA). The transformed feature vectors are then respectively classified by individual K-Nearest Neighbor (KNN) classifiers. Lastly, the final classification decision is made by a fusion technique. Experimental results are provided to prove the superiority of the proposed method.