Comparison of PCA, LDA and GDA for palmprint verification

Human Activity Recognition Using Spatiotemporal 3-D Body Joint Features with Hidden Markov Models

“…For video-based HAR, the use of binary silhouettes is considered the most-popular approach [1]- [11]; for instance, in [4], where binary pixel-based mesh features were extracted from every image, the authors used binary silhouettes for HAR. In [5] and [6], the authors adopted Principal Component (PC)-based binary silhouette features to recognize view-invariant human activities.…”

“…Basically, the functionality of an LDA is based on the class information that projects data onto a subspace by using the criterion that tries to maximize the between-class scatterings and minimize the within-class scatterings of the projected data. A generalized discriminant analysis (GDA) [11] that tries to separate the class samples using nonlinear subspace, however, can be preferable to an LDA in terms of its applicability regarding activity features; therefore, a GDA can be considered a robust tool for the classification of human-activity features.…”

“…The recognition of a variety of human activities from video has emerged as a key research area in computer vision, image processing, and human-computer interaction (HCI). Over the last decade, human activity recognition (HAR) systems have therefore attracted a great deal of attention from a community of respected researchers due to their applications in many areas of pattern recognition and computer vision [1]- [11]. The accurate recognition of human activities, however, is still considered a major concern for most of these researchers due to the lack of accuracy that can occur because of a variety of causes such as a failed efficient-feature extraction, a low variance among the features of different activities, and a high variance among the features of the same activity class.…”