Sparse Modeling of Human Actions from Motion Imagery

Abstract

In treating Human Immunodeficiency Virus (HIV) infection, strict adherence to drug therapy is crucial in maintaining a low viral load, but the high dosages required for this often have toxic side effects which make perfect adherence to Antiretroviral Therapy (ART) unsustainable. The imperfect patient adherence to ART and the development of resistant strains in the viral load has led to the development of alternative treatments that incorporate immunological response. This paper investigates theoretically and numerically the effect of immune effectors, such as the cytotoxic lymphocyte (CTL), in modeling HIV pathogenesis; our results suggest the significant impact of the immune response on the control of the virus during primary infection. Qualitative aspects (including positivity, boundedness, stability, uncertainty, and sensitivity analysis) are addressed. Additionally, by introducing drug therapy, we analyze numerically the model to assess the effect of treatment consisting of a combination of several antiretroviral drugs. Nevertheless, even in the presence of drug therapy, ongoing viral replication can lead to the emergence of drug-resistant virus variances. This fact is addressed in our model by including two viral strains, wild-type and drug-resistant. Our results show that the inclusion of the CTL compartment produces a higher rebound for an individual's healthy helper T-cell compartment than does drug therapy alone. Furthermore, we quantitatively characterize successful drugs or drug combination scenarios for both strains of virus.

Discriminative Dictionary Learning With Motion Weber Local Descriptor for Violence Detection

“…In class-specific dictionary learning, a dictionary, in which atoms are predefined to correspond to subject class labels, is learned and thus the class-specific reconstruction error could be used for classification [51], [56], [57], [58], [59], [60]. By adding a discriminative reconstruction penalty term in the KSVD model [50], Mairal et al [58] presented a dictionary learning algorithm for texture segmentation and scene analysis.…”

Action recognition using global spatio-temporal features derived from sparse representations

“…Guo et al proposes a framework based on sparse representation of region covariance descriptors, computed on optical flow features [24] and silhouette tunnels [25]. Castrodad et al [26] proposes a deep-layered discriminative approach for classifying human actions based on learned basis vectors or action primitives for each action category. These approaches try to capture the most non-redundant or informative spatiotemporal features representative of a certain class (top-down).…”

Video-Based Human Behavior Understanding: A Survey

“…Given the redundancy of information collected through spatio-temporal features, Castrodad and Sapiro [96] propose a sparse coding pipeline, by classifying the incoming videos using a dictionary of learned primitives. In spite of the simplicity of the method, the achieved results reach very high performances, scoring 100% on the KTH and UT datasets, and 96% on the UCF-Sports dataset.…”