Interactive Displays offer a wealth of opportunities for applications dealing with human computer interaction, personalized content delivery, entertainment, education, customer service and smart spaces. This talk discusses recent research performed at RIT’s Real Time Computer Vision Lab on human action classification, such as face detection, tracking, pose estimation and expression recognition, that can be used to support Interactive Display applications. Manifold learning and Random Projections are dimensionality reduction techniques that are considered for modeling and efficient processing. Manifold learning identifies non-linear structures embedded in the data and may be utilized in a variety of contexts. Random projections allow a data-independent transformation that preserves distances and can be effective for classification and tracking. In addition to generating elegant signal representations, such dimensionality reduction techniques offer computational efficiency that allows potential extensions to mobile, resource constrained systems.