Geometric Media Lab
Our approach to media processing and analysis pays attention first to phenomena, and looks for ways to detect, analyze the phenomena using mathematical means yielding the greatest understanding. To push the boundaries of understanding the underlying phenomena of diverse media — images, video, computational sensors, wearables, we draw from mathematical methodologies from statistics, optimization, differential geometry, and topology in a manner that is motivated by physical constraints, invariance requirements, or other phenomenological considerations. Collaborations with researchers from diverse areas including rehabilitation, health promotion and well-being, and media-arts, motivate the development of new algorithmic advances, as well as present challenging use-cases.
Rather than looking for applications of a fixed special computational method, we start with empirical phenomena such as human movement, and find the mathematics / physics, systems techniques most appropriate for that phenomena. This abductive analysis is grounded in two ways: we check our understanding against practical applications; and we create techniques that can be implemented computationally and assessed by criteria like real-time, low latency, bandwidth, etc.