Stationary Subspace Analysis for Video Classification

Mehrtash Harandi (NICTA)

COMPUTER VISION AND ROBOTICS SERIES

DATE: 2013-06-13
TIME: 16:00:00 - 17:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.

ABSTRACT:
While subspace analysis for non-temporal data (e.g. still images) has a long history in machine learning and computer vision communities, similar treatments for temporal data (e.g. Videos) have received comparatively little attention. Through the notion of signal stationarity, in this talk, methods that explicitly take advantage of temporal information will be introduced (hence Stationary Subspace Analysis or SSA). More specifically, given a signal in R^n, I will show how it can be decomposed into d, d< BIO:
Mehrtash is a research fellow at NICTA CVRG and interested in Riemannian geometry, "The Simpsons" and "The Big Bang Theory". He used to be a good chess player!