Human Activity Recognition in Real-World Video Sequences(Advert Reference: EE/DRFCOM7P/57658)

Human Action Recognition has recently become a very popular topic in computer vision due to its clear practical applications and rapidly growing computing power. It is concerned with accurate classification/regression of human actions to provide a semantically meaningful representation.

Some more commonly proposed uses for action recognition include: surveillance, where action recognition can be used in the identification and prevention of crime and anti-social behaviour; for monitoring in care homes so that residents need not be continually watched by nurses; for content-based retrieval of video documents containing human actions on websites such as YouTube. Many algorithms in action recognition have been proposed in recent years however the current state-of-the-art is far from solving most of these problems. Existing algorithms are based on simple, specific cues, such as abnormal trajectory tracking, and lack the sophistication to identify complex or highly variable activities. Additionally, and in part because of such simple algorithms, most existing systems are not reliable enough for most real-world uses; for example, while we already have algorithms capable of reasonably accurate fall detection, it is unlikely these systems would be trusted in the real world until they can demonstrate a near-human level of accuracy.

This project will explore more advanced machine learning and dimensionality reduction techniques for reliably recognising human actions and activities in challenging real-world video sequences.

Applicants should hold a first or upper second class honours degree (in a relevant subject) from a British higher education institution, or equivalent. Students who are not UK/EU residents are eligible to apply, provided they hold the relevant academic qualifications, together with an IELTS score of at least 6.5.

Informal enquiries regarding this studentship should be made to Professor Ling Shao ([Email Address Removed])

To apply, contact Karen Vacher to request the appropriate application form, quoting the advert reference above, via email to [Email Address Removed] or by using the application link on this page.

Deadline for applications: 25 November 2014
Interview date (if known): 30 November 2014
Start Date: 1 February 2015