Output details
15 - General Engineering
University of Oxford
An extreme function theory for novelty detection
This paper describes a functional theory of extreme value statistics and was published in the leading IEEE journal for Signal Processing. This allows, for the first time, the field of extreme value statistics to be used with probability distributions over functions, rather than over data points. The work demonstrated that abnormal time-series describing human vital-sign data could be identified, and led to a 2012 patent application (GB 1215649.3) currently at PCT stage. This publication provided essential proof-of-concept for securing DA Clifton's five-year Royal Academy of Engineering Research Fellowship, consolidated into a permanent appointment at Oxford.