Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Southampton
Gait feature subset selection by mutual information
Significance of output:
This is the first work on feature selection using entropy as a basis for feature set selection, in particular by approximation of the high-dimensional data by means of lower-dimensional mutual information estimates. Entropy is fundamental in information theory but had not previously enjoyed deployment in feature selection – a standard approach in pattern recognition. The paper was invited as a best paper from IEEE BTAS 2007. The paper’s approach has enjoyed wide usage in areas such as vehicle recognition, gait biometrics and sensor selection and has inspired leading researchers in the area (e.g. IEEETSMC(A) 40(3)-2010 http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5395688 ; Entropy-12(10)-2010 http://www.mdpi.com/1099-4300/12/10/2144).
