DefinePK

Abstract

The aim of this paper is to investigate the viability to recognise an individual by gait where this individual is having a twin sibling where he/she is sharing similar facial features, hence may render a challenge to face recognition systems. To facilitate the investigation, supervised classification experiments are devised to compare the gait of both twins and non-twins siblings and additionally compare both of these with the inclusion of an early SOTON data set to synthesise a small population data set. An important component of this investigation is to firstly develop data sets consisting of twins and non-twins siblings, which are video-based data sets. After extraction of lower limbs kinematics signals from the videos, to learn the recognition viability performance, this paper performs classification with a leave-one-out cross validation on the data sets. The best correct classification rates using the proposed descriptor is 95%, 76%, 82%, and 74% respectively for data sets containing of only non-twins siblings, only twins, both non-twins and twins, and both non-twins and twins including SOTON data set.