A challenging gait database for office surveillance

Abstract

Gait recognition has become increasingly important in the area of automated human identification. Several databases have been established to enable researchers to evaluate the different covariates affecting the performance of gait recognition. Recently, most of the factors like clothing variation, speed changes, and multiple view angles have been sufficiently addressed. We think it is time to introduce a more challenging database to encourage researchers to look into the harder aspects of gait recognition. The proposed database simulates an office environment as surveillance cameras are often used in offices. The database include difficult conditions like walking with umbrella which portraits a “three-legged” walking pattern, load carriage that occludes the body part, and footwear like high-heeled shoes which alter the way that people walk. We believe the proposed database containing reasonable amount of testing samples will be able to provide an avenue for the research community to advance in the gait recognition field.