Spatial-temporal feature learning for Human Identification at a distance

Abstract

Gait, as a powerful tool for video Surveillance, has gained increasing attention recently. Comparing with other bio-metric such as the face, iris, fingerprint, gait has its' unique advantage in recognizing people at a distance. Gait recognition is hard to impersonate, and it can be done without the cooperation of the customer. As surveillance cameras are everywhere in life, gait is significant to recognize pedestrians from a video has increasingly demonstrated the importance. This work reviews deep learning-relative works in vision-based gait recognition. Meanwhile, in addition to common issues such as Clothing and View-variance, we also release other new concerns. Lastly, data protection and privacy are discussed.

Keywords: Human Identification at Distance; Smart Surveillance; Gait Recognition; Biometrics Authentication; Deep Learning.