A Regression Approach for Robust Gait Periodicity Detection with Deep Convolutional Networks
This paper presents a regression approach to gait periodicity detection via fitting gait sequence to a sine function by deep convolutional neural networks. The key idea is to model the gait fluctuation as a sinusoidal function because of similar periodic regularity. Each frame of the gait video corresponds to a function value that can represent its periodic features. Convolutional network serves to learn and locate a frame in a gait cycle. To the best of our knowledge, it is the first work based on deep neural networks for gait period detection in the literature. An extensive empirical evaluation is provided on the CASIA-B dataset in terms of different views and network architectures with comparison to the existing works. The results show the good accuracy and robustness of the proposed method for gait periodicity detection.
Gait period detection Deep convolutional neural networks Gait recognition Biometrics technology
Kejun Wang Liangliang Liu Xinnan Ding Yibo Xu Haolin Wang
College of Automation,Harbin Engineering University,Harbin,China
国际会议
广州
英文
146-156
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）