TY - GEN
T1 - Video-based face recognition by Auto-Associative Elman Neural network
AU - Zhang, Bailing
AU - Zhou, Juntao
N1 - Publisher Copyright:
© 2014 IEEE.
PY - 2014/1/6
Y1 - 2014/1/6
N2 - While classical face recognition (FR) technologies are mainly based on static images, video-based FR is concerned with the matching of two image sets containing facial images captured from each video. Video based FR is supposed to be advantageous as it takes more abundant information in to account to improve accuracy and robustness. Though many methods have been proposed, there still exists a variety of challenges such as the variation in poses and occlusion. In this paper, we proposed a simple video-based face recognition system by proposing an Auto-Associative Elman Network (AAEN) for the comparison of facial image sequences from videos. AAEN is designed to reconstruct its inputs, while compressing the data to a lower-dimensionality in the hidden layer. In the recognition system, faces are first detected by applying the Viola-Jones algorithm and then tracked by exploiting Kalman filtering. We tested our method in two experimental settings, using a webcam for the simulation of video conferencing and a surveillance camera for indoor environments. Experiment results demonstrated that the proposed AAEN model can efficiently handle the temporal face sequences for the recognition task. The average recognition accuracies for the two experimental settings are 90.2% and 86.4% respectively.
AB - While classical face recognition (FR) technologies are mainly based on static images, video-based FR is concerned with the matching of two image sets containing facial images captured from each video. Video based FR is supposed to be advantageous as it takes more abundant information in to account to improve accuracy and robustness. Though many methods have been proposed, there still exists a variety of challenges such as the variation in poses and occlusion. In this paper, we proposed a simple video-based face recognition system by proposing an Auto-Associative Elman Network (AAEN) for the comparison of facial image sequences from videos. AAEN is designed to reconstruct its inputs, while compressing the data to a lower-dimensionality in the hidden layer. In the recognition system, faces are first detected by applying the Viola-Jones algorithm and then tracked by exploiting Kalman filtering. We tested our method in two experimental settings, using a webcam for the simulation of video conferencing and a surveillance camera for indoor environments. Experiment results demonstrated that the proposed AAEN model can efficiently handle the temporal face sequences for the recognition task. The average recognition accuracies for the two experimental settings are 90.2% and 86.4% respectively.
KW - Elman neural network
KW - Video-based Face recognition
KW - auto-associative memory
KW - face detection
KW - face tracking
UR - http://www.scopus.com/inward/record.url?scp=84946530866&partnerID=8YFLogxK
U2 - 10.1109/CISP.2014.7003755
DO - 10.1109/CISP.2014.7003755
M3 - Conference Proceeding
AN - SCOPUS:84946530866
T3 - Proceedings - 2014 7th International Congress on Image and Signal Processing, CISP 2014
SP - 89
EP - 93
BT - Proceedings - 2014 7th International Congress on Image and Signal Processing, CISP 2014
A2 - Wan, Yi
A2 - Sun, Jinguang
A2 - Nan, Jingchang
A2 - Zhang, Quangui
A2 - Shao, Liangshan
A2 - Wang, Lipo
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2014 7th International Congress on Image and Signal Processing, CISP 2014
Y2 - 14 October 2014 through 16 October 2014
ER -