Face micro-expression is crucial for feeling perception and yet demanding due to the high dimension nature and the increasingly request for the recognition accuracy. The tradeoff between accuracy and efficiency by Deep Belief Network is a challenging. This paper shows that a two-stage strategy can achieve both speedup and high accuracy. With it, an efficient facial micro-expression algorithm is proposed that consists of Double Weber Local Descriptor devised in this paper firstly for extracting initial texture local features, and Deep Belief Net for more global feature and less computation dimension. The experiments with JAFFE database show that the average recognition rate by the new algorithm is up to 92.66%, and the rate of neutral facial expression is nearly 100%. Compared with LBP, LDP, PCA, Gabor wavelet and Weber local descriptor combined with DBN, the new algorithm of the introduction of Double Weber Local Descriptor into DBN has higher recognition rate.
[1]
Zhenhua Guo,et al.
A Completed Modeling of Local Binary Pattern Operator for Texture Classification
,
2010,
IEEE Transactions on Image Processing.
[2]
Matti Pietikäinen,et al.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2009, TPAMI-2008-09-0620 1 WLD: A Robust Local Image Descriptor
,
2022
.
[3]
Matti Pietikäinen,et al.
Discriminative features for texture description
,
2012,
Pattern Recognit..
[4]
Tian Yantao,et al.
Facial Expression Recognition Method Based on Gabor Multi-orientation Features Fusion and Block Histogram
,
2011
.
[5]
Junmo Kim,et al.
Development of deep learning-based facial expression recognition system
,
2015,
2015 21st Korea-Japan Joint Workshop on Frontiers of Computer Vision (FCV).