A Development of Directional Gradient Local Ternary Pattern for Facial Expression Recognition

Abstract

Humans can easily determine ones’ gender, identity and ethnicity with highest accuracy as compared to face expression recognition.. This makes development of automatic face expression recognition techniques that surpass human performance an attractive yet challenging task. Facial expressions recognition requires extraction of robust and reliable expression discriminative features. Local binary patterns (LBP) sensitivity to noise makes it insufficiently reliable in capturing expression discriminative features. Although a local ternary pattern (LTP) is insensitive to noise, it uses a single static threshold for all images regardless of varied image conditions. Local directional patterns (LDP) uses k directional responses to encode image gradient and disregards not only central pixel in the local neighborhood but also 8 − kdirectional responses. Every pixel in an image carry subtle information. Discarding 8 − k directional responses lead to lose of discriminative texture features. Each pair are compared and the bit corresponding to the maximum value in the pair is set to 1 while the resultant binary code is converted to decimal and assigned to the central pixel. Local ternary directional patterns (LTDP) first get the difference between neighboring pixels and central pixel in 3×3 image region. These differential values are convolved with Kirsch edge detectors to obtain directional responses. These responses are normalized and used as probability of an edge occurring towards a respective direction. An adaptive threshold is applied to deriveLTDP code. The LTDP code is split into its positive and negative LTDP codes. Histograms of negative and positive LTDP encoded images are concatenated to obtain texture feature.. This study proposes fusion of different facial features to enhance their discriminative power. Experimental results show that face component comparing with whole face achieve lower MAE compared to single feature performance. Finally the study used Alex-Net, Vgg-16 & Resent which applied on images to extract the feature of facial images and obtained features in the last fully connected layer which are used as input to SVM classifier for producing the final classification result. The study shows that when this study used CK & JEFFE datasets and SVM classifier in whole face the accuracy is 99.3% but when it used facial component the accuracy is 99.7% this results show that the accuracy gradually increas