Image-Based Facial Expression Recognition Using Local Neighborhood Difference Binary Pattern

Automatic facial expression recognition (FER) has gained enormous interest among the computer vision researchers in recent years because of its potential deployment in many industrial, consumer, automobile, and societal applications. There are a number of techniques available in the literature for FER; among them, many appearance-based methods such as local binary pattern (LBP), local directional pattern (LDP), local ternary pattern (LTP), gradient local ternary pattern (GLTP), and improved local ternary pattern (IGLTP) have been shown to be very efficient and accurate. In this paper, we propose a new descriptor called local neighborhood difference binary pattern (LNDBP). This new descriptor is motivated by the recent success of local neighborhood difference pattern (LNDP) which has been proven to be very effective in image retrieval. The basic characteristic of LNDP as compared with the traditional LBP is that it generates binary patterns based on a mutual relationship of all neighboring pixels. Therefore, in order to use the benefit of both LNDP and LBP, we have proposed LNDBP descriptor. Moreover, since the extracted LNDBP features are of higher dimension, therefore a dimensionality reduction technique has been used to reduce the dimension of the LNDBP features. The reduced features are then classified using the kernel extreme learning machine (K-ELM) classifier. In order to, validate the performance of the proposed method, experiments have been conducted on two different FER datasets. The performance has been observed using well-known evaluation measures, such as accuracy, precision, recall, and F1-score. The proposed method has been compared with some of the state-of-the-art works available in the literature and found to be very effective and accurate.