Effective approach for facial expression recognition using hybrid square-based diagonal pattern geometric model

Abstract

Facial expressions convey human emotions in an expressive way. The development of an automated system to recognize the facial expressions is the difficult task. Automatic Facial Expression Recognition (FER) is an imperative process that leads to next-generation Human-Machine Interaction (HMI) for clinical practice and behavioral description. Segment detection and the extraction of relevant information from the images are the major issues to design an effective FER system. The creation of the suitable system that addresses these issues is the basic stage to achieve the accurate HMI models. The in-depth information analysis and maximization of labeled database are the real problems in the domain of FER approaches. A novel framework based on Square-Based Diagonal Pattern (SBDP) method on Geometric model called Geometric Appearance Models (GAM) has been presented through this paper that extracts the in-depth detailed of the features. The framework adopts the co-training by using detailed information from RGB-D images. The performance analysis of proposed SBDP-GAM regarding identification rate, sensitivity, accuracy and error rates with the RGB-D images shows the effectiveness diagonal patterns in facial expression identification. The comparative analysis of proposed SBDP-GAM model with the traditional methods regarding the recognition rate, error rate and F-score and on RGB-D images from EURECOMM database states the effectiveness of the proposed method. Moreover, the comparison of proposed SBDP-GAM with existing Support Vector Machine (SVM) regarding the acceptance rate (FAR, FRR, GAR) measures for biographer RGB-D image database proves the effectiveness of SBDP-GAM in FER applications.