Improving Facial Emotional Recognition Using Convolution Neural Network with Minimal Layer

Abstract

Human emotions identification has many applications, including human-computer interaction, illogical analysis, medical diagnosis, data-driven animation, and human-robot interaction. This paper presents a classification model, ConvNet that extracts features from facial images using techniques such as local binary patterns (LBP), convolutional neural networks (CNN), and region-based oriented FAST and rotational BRIEF (ORB). This model converges quickly. Experiment show that ConvNet outperforms existing methods with a precision of 98.13% on the CK+ dataset and 92.05 % on the JAFFE dataset.