HOG-based Emotion Recognition Using One-Dimensional Convolutional Neural Network

This work proposes an emotion detection approach using Histogram of Oriented Gradients algorithm. Emotion detection is a crucial area since the emotions are extremely person dependent and finding them is hard with various lightning and illumination changes. Most of the works in this field focus on predicting the emotion using the facial region. In the proposed work, emotion detection is done using the mouth region. The dataset is comprised of mouth images containing emotions such as happy, normal and surprised in the form of video frames. The mouth regions are detected using the Haar-Based Cascade classifier at 20 frames per second. The HOG features are then extracted to detect three emotions namely Happy, Normal and Surprised. These HOG features are then trained using One-Dimensional Convolutional Neural Network (1D-CNN). The experimental results show that the proposed system can identify the emotions which gave improved performance than the earlier works.