A novel approach to detect sentimental analysis using machine learning

Abstract

ABSTRACT newlineAutomated facial expression analysis for Emotion Recognition (ER) system is an emerging research area towards creating socially intelligent systems. Facial expression recognition is the most important criteria for effective Human Computer Interaction (HCI) as well as a medium to understand and communicate with children who cannot emote verbally. The eye region, often considered integral for ER by psychologists and neuroscientists, has received very little attention in engineering and computer sciences. Using eye region as an input signal presents several benefits for low-cost, non-intrusive ER applications. This work proposes two frameworks towards ER from eye region images. The first framework uses Local Binary Patterns (LBP) as the feature extractor on grayscale eye region images. The features extracted were then tested on standard classifiers i.e., Support Vector Machine (SVM) and K-Nearest Neighbourhood (KNN) classifiers. Facial expression images from JAFFE and Cohn-Kennedy databases were utilized for training as well as testing. The results validate the eye region as a significant contributor towards communicating the emotion in the face by achieving high person-dependent accuracy. The Gabor filter and radial encode is used firstly to divide the expression image into local regions, then PCA and LDA is adopted for feature selection, finally the extended nearest neighbor algorithm is applied to classify the facial expression data. Experiments and analysis conducted on Japanese Female Facial Expression (JAFFE) Database show that this method achieves better efficiency and effectiveness. newlineThree methods based on local or global facial features are implemented using simulation environment provided by Math Works. In this thesis, Method I is referred as ZMNB, Method II is referred as GFSVM, Method III is referred as GFLVQ, and Method IV is referred as HM I and Method V or proposed method is referred as Emotional Affect Recognition through Hybrid Approach (EARTHA). newlineA novel EARTHA algorithm is developed in s