Non verbal approach for emotion detection

Abstract

Non verbal approaches of emotion detection plays vital role in various applications like E-learning, automatic pain monitoring, driver alert system, cognitive assessment etc. which are developed to enhance quality of human life. Facial expressions based approach is one of the very effective approaches which is used widely on standalone basis or combined with other approaches known as multimodal techniques of emotion detection. The paper discusses facial expressions based emotion detection which uses patch based face features and SVM Classifier. The experimentation carried out on JAFFE database provides 90.65 % average accuracy for emotion detection for basic emotions happy, anger, sad, surprise, disgust, fear and neutral. The other performance parameters through experimentation are obtained as Average True positive rate is 90.5942 %, average false positive rate is 9.3671 % and average false negative rate is 9.4 %. The average feature extraction time is 18.49 s and emotion detection time through these extracted features is 1.1 s.