Human emotion recognition with morphological segmentation of facial features using elm

Abstract

Human emotion detection has been a challenging topic in the field of human-computer interaction. To develop a more natural interaction between human and computer it is expected that the computer is able to perceive and respond to human emotion. In this paper we provide a better approach to predict human emotions accurately. The proposed system employs ELM as its learning algorithm because of its flexible optimization constraints compared to other algorithms like SVM, CNN, etc. In this framework CLAHE is used to normalize the sequences extracted from the Cohn-Kanade dataset. HAAR Classifier is used to detect the edge lines. Gabor filter and two dimensional principle component analysis (2DPCA) are used for feature extraction. ELM is then applied to classify the features. The experiments of facial emotion recognition are performed using Cohn-Kanade dataset, in which 95% recognition rate is achieved. This system provides promising results implemented in personalization face case which can be utilized in developing personalised applications to detect six basic human emotions namely anger, disgust, fear, happiness, sadness, surprise.