ROLE OF FACE FEATURE CLASSIFICATION FOR THE DETECTION AND RECOGNITION

Abstract

Humans easily recognize emotions from facial expressions, while computers struggle with it. Emotion recognition from facial expressions is vital for human-computer interaction, using machine learning in most studies. Computer vision research on simple emotion recognition remains difficult. Surprise, sadness, fear, contempt, happiness, & anger are among examples. Recently, deep learning has been discussed as a possible answer to several practical problems, one of which is recognition of emotion. In this paper, we enhance the CNN (Convolutional Neural Network) method for distinguishing seven basic emotions and tested the effects of various preprocessing procedures on classification accuracy. Through the application of facial traits and emotions, this research aims to improve emotion recognition. Computers draw more accurate inferences about a person's mental state and provide more appropriately customised replies if they can recognise or elicit facial expressions that suggest such states. Therefore, we investigate whether a deep learning approach employing a MCNN (Modified Convolutional Neural Network) may enhance emotion recognition from facial data. After the input image has been cleaned of noise using the preprocessing method and features have been extracted during the pretraining phase, face detection may be performed with high accuracy. 7 emotions represented by the FACS (Facial Action Coding System) are the basis for the MCNN we propose for classifying different facial expressions. To classify facial features, our proposed MCNN attained an impressive 97.1% accuracy.

Keywords: Classifying, face features, Recognition, Detection