Facial Expression Recognition Using Convolutional Neural Network
The recognition of facial expression is a key step in nonverbal human communication in the field of computer vision. The Facial Expression Recognition (FER) system reflects emotions as well as mental activities, social interaction with people and expressing positive or negative attitude in general. It has many applications like human machine interaction, understanding human behavior and emotions, mental state identification, drowsiness detection for driving safety and health care. One of the key challenges in the FER system is the dynamic variation while capturing the image. Convolutional Neural Networks (CNN) for facial expression recognition task has been developed. The main goal is to classify each facial image in to one of the seven facial emotions categories like a happiness, sadness, surprise anger, fear, disgust and neutral. Convolutional Neural Network (CNN) can be used to find auto learned features and classify the features of facial expression from human face. The algorithm is tested on various standard databases like Japanese Female Facial Expressions (JAFFE) which includes 213 images, Cohn-Kanade (CK) which includes 7809 images and Cohn-Kanade (CK+) which includes 10584 images of different facial expressions. Experimental results show that CNN has good performance on facial expression recognition which results in 95.20% recognition accuracy on CK+ database.
Cite this Article
Kajal Parmar, Heena Kher, Mikita Gandhi. Facial Expression Recognition Using Convolutional Neural Network. Journal of Open Source Developments. 2019; 6(1): 18–27p.
- There are currently no refbacks.