Abstract
Smartphone is a good indicator of mental status that has been employed in healthcare environments in order to assess the mental status of elderly patients. In such environments, the mental status of humans can be analyzed by expressions. Expressions have a significant role in improving the level of interaction between human-to-human communications. The features from human mouth, eyes, and eyebrow are considered the most informative features for expressions recognition. Commonly, these systems were tested on publicly available datasets which were collected using a fixed camera with static background. In this work, we proposed a real-time facial expression recognition (FER) system using smartphone camera. In order to make the system efficient and robust, we have extracted the features only from the contributing parts of the face, for which the angles and distances were measured. Then for classification, we used support vector machine (SVM) under 10$-$fold cross validation setting. The system was tested and validated in real time by 10 university students (which are not professional). The weighted average recognition rate for the proposed smartphone-based FER system is 85.6% for 5 basic expressions, which is a significant improvement in real-time domain.