A Deep Learning Model for Human Emotion Recognition on Small Dataset

Humans express their emotions through facial expressions. On the other hand, facial expression recognition has remained a difficult and fascinating subject in computer vision. For recognition of emotions is difficult because of the lack of a landmark demarcation between the emotions on the face, as well as the complexity and variety. In this paper, the human emotional states through facial expression are finding through the Convolutional neural network model. Firstly, the images have been taken from the publically Jaffe (Japanese female facial expression) and KDEF (Karolinska Directed Emotional Faces) dataset. After the dataset is taken the threshold technique has been applied for removing the background in the image for improving accuracy. Therefore, the proposed CNN model achieves higher accuracy as compared toprevious state-of-the-art techniques for emotion recognition.