Human Emotion Recognition using Deep Learning

As the world is developing, so are the needs to understand human emotional behavior in order to decrease the amount of crimes, increase the business sales and various medical purposes. This research paper tests the various algorithms present in today’s field which were used for human emotion recognition by which they achieve certain precision of understanding and recognizing the emotion of the human more correctly. Such systems attract lots of attention in the present world as this system will help the business grow and understand the present mood of their customers. The proposed system of Human Emotion Recognition (HER) consists of three very simple parts , the initial phase is face detection, second is feature extraction and finally recognition and output. Basically the system uses 3 layers of CNN as well as a system of OpenCV algorithms. The dataset which we used is of Keras but the color detection can be done through the Kaegal color model. The AAM (Active Appearance Model) method is used to extract and save various physical features and distance among them as a dataset to understand how faces change. Through understanding the Euclidean distance the output of pictures can be understood. The current recognition rate of this proposed system can be around 91% to 94%.Artificial Neuro-Fuzzy Inference System (ANFIS) is used to further modify the model results.