Abstract:
Multiple researchers, maximum human communication is non-verbal basis. The traditional Human Computer Interaction (HCI) focuses only on the intended input and overlooks the information delivered non-verbally. So, it is safe to say that there is need of a system which helps to recognize and perceive the objective and emotions expressed by social indicators. For the last many years, research on Facial Expressions Recognition (FER) has been under consideration in computer vision field but there are still many answerable questions outstanding. Through this research few of those questions are tried to be figured out. Emotions are said to be helpful in different fields such as biomedical engineering, psychology, neuroscience and medical health. They are helpful in analyzing many diseases in medical filed. In last few year Deep Learning is one of the major progressing field used for image classification problem. Through this research, a Convolutional Neural Network (CNN) based architecture has been proposed, which can be used for facial expression recognition problems. Emotions was classified into 7 classes include happy, sad, neutral, angry, fear, surprise, disgust. Computed results were very effective in observing human behavior which, as a result, would be helpful in psychological disorders. An independent method was proposed for this work. In First Method, using autoencoders to develop a unique representation of sentiments, while in second part using 8-layer convolutions neural network (CNN). These Methods use Google facial expression comparison data set. Computed results indicated that with better tuning of data, this model can perform accurately. So, through the simulation study, it was also observed that the prediction performance of these proposed approaches was far better in each class as compare to existing approaches. Overall, these proposed approaches played an import role in emotion recognition which can help significantly in identification of biological disorders, Computational Biology, Molecular Biology, Bioinformatics. Also, this might help in applications related to emotion recognition. In addition, the related web predictors used in this study provide sufficient information to researchers and academicians in future research