Real Time Face Expression Classification Using Convolutional Neural Network Algorithm

The study of facial expression has been exist since the Aristotelian erain which it explains the expression into a complicated nonverbal research. Expression signals are instinctive behaviors, not learning outcomes, so machines will not be easy to recognize the varied expressions of people. In this study, the detection of accuracy levels in the form of predictions of class expressions of pleasure, sadness, surprise, disgust, neutral, anger and fear in real time with captured face images from the camera of an Android mobile phone isconducted. This research is important as the first step in detecting individual lies in terms of non-verbal communication and how accurately it detects someone's expression. Broadly speaking, the research process is divided into 2, an approach to detect face position with Viola Jones and extraction and classification with deep learning Convolutional Neural Network to produce predictions of expressions from machine learning. In CNN there are several layers for the filtering process by conducting the learning process to find the best representation. Datasets used face image capture in real time using a mobile camera with a total of 41 try expression detection.The result of the study shows that the performance of the Viola Jones and CNN method produces the highest level of accuracy, that is 98,7% for happiness expression with 9 times detection.