Video Based Sentiment Analysis

It is generally said, that our inner emotions and feelings are mirrored on our faces. For better communication, Face Expressions Recognition plays a vital role and can be used in the study of human behaviour and psychological characteristics. In the domain of Machine Learning, this concept of reading and understanding human expressions falls into the category of Cognitive Systems and the study of the thought process of humans by understanding the human behaviour and emotions using the Machine Learning Algorithms is known as Cognitive Science.The technology of recognition of facial expression is basically a Sentiment Analysis tool that can detect six universal expressions: “happiness”, “anger”, “sadness”, “surprise”, “fear” and “disgust”. In this paper, the Sentiment analysis of a prerecorded video has been done using the algorithm of Face Emotion Recognition (FER) to depict the intensity of an emotion. The goal is to identify the emotions present in the pre-recorded video (input data) in order to emulate the human thought process. The approach used in this paper to do the video sentiment analysis is Statistical Approach which uses the algorithms of Machine Learning and Deep Learning such as CNN to detect the sentiments precisely. The result is visible in the form of graphs and it depicts various sentiment intensities. We can see every emotion plotted against time in the graph. In addition, the output is a replication of the original video with a box around the detected face exhibiting real emotions within the video.