A Sentiment Analysis Study on Recognition of Facial Expressions: Gauss and Canny Methods

Abstract

Human-computer interaction has been the focus of today's current researches. Human-computer interaction is accepted as a multidisciplinary field that takes place through interfaces. These interfaces can sometimes be software functions, or sometimes they can be interact provided with hardware components. Facial expressions give information about people's emotions play an important role in sentiment recognition. Today, facial expressions are used in many fields such as education, psychological studies, virtual reality, robotics, facial animation, health and law, and the need for analysis of facial expressions in many areas is increasing. In addition, the analysis of human facial expressions with computers is a remarkable research area, but it is considered a challenging problem. In this context, it is necessary to analyze facial expressions accurately and quickly by software. In this study, sentiment recognition from facial expressions (sad, happy, scared, confused) was performed using 50 different images obtained from various databases and internet sources. With digital image processing techniques, improved images can be obtained and feature extraction can be made. In this research, digital image processing functions and MATLAB programming language of MATLAB 2018 program, which provides advanced programming for scientific studies, were used. Image noise was removed with the Gauss filter, and edge detection operations were performed with the Canny method. Geometric ratios were used to eliminate errors. As a result of the study, it was determined that sentiment recognition procedures performed on images with similar facial expressions made incorrect sentiment classification. However, it has been observed that face recognition with MATLAB functions and MATLAB programming has generally produced successful results.