A new algorithm for the accelerated statistical analysis of facial expressions on video

Facial expressions are universal symbols of emotions that provide cohesion to interpersonal communication. Facial expression analysis has widespread range of application in areas such as analysis of human behaviors, human-human interaction and human-computer interaction. Lately, especially due to developments in human-computer interaction, the understanding of human emotions by computer has become an indispensable issue. Besides, analysis and recognition of facial expressions has prevailed in various areas such as security, psychology, education, health, entertainment, and robotics. For these reasons, the analyzing of facial expressions quickly and the recognition of facial expressions correctly according to the analyzed facial expressions play a critical role for many software systems in different applications. In this thesis, a new algorithm is proposed for the acceleration of video-based facial expression analysis that is performed using cubic Bezier curves. By reducing the total number of analyzed frames, performance evaluation of the expression analysis accelerated with parallel thread on multi-core computer was performed. Additionally, the results of frames which were found to be incorrect were fixed by performing error analysis on the results of statistical analysis.