A multicore accelerated implementation for statistical analysis of facial expressions in video

Recently, the analysis of facial expressions has become one of the important research areas in computer vision and image processing. Analysing facial expressions quickly and accurately play a critical role in many software systems for various applications. In this study, a new algorithm is proposed for the acceleration of video-based facial expression analysis that is performed using cubic Bézier curves. Performance evaluation of the expression analysis accelerated with parallel threads on multi-core computers has been presented. The experiments have been conducted with eNTERFACE'05 emotion video database. Experimental results were obtained using quad-core processor with Hyper Threading technology. According to experimental results, quad core processor using two threads produced about 1.8 fold speed-up and four threads produced about 2.9 fold speed-up while eight threads produced about 3.5 fold speed-up has been obtained.