Skip to Main Content
Particle Swarm Optimization (PSO) algorithm has been widely recognized as an efficient algorithm with applicability is many areas. Recently, we proposed the Guided Particle Swarm Optimization (GPSO) algorithm, which is a modification to PSO designed for facial emotion detection. GPSO involves keeping track of relevant points, called Action Units (AUs), which are specified on the face of a subject. Swarm of particles was defined such that each particle has a component within the neighborhood of each AU. We implement GPSO into a facial emotion detection system that can detect the six basic universal emotions. The system was tested on a variety of subjects and the results realized were very promising. However, application of the system was limited to pre-recorded video clips because the AUs must be pre-processed to obtain their positions over time, which were then fed as input to the system. In this paper, we present an improvement we made to the system by applying Lucas-Kanade (LK) optical flow algorithm. LK algorithm allowed us to keep track of the positions of the AUs in real-time, thereby eliminating the need for preprocessing. The improved system is now a real-time system that works on video streams to identify facial emotions and achieved the same promising detection success rates as the original system.