SUMMARY
Facial expressions remain a significant component of human-to-human interface and have the potential to play a correspondingly essential part in human-computer interaction. Support Vector Machine (SVM) by the virtue of its application in a various domain such as bioinformatics, pattern recognition, and other nonlinear problems has a very good generalization capability. However, various studies have proven that its performance drops when applied to problems with large complexities. It consumes a large amount of memory and time when the number of dataset increases. Optimization of SVM parameter can influence and improve its performance.Therefore, a Culture Particle Swarm Optimization (CPSO) techniques is developed to improve the performance of SVM in the facial expression recognition system. CPSO is a hybrid of Cultural Algorithm (CA) and Particle Swarm Optimization (PSO). Six facial expression images each from forty individuals were locally acquired. One hundred and seventy five images were used for training while the remaining sixty five images were used for testing purpose. The results showed a training time of 16.32 seconds, false positive rate of 0%, precision of 100% and an overall accuracy of 92.31% at 250 by 250 pixel resolution. The results obtained establish that CPSO-SVM technique is computational efficient with better precision, accuracy, false positive rate and can construct efficient and realistic facial expression feature that would produce a more reliable security surveillance system in any security prone organization.