Facial Expression Recognition Based on WAPA and OEPA Fastica
Humayra Binte Ali, David M W Powers · International Journal of Artificial Intelligence & Applications · 2014
Face is one of the most important biometric traits for its uniqueness and robustness.For this reason researchers from many diversified fields, like: security, psychology, image processing, and computer vision, started to do research on face detection as well as facial expression recognition.Subspace learning methods work very good for recognizing same facial features.Among subspace learning techniques PCA, ICA, NMF are the most prominent topics.In this work, our main focus is on Independent Component Analysis (ICA).Among several architectures of ICA, we used here FastICA and LS-ICA algorithm.We applied Fast-ICA on whole faces and on different facial parts to analyze the influence of different parts for basic facial expressions.Our extended algorithm WAPA-FastICA and OEPA-FastICA are discussed in proposed algorithm section.Locally Salient ICA is implemented on whole face by using 8x8 windows to find the more prominent facial features for facial expression.The experiment shows our proposed OEPA-FastICA and WAPA-FastICA outperform the existing prevalent Whole-FastICA and LS-ICA methods. .