Independent Component Analysis and Number of Independent Basis Vectors

Abstract

Abstract Among the various biometric systems, face recognition is an important area of research due to its applications in Human Computer Interaction, biometrics and security. It is one of the most popular research areas in the field of computer vision and pattern recognition. This paper addresses the use of Independent Component Analysis (ICA) for recognizing human faces. It is implemented using InfoMax algorithm. Face recognition performance is evaluated using Architecture-I which treats images as random variables and pixels as outcomes. We are observing the sensitivity of ICA to the dimensionality of final subspace. Experiments are carried out on ORL face database which consists of 400 face images. We presented recognition rate of the system corresponding to number of independent basis vectors along with the energy retained in number of eigenvectors of underlying Principal Component Analysis (PCA) subspace. Our results show that the performance of face recognition using ICA increases with the number of statistically independent basis vectors.