The Impact of the Number of Eigen-Faces on the Face Recognition Accuracy using Different Distance Measures

The embedded and real-time systems are the main motivation for this research where the computations are critical to be reduced as much as possible. Face recognition method using eigen-faces yields good accuracy if enough eigen-faces are considered in the classification process. The more eigen-faces used, the more computation power is needed. In this paper, the main goal is to investigate the trade-off between the used number of eigen-faces and the accuracy and the needed computation power of face recognition. Three different distance measures are studied. Namely: Euclidean, block-city, and chess board distances are used. It is concluded that there is some optimum number of eigen-faces that provides the highest recognition rate and acceptable execution time. Moreover, the best number of eigen-faces highly depends on the selected distance measure.