A methodology for improving recognition rate of linear discriminant analysis in video-based face recognition using support vecto