Research on face recognition is one of the
interesting biometric technologies to study since human can
recognize faces of a particular physical characteristic. There are
many methods used in face recognition process, but different
lightning, angle and expression are problems to get optimal
accuration for face recognition. From those problems, HMM
method is a good method to deal with it if it is combined with
the right feature extraction. This research will combine Gabor
Wavelet and SVD methods with single feature extraction using
SVD applied in same variable and databases. Gabor Wavelet is
used to eliminate the variability caused by the illumination
contrast and a slight shift as well as image deformation. The
SVD process will make an image that will be processed to be
more details with the extraction of two-dimensional into onedimensional
images. HMM is used for face recognition method
that utilizes Baum-Welch algorithm for the training process and
Viterbi in the testing process. Based on the combined feature
extraction applied to the 4 databases, each 100 facial image
produces accuracy 81% AMP, 86% Jaffe, 96% ORL and 59%
YALE accuracy. In the future, besides the needs in the accuracy
or research, specific researches on why there are differences of
test results in each facial image database is also needed to
conduct.
