![](bg6.png)
Table 2 shows a comparison of the different face
recognition techniques on the ORL face database
which reported their computational cost. As we can
see from Table 2 our system has a recognition rate
of 99% and a low computational cost. Proposed
system was implemented in Matlab 7.1 and tested on
a machine with CPU Pentium IV 2.8 GHz with 512
Mb Ram and 1 Mb system cache.
Finally we tested our system on YALE face
database. The Yale face database contains 165
images of 15 subjects. There are 11 images per
subject with different facial expressions or lightings.
Figure 10 shows the 11 images of one subject. We
resized YALE database from
195231× into 6464
jpeg face images. No other changes like background
cutting or cropping the images were preformed. We
obtained our system results on 1 image to 10 images
for training using 960 symbols. Table 3 shows
Comparative results on this database.
Table 2: Comparative computational costs and recognition
results of some of the other methods as reported by the
respective authors on ORL face database.
Method Recog.(%) Train. time
per image
Recog. time
per image
PDBNN
(Lin et al., 1997)
96 20min
0.1 sec.
n-tuple
(Lucas, 1997)
86 0.9 sec. 0.025 sec.
Pseudo-2D HMM
(Samaria and Young,
1994)
95 n/a 240 sec.
DCT-HMM (Kohir
and Desai, 1998)
99.5 23.5 sec. 3.5 sec.
Proposed method 99 0.63 sec. 0.28 sec.
Figure 10: A class of the YALE face database.
Table 3: Experiments on YALE face database. Our
accuracy obtained on
6464 ×
resolution face images.
# of train
image(s)
MRF
(Huang et al.,
2004)
PCA
(Huang et al.,
2004)
Proposed
method
1 81.6% 60.04% 78%
2 93.11% 75.2% 82.22%
3 95.17% 79.03% 90.83%
4 95.9% 79.75% 94.29%
5 96.11% 81.13% 97.78%
6 96.67% 81.15%
100%
7 98.67% 81.9% 100%
8 97.33% 81.24% 100%
9 97.33% 81.73% 100%
10 99.33% 81.73% 100%
4 CONCLUSIONS
A fast and efficient system was presented. Proposed
system used SVD for feature extraction and 1-D
HMM as classifier. The evaluations and
comparisons were performed on the two well known
face image databases; ORL and YALE. In both
databases, approximately having a recognition rate
of 100%, the system was very fast. This was
achieved by resizing the images to smaller size and
using a small number of features.
Future work will be directed towards evaluating
the proposed system on larger face databases.
REFERENCES
Rabiner, L. R., 1989. ‘A tutorial on Hidden Markov
Models and Selected Applications in Speech
Recognition’, IEEE Proceedings, Vol. 77, No. 2, pp.
257-285.
Bicego, M., Castellani, U., and Murino, V., 2003. Using
Hidden Markov Models and Wavelets for face
recognition. In Proceedings IEEE International
Conference on Image Analysis and
Processing(ICIAP), 0-7698-1948-2.
Samaria, F.S. and Young, S., 1994. ‘HMM-based
Architecture for Face Identification’, Image and
Vision Computing, Vol. 12, No. 8,pp. 537-543.
Haralick, R.M. and Shapiro, L.G., 1992. Computer and
Robot Vision, Volume I. Addison-Wesley.
Kanade, T., 1973. “Picture Processing by Computer
Complex and Recognition of Human Faces,” technical
report, Dept. Information Science, Kyoto Univ.
Wiskott. L., Fellous, J.-M., Krüger, N., and vondder
malsburg, C., 1997. Face Recognition by Elastic
Bunch Graph Matching. IEEE Transaction on Pattern
Analysis and Machine Intelligence, 19(7):775-779.
Klema, V. C., and Laub, A. J., 1980. The Singular Value
Decomposition: Its Computation and Some
Applications. IEEE Transactions on Automatic
Control, 25(2):0018–9286.
Huang, R., Pavlovic, V., and Metaxas, D. N., 2004. A
Hybrid Face Recognition Method using Markov
Random Fields. IEEE, 0-7695-2128-2.
Kohir, V. V., and Desai, U. B., 1998. Face recognition
using DCTHMM approach. In Workshop on Advances
in Facial Image Analysis and Recognition Technology
(AFIART), Freiburg, Germany.
Turk, M., and Pentland, A., 1991. “Eigenfaces for
Recognition,” J. Cognitive Neuroscience, vol. 3, pp.
71-86.
Lin, S., Kung, S., and Lin, L., 1997. Face
Recognition/Detection by Probabilistic Decision-
Based Neural Network. IEEE Trans. Neural Networks,
8(1):114–131.
Lucas, S. M., 1997. Face recognition with the continuous
n-tuple classifier. In Proceedings of British Machine
Vision Conference.
A NEW FACE RECOGNITION SYSTEM - Using HMMs Along with SVD Coefficients
205