In this paper, a new character recognition method based on topological
properties, positions of starting point, statistical analysis, and sh
ape recognition is proposed for recognition of unconstrained handwritt
en digits based on various contour information. First, input digits ar
e classified into three groups according to their topological properti
es. In group 1, the features of normalized length of the digit contour
, the normalized area of the digit and Fourier descriptors of the oute
r contour of a digit are used for recognition of digits. In group 2, t
he relative position of outer and interior contour centroids, the posi
tion of interior contour centroid, the mean and the variance of the ou
ter contour distance function, and Fourier descriptors of the outer co
ntour of a digit are used for the recognition of digits. In group 3, u
sually there is only one digit 8, but because of the complexity of wri
ting styles, some other digits are still classified into this group. T
he final recognition is based on shape comparison of the input digit w
ith models. In the recognition process, some special models are establ
ished and used for recognition of broken digits and digits whose topol
ogical properties are destroyed. In our experiment, 1000 digits from t
he NIST database are used for training and 5278 unseen digits are used
for testing. The recognition rate has reached 98.5% with a reliabilit
y rate of 99.09%, a substitution rate of 0.91% and a rejection rate of
0.59%. (C) 1997 Pattern Recognition Society. Published by Elsevier Sc
ience Ltd.