패턴 네트워크와 유전 목 지도의 동적 확장을 이용한 특징 자동 선택 알고리즘

다국어 초록 (Multilingual Abstract)

In this paper, we suggest method which has automatic selected feature extracting algorithms for every pattern. It joins pattern network and dynamic extension of genetic tree maps (DGTM). The latter in the map has simple add and update action for needed features and has fast training method. We define DCOM (Distributed Component Object Model) pattern network model consists of server and client. The client connects feature extracting algorithms in pattern server. The suggested method can be expand to every pattern category such as video, fingerprint, character, etc.. The experiment analyzed algorithm performances about numeral and face patterns.

참고문헌 (Reference)

1 하성욱, "유전 목 지도의 동적 확장" 한국정보과학회 29 (29): 386-395, 2002

2 L.D.Davis, "The Handbook of Genetic Algorithms" Van Nostrand Reinhold 1991

3 Bilan Zhu, "Segmentation of On-Line Handwritten Japanese Text Using SVM for Improving Text Recognition" (1) : 208-219, 2006

4 El-Sayed M. El-Alfy, "Offline Recognition of Handwritten Numeral Characters with Polynomial Neural Networks Using Topological Features" 173-183,, 2010

5 K.M.M.R.Batuwita, "New Segmentation Algorithm for Individual Offline Handwritten Character Segmentation" 3614 : 215-229, 2005

6 Lajish, V.L, "Handwritten Character Recognition Using Gray-scale Based State-Space Parameters and Class Modular NN" 374-379, 2008

7 Venkataramani, B, "FPGA implementation of isolated digit recognition system using modified back propagation algorithm" 1-6, 2008

8 Eric Harmon, "Delphi COM Programming" Macmillan Technical Publishing 2000

9 Jun Zhang, "An automatic segmentation technique for color images based on SOFM neural network" 3528-3533, 2009