• Proceedings of the Korean Information Science Society Conference
• /
• 2008.06c
• /
• pp.569-573
• /
• 2008

패턴인식 문제를 다루는 연구에서 인식 성능을 향상시키고자 베이스 에러율의 상한인 조건부 엔트로피를 응용하는 시도가 있었다. 본 논문에서는 다수의 인식기로 구성된 다수 인식기 시스템이 우수한 성능을 보이도록 인식기의 집합을 결정하는 문제에서 이러한 조건부 엔트로피의 최소화를 통하여 시도한 방법과 다른 방법들을 간단하고 분명한 예제를 통하여 비교, 분석해 보았다. 다수 인식기의 결합 방법으로 대표적인 투표 기법과 조건부 독립 가정의 베이지안 기법을 사용하였으며, 조건부 엔트로피의 최소화를 통하여 인식기의 집합을 결정하는 방법에 대한 유용성을 확인할 수 있었다.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2005.04a
• /
• pp.403-414
• /
• 2005

Most of the currently used spam-filters are based on a Bayesian classification technique, where some serious problems occur such as a limited precision/recall rate and the false positive error. This paper addresses a solution to the problems using a modified Bayesian classifier based on chi-square statistics. The resulting spam-filter is more accurate and flexible than traditional Bayesian spam-filters and can be a personalized one providing some parameters when the filter is teamed from training data.

• Journal of KIISE:Software and Applications
• /
• v.28 no.9
• /
• pp.681-687
• /
• 2001

In order to raise a class discrimination power by combining multiple classifiers under the Bayesian decision theory, the upper bound of a Bayes error rate bounded by the conditional entropy of a class variable and decision variables obtained from training data samples should be minimized. Wang and Wong proposed a tree dependence first-order approximation scheme of a high order probability distribution composed of the class and multiple feature pattern variables for minimizing the upper bound of the Bayes error rate. This paper presents an extended high order product approximation scheme dealing with higher order dependency more than the first-order tree dependence, based on the minimization of the upper bound of the Bayes error rate. Multiple recognizers for unconstrained handwritten numerals from CENPARMI were combined by the proposed approximation scheme using the Bayesian formalism, and the high recognition rates were obtained by them.

• The KIPS Transactions:PartB
• /
• v.11B no.6
• /
• pp.667-674
• /
• 2004

Extraction of texts from images is a fundamental and important problem to understand the images. This paper suggests a text region verification method by statistical means of stroke features of the characters. The method extracts 36 dimensional features from $16\times16$sized text and non-text images using wavelet transform - these 36 dimensional features express stroke and direction of characters - and select 12 sub-features out of 36 dimensional features which yield adequate separation between classes. After selecting the features, SVM trains the selected features. For the verification of the text region, each $16\times16$image block is scanned and classified as text or non-text. Then, the text region is finally decided as text region or non-text region. The proposed method is able to verify text regions which can hardly be distin guished.