• Proceedings of the Korea Information Processing Society Conference
• /
• 2008.05a
• /
• pp.148-151
• /
• 2008

어려운 패턴인식 문제를 다루기 위하여, 다수 인식기를 사용하는 다수 인식기 시스템의 개발에 관한 연구가 활성화 되었으나, 다수 인식기 시스템의 효율적인 구축에 관한 체계적인 시도는 그리 많지 않았다. 다수 인식기 시스템의 효율성은 인식기 집합에 포함되는 인식기의 선택 방법과 선택된 인식기들의 결합 방법에 의해서 결정되는 시스템의 인식 성능으로 판단될 수 있다. 따라서, 이들 요인을 고려하여 효율성이 높은 다수 인식기 시스템을 구축하는 방법을 살펴보고자 한다.

• Journal of KIISE:Computing Practices and Letters
• /
• v.14 no.3
• /
• pp.326-330
• /
• 2008

The study on combining multiple classifiers in the field of pattern recognition has mainly focused on how to combine multiple classifiers, but it has gradually turned to the study on how to select multiple classifiers from a classifier pool recently. Actually, the performance of multiple classifier system depends on the selected classifiers as well as the combination method of classifiers. Therefore, it is necessary to select a classifier set showing good performance, and an approach based on information theory has been tried to select the classifier set. In this paper, a classifier set candidate is made by the selection of classifiers, on the basis of mutual information between classifiers, and the classifier set candidate is compared with the other classifier sets chosen by the different selection methods in experiments.

• 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.

• Journal of KIISE:Software and Applications
• /
• v.31 no.5
• /
• pp.577-585
• /
• 2004

Storing and estimating the high order probability distribution of classifiers and class labels is exponentially complex and unmanageable without an assumption or an approximation, so we rely on an approximation scheme using the dependency. In this paper, as an extended study of the second-order dependency-based approximation, the probability distribution is optimally approximated by the third-order dependency. The proposed third-order dependency-based approximation is applied to the combination of multiple classifiers recognizing handwritten numerals from Concordia University and the University of California, Irvine and its usefulness is demonstrated through the experiments.

• 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.

• Journal of KIISE:Software and Applications
• /
• v.27 no.8
• /
• pp.855-863
• /
• 2000

Although Behavior-Knowledge Space (BKS) method, one of well known decision combination methods, does not need any assumptions in combining the multiple experts, it should theoretically build exponential storage spaces for storing and managing jointly observed K decisions from K experts. That is, combining K experts needs a (K+1)st-order probability distribution. However, it is well known that the distribution becomes unmanageable in storing and estimating, even for a small K. In order to overcome such weakness, it has been studied to decompose a probability distribution into a number of component distributions and to approximate the distribution with a product of the component distributions. One of such previous works is to apply a conditional independence assumption to the distribution. Another work is to approximate the distribution with a product of only first-order tree dependencies or second-order distributions as shown in [1]. In this paper, higher order dependency than the first-order is considered in approximating the distribution and a dependency-based framework is proposed to optimally approximate the (K+1)st-order probability distribution with a product set of dth-order dependencies where ($1{\le}d{\le}K$), and to combine multiple experts based on the product set using the Bayesian formalism. This framework was experimented and evaluated with a standardized CENPARMI data base.

• Annual Conference on Human and Language Technology
• /
• 1997.10a
• /
• pp.175-182
• /
• 1997

지금까지의 한글 문자인식 후처리 연구분야에서 미등록어와 비문맥적 오류 문제는 아직까지 잘 해결하지 못하고 있는 문제이다. 본 논문에서는 단어로서 가능한지를 결정하는 기준으로 확률적 음절 결합 정보를 사용하여 형태소 분석 기법만을 사용했을 때 발생할 수 있는 미등록어 문제를 해결하고, 통사적 기능의 어말 어휘를 고려한 문맥 결합 정보를 이용함으로써 다수의 후보 어절 가운데에서 최적의 후보 어절을 선택하는 방법을 제안한다. 제안된 시스템은 인식기에서 내보낸 후보 음절과 학습된 혼동 음절을 조합하여 하나 이상의 후보 어절을 생성하는 모듈과 통계적 언어 정보를 이용하여 최적의 후보 어절을 선정하는 모듈로 구성되었다. 실험은 1000만 원시 코퍼스에서 추출한 음절 결합 정보와 17만 태깅된 코퍼스에서 추출한 어절 결합 정보를 사용하였으며, 실제 인식 결과에 적용한 결과 문자 단위에서는 94.1%의 인식률을 97.4%로, 어절 단위에서는 87.6%를 96.6%로 향상시켰다. 교정률과 오교정률은 각각 문자 단위에서 56%와 0.6%, 어절 단위에서 83.9%와 1.66%를 보였으며, 전체 실험 어절의 3.4%를 차지한 미등록어 중 87.5%를 올바로 인식하는 한편, 전체 오류의 20.3%인 비문맥 오류에 대해서 91.6%를 올바로 교정하는 후처리 성능을 보였다.

• Journal of Korea Society of Industrial Information Systems
• /
• v.12 no.4
• /
• pp.15-24
• /
• 2007

The recognition of postal address images is indispensable for the automatic sorting of postal envelopes. The process of the address image recognition is composed of three steps-address image preprocessing, character recognition, address interpretation. The extracted character images from the preprocessing step are forwarded to the character recognition step, in which multiple candidate characters with reliability scores are obtained for each character image extracted. aracters with reliability scores are obtained for each character image extracted. Utilizing those character candidates with scores, we obtain the final valid address for the input envelope image through the address interpretation step. The envelope sorting rate depends on the performance of all three steps, among which character recognition step could be said to be very important. The good character recognizer would be the one which could produce valid candidates with very reliable scores to help the address interpretation step go easy. In this paper, we propose the method of generating character candidates with reliable recognition scores. We utilize the existing MLP(multilayered perceptrons) neural network of the address recognition system in the current automatic postal envelope sorters, as the classifier for the each image from the preprocessing step. The MLP is well known to be one of the best classifiers in terms of processing speed and recognition rate. The false alarm problem, however, might be occurred in recognition results, which made the address interpretation hard. To make address interpretation easy and improve the envelope sorting rate, we propose promising methods to reestimate the recognition score (confidence) of the existing MLP classifier: the generation method of the statistical recognition properties of the classifier and the method of the combination of the MLP and the subspace classifier which roles as a reestimator of the confidence. To confirm the superiority of the proposed method, we have used the character images of the real postal envelopes from the sorters in the post office. The experimental results show that the proposed method produces high reliability in terms of error and rejection for individual characters and non-characters.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.3
• /
• pp.75-81
• /
• 2011

Protein molecules are combined with another ones which have similar shapes at pocket positions. The pocket positions can be good references to describe the shapes of protein molecules. Harris corner detector is commonly used to detect feature points of 2 or 3D objects. Feature points can be found on the pocket areas and the points which have high derivatives. Generally speaking, the densities of feature points are relatively high at pocket areas because the shapes of pockets are concave. The pocket areas can be decided by the subdivision of voxel cubes which include feature points. The Euclidean distances between feature points and the central coordinate of the decided pocket area are calculated and sorted. The graph of sorted distances describes the shape of a protein molecule and the distribution of feature points. Therefore, it can be used to classify protein molecules by their shapes. Even though the shapes of protein molecules have been distorted with noises, they can be recognized with the accuracy more than 95 %. The accurate shape recognition provides the information to predict the binding properties of protein molecules.

• Journal of Broadcast Engineering
• /
• v.20 no.6
• /
• pp.938-949
• /
• 2015

In this paper, we propose a robust speed-limit sign recognition system which is durable to any sign changes caused by exterior damage or color contrast due to light direction. For recognition of speed-limit sign, we apply CNN which is showing an outstanding performance in pattern recognition field. However, original CNN uses multiple hidden layers to extract features and uses fully-connected method with MLP(Multi-layer perceptron) on the result. Therefore, the major demerit of conventional CNN is to require a long time for training and testing. In this paper, we apply randomly-connected classifier instead of fully-connected classifier by combining random forest with output of 2 layers of CNN. We prove that the recognition results of CNN with random forest show best performance than recognition results of CNN with SVM (Support Vector Machine) or MLP classifier when we use eight speed-limit signs of GTSRB (German Traffic Sign Recognition Benchmark).