• 정보학연구
• /
• 제9권4호
• /
• pp.65-74
• /
• 2006

The back propagation of neural networks has the problems of falling into local minimum and delay of the speed by the iterative learning. An algorithm to solve the problem and improve the speed of the learning was already proposed in[8], which updates the learning parameter related with the connection weight. In this paper, we propose the algorithm generating initial weight to improve the efficiency of the algorithm by offering the difference between the input vector and the target signal to the generating function of initial weight. The algorithm proposed here can classify more than 98.75% of the handwritten digits and this rate shows 30% more effective than the other previous methods.

• Journal of Information Processing Systems
• /
• 제13권1호
• /
• pp.174-183
• /
• 2017

In this paper, we propose a novel feature for recognizing handwritten Odia numerals. By using polygonal approximation, each numeral is segmented into segments of equal pixel counts where the centroid of the character is kept as the origin. Three primitive contour features namely, distance (l), angle (${\theta}$), and arc-tochord ratio (r), are extracted from these segments. These features are used in a neural classifier so that the numerals are recognized. Other existing features are also considered for being recognized in the neural classifier, in order to perform a comparative analysis. We carried out a simulation on a large data set and conducted a comparative analysis with other features with respect to recognition accuracy and time requirements. Furthermore, we also applied the feature to the numeral recognition of two other languages-Bangla and English. In general, we observed that our proposed contour features outperform other schemes.

• 한국멀티미디어학회논문지
• /
• 제14권5호
• /
• pp.642-649
• /
• 2011

Among the various aspects of offline handwritten character patterns, it is the great variety of writing styles and variations that renders the task of computer recognition very hard. The immense variety of character shape has been recognized but rarely studied during the past decades of numerous research efforts. This paper tries to address the problem of measuring image distortions and handwritten character patterns with respect to reference patterns. This work is based on mass-spring mesh model with the introduction of simulated electric charge as a source of the external force that can aid decoding the shape distortion. Given an input image and a reference image, the charge is defined, and then the relaxation procedure goes to find the optimum configuration of shape or patterns of least potential. The relaxation process is based on the fourth order Runge-Kutta algorithm, well-known for numerical integration. The proposed method of modeling is rigorous mathematically and leads to interesting results. Additional feature of the method is the global affine transformation that helps analyzing distortion and finding a good match by removing a large scale linear disparity between two images.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 The Third Asian Fuzzy Systems Symposium
• /
• pp.707-712
• /
• 1998

This paper presents several softcomputing techniques such as neural networks, fuzzy logic and genetic algorithms : Neural networks as brain metaphor provide fundamental structure, fuzzy logic gives a possibility to utilize top-down knowledge from designer, and genetic algorithms as evolution metaphor determine several system parameters with the process of bottom up development. With these techniques, we develop a pattern recognizer which consists of multiple neural networks aggregated by fuzzy integral in which genetic algorithms determine the fuzzy density values. The experimental results with the problem of recognizing totally unconstrained handwritten numeral show that the performance of the proposed method is superior to that of conventional methods.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 1998년도 하계종합학술대회논문집
• /
• pp.705-708
• /
• 1998

In this paper, a new algorithm is proposed that the template matching is used to devide SOFM (self-organizig feature map) for fast learning and to extract features for considering input data types. In order to verify the superoprity of the proposed algorithm, applied to the recognition of handwritten numerals. Templates of handwritten numerals are created by a line of external-contact.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 1999년도 추계종합학술대회 논문집
• /
• pp.1019-1022
• /
• 1999

In this paper, we analysis the effect of a dictionary in a handwritten Hangul word recognition problem in terms of its size and the length of the words in it. With our experimental results, we can account for the word recognition rate depending not only on character recognition performance, but also much on the amount of the information that the dictionary contains, as well as the reduction rate of a dictionary.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 1999년도 추계종합학술대회 논문집
• /
• pp.468-471
• /
• 1999

In this paper, we present a solution for combining multiple neural networks. Each neural network is trained with different features. And the neural networks are combined by four methods. The recognition rates by four combination methods are compared. The experimental results for handwritten digit recognition shows that the combination at hidden layers by single layer neural network is superior to any other methods. The reasons of the results are explained.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제30권10호
• /
• pp.904-911
• /
• 2003

본 논문에서는 우편봉투 상에 기입된 인쇄체 한글 주소와 필기체 한글 주소를 효과적으로 구분할 수 있는 방법을 제안한다. 문자인식 모듈을 포함하는 각종 응용 시스템에서 입력 영상이 인쇄체인지 필기체인지 구분하는 것은 매우 중요하다. 이는 대부분의 경우 인쇄체 영상과 필기체 영상이 갖는 특징이 상이하여, 각 영상에서의 문자 및 문자열 분리 방법, 문자 인식 방법 둥이 매우 상이하게 개발되기 때문이다. 본 논문에서 제안한 구분 방법은 연결요소 추출 및 병합, 특징 추출, 영상 구분 순으로 수행된다. 연결요소 추출 및 병합 단계에서는 입력영상으로부터 연결요소를 추출한 후 일부 연결요소들에 대하여 병합을 시도하며, 특징 추출 단계에서는 병합결과 얻어진 연결요소들의 그룹들로부터 폭과 위치에 관련된 특징을 추출하고, 영상 구분 단계에서는 추출한 특징을 입력으로 제공받는 다충퍼셉트론을 사용하여 구분을 시도한다. 제안한 방법의 우수성을 증명하기 위해 실제 우편물로부터 추출된 3,147개의 한글 주소 영상을 사용하여 실험한 결과, 98.85%의 구분률을 보여주었다.

• 한국인터넷방송통신학회논문지
• /
• 제10권1호
• /
• pp.63-68
• /
• 2010

최근, 유비쿼터스 시대로의 도약을 위하여 모바일 기기의 입력 인터페이스에 대한 연구가 활발하게 진행되고 있으며, 기존의 마우스, 키보드뿐만 아니라 필기, 음성, 시각, 터치와 같이 다분야로 세분화되어 새로운 인터페이스가 연구되고 있다. 특히 소형 모바일 기기에서는 크기로 인하여 추가가능한 장치의 제약이 심하기 때문에, 작은 화면에서도 효율적인 입력 인터페이스로 필기 인식이 대두되고 있다. 필기 인식에 대한 이전 연구는 2차원 영상을 기반으로 하거나 벡터로 입력받은 필기데이터를 인식하는 알고리즘이 대부분이며, 필기 인식 알고리즘의 정확성을 향상시키는 연구에만 초점을 두고 있는 실정이다. 또한 실제 필기 입력하는 경우에는 현재 문자가 영문 대/소문자, 한글, 숫자 등의 어느 분류에 속하는지 선택해야하는 문제가 있다. 따라서 본 논문에서는 입력된 필기 데이터를 기반으로 형상 분석을 통하여, 영문이나 한글의 여부를 판단하고 언어식별이 가능한 시스템을 제안하였다. 제안 기법은 벡터 단위의 집합으로 필기 데이터를 취급하여 각 벡터 간의 상호관계와 방향성을 분석함으로써 효율적인 언어식별을 가능하도록 하였다.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2004년도 학술대회지
• /
• pp.814-818
• /
• 2004

Off-line handwritten character recognition is in difficulty of incomplete preprocessing because it has not dynamic and timing information besides has various handwriting, extreme overlap of the consonant and vowel and many error image of stroke. Consequently off-line handwritten character recognition needs to study about preprocessing of various methods such as binarization and thinning. This paper considers running time of watershed algorithm and the quality of resulting image as preprocessing For off-line handwritten Korean character recognition. So it proposes application of effective watershed algorithm for segmentation of character region and background region in gray level character image and segmentation function for binarization image and segmentation function for binarization by extracted watershed image. Besides it proposes thinning methods which effectively extracts skeleton through conditional test mask considering running time and quality. of skeleton, estimates efficiency of existing methods and this paper's methods as running time and quality. Watershed image conversion uses prewitt operator for gradient image conversion, extracts local minima considering 8-neighborhood pixel. And methods by using difference of mean value is used in region merging step, Converted watershed image by means of this methods separates effectively character region and background region applying to segmentation function. Average execution time on the previous method was 2.16 second and on this paper method was 1.72 second. We prove that this paper's method removed noise effectively with overlap stroke as compared with the previous method.