• The Korean Journal of Applied Statistics
• /
• v.25 no.6
• /
• pp.1009-1018
• /
• 2012

Persistence homology (a type of methodology in computational algebraic topology) can be used to capture the topological characteristics of functional data. To visualize the characteristics, a persistence diagram is adopted by plotting baseline and the pairs that consist of local minimum and local maximum. We use the bottleneck distance to measure the topological distance between two different functions; in addition, this distance can be applied to multidimensional scaling(MDS) that visualizes the imaginary position based on the distance between functions. In this study, we use handwriting data (which has functional forms) to get persistence diagram and check differences between the observations by using bottleneck distance and the MDS.

• Journal of Korea Multimedia Society
• /
• v.11 no.6
• /
• pp.807-815
• /
• 2008

A simple efficient method of spotting and recognizing hand gestures in video is presented using a network of hidden Markov models and dynamic programming search algorithm. The description starts from designing a set of isolated trajectory models which are stochastic and robust enough to characterize highly variable patterns like human motion, handwriting, and speech. Those models are interconnected to form a single big network termed a spotting network or a spotter that models a continuous stream of gestures and non-gestures as well. The inference over the model is based on dynamic programming. The proposed model is highly efficient and can readily be extended to a variety of recurrent pattern recognition tasks. The test result without any engineering has shown the potential for practical application. At the end of the paper we add some related experimental result that has been obtained using a different model - dynamic Bayesian network - which is also a type of stochastic model.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2019.10a
• /
• pp.63-65
• /
• 2019

본 논문은 개발하고자 하는 기계학습 기반 한글 필기 인식 시스템의 첫 연구 결과를 담고 있다. 즉, 기계학습을 위해서는 학습용 및 테스트용 필기 데이터가 아주 많이 필요하므로, 이를 수집하고 전처리하는 방법을 제안하였다. 한글의 한 글자는 자음과 모음을 결합하여 생성되는데, 실제 만 개 이상의 글자가 생성될 수 있다. 따라서 각각의 글자 데이터를 수집하는 대신, 수집한 글자 데이터로부터 초성, 중성, 종성을 구분하여 최종적으로 자음, 모음 데이터로 저장하고자 한다. 아직 초기 연구이므로, 다양한 경우에 대한 분석이나 실험 결과는 없지만, 이를 활용하여 온라인 필기 인식 모델에 적용하여 인식 성능을 높이기 위한 추후 연구의 기반으로 활용하고자 한다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2000.04a
• /
• pp.890-895
• /
• 2000

본 논문에서는 인공신경망을 이용한 자모 검증을 HMM 기반 온라인 한글인식 시스템에 적용하는 방법론을 제시한다. 본 시스템에서 각각의 자모는 한 개의 HMM 모델과 한 개의 인공신경망 검증기를 갖는다. 자모 검증기는 HMM 네트웍이 생성한 자모 후보 가정을 입력으로 받은 후, 이 가정의 타당성에 대한 사후 확률을 출력한다. 이 사후 확률은 Viterbi 탐색시 탐색 경로에 반영된다. 기존 HMM 시스템의 국소적 특징의 한계를 보완하기 위하여, 한글 자모의 기본획 분석에서 얻어진 구조적, 전역적 특징이 자모 검증기에 사용되었다. 한글 낱자인식에 대한 실험 결과 HMM 기반 인식기에 자모 검증기를 도입함으로서 38.5%의 인식 오류를 줄일 수 있었다.

• Journal of the Korean Institute of Plant Engineering
• /
• v.23 no.4
• /
• pp.29-38
• /
• 2018

Through process quality information, the time required for process quality analysis has been drastically shortened, the process defect rate has been reduced, and the manufacturing lead time has been shortened and the on-time delivery rate has been improved. Therefore, The purpose of this study is to develop a quality information analysis system model that effectively shortens the time required for process quality analysis in automobile safety belt parts manufacturing process. As a result of experiments on communication operation between manufacturing execution system (MES) quality server, injection machine control computer, injection machine programmable logic controller (PLC) and terminal, in analyzing quality information, the conventional handwriting input method took an average of 20 minutes, but the new multi-network method took about 2 minutes on average. In addition, the process defect rate was reduced by 13% and the manufacturing lead time was shortened from 28 hours to 20 hours. The delivery compliance rate improved from 96 to 99%.

• Journal of Multimedia Information System
• /
• v.8 no.1
• /
• pp.31-34
• /
• 2021

The classical HMM is defined by a parameter triple �� = (��, A, B), where each parameter represents a collection of probability distributions: initial state, state transition and output distributions in order. This paper proposes a new stationary parameter e = (e1, e2, …, eN) where N is the number of states and et = P(|xt = i, y) for describing how an input pattern y ends in state xt = i at time t followed by nothing. It is often said that all is well that ends well. We argue here that all should end well. The paper sets the framework for the theory and presents an efficient inference and training algorithms based on dynamic programming and expectation-maximization. The proposed model is applicable to analyzing any sequential data with two or more finite segmental patterns are concatenated, each forming a context to its neighbors. Experiments on online Hangul handwriting characters have proven the effect of the proposed augmentation in terms of highly intuitive segmentation as well as recognition performance and 13.2% error rate reduction.

• Journal of Information Processing Systems
• /
• v.18 no.2
• /
• pp.282-291
• /
• 2022

This research applies a pre-trained bidirectional encoder representations from transformers (BERT) handwriting recognition model to predict foreign Korean-language learners' writing scores. A corpus of 586 answers to midterm and final exams written by foreign learners at the Intermediate 1 level was acquired and used for pre-training, resulting in consistent performance, even with small datasets. The test data were pre-processed and fine-tuned, and the results were calculated in the form of a score prediction. The difference between the prediction and actual score was then calculated. An accuracy of 95.8% was demonstrated, indicating that the prediction results were strong overall; hence, the tool is suitable for the automatic scoring of Korean written test answers, including grammatical errors, written by foreigners. These results are particularly meaningful in that the data included written language text produced by foreign learners, not native speakers.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.7
• /
• pp.1098-1101
• /
• 2022

With the development of computer music notation programs, when drawing sheet music, it is often drawn using a computer. However, there are still many use of hand-written notations for educational purposes or to quickly draw sheet music such as listening and dictating. In previous studies, OMR focused on recognizing the printed music sheet made by music notation program. the result of handwritten OMR with camera is poor because different people have different writing methods, and lens distortion. In this study, as a pre-processing process for recognizing handwritten music sheet, we propose a method for recognizing a staff using linear regression and a method for recognizing a bar using CNN. F₁ scores of staff recognition and barline detection are 99.09% and 95.48%, respectively. This methodologies are expected to contribute to improving the accuracy of handwriting.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2022.11a
• /
• pp.225-227
• /
• 2022

다양한 분야에서 전자기기들을 사용함으로 인해 문서를 작성할 때 디지털 글꼴을 통해 작성하게 되는데, 이로 인해 글꼴을 종류가 여러 형태로 증가하면서 다양한 글꼴들을 사용하고 있다. 하지만, 글꼴마다 저작권을 가지고 있어서 마음에 든다고 해서 함부로 사용할 수도 없는 것이 문제점이다. 또한, 한글은 다른 언어에 비해 글자 조합방식이 많아서 폰트로 제작하기엔 많은 시간과 비용이 든다는 문제도 있다. 이러한 문제들을 해결하기 위해서 딥러닝을 통해 글꼴을 제작하게 된다면 적은 글자를 입력해 많은 글자의 결과를 도출함으로써, 시간과 비용을 절감해 효율적으로 만들고자 하였다. 이에 본 논문은 GAN을 기반으로 한 손글씨 폰트 제작을 하는 가운데 글꼴을 만들기 위해 입력에 어떤 글자들이 필요한 지에 대해 연구하였다. 다양한 분석적 요소를 갖고 실험을 하여 입력에 따라 결과가 어떻게 달라지는지를 알아보았고 이를 바탕으로 글꼴을 생성하였다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.561-562
• /
• 2021

CNN (Convolutional Neural Network) is widely used to recognize various images. In this presentation, a single digit handwritten by humans was recognized by applying the CNN technique of deep learning. The deep learning network consists of a convolutional layer, a pooling layer, and a platen layer, and finally, we set an optimization method, learning rate and loss functions.