• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1997.11a
• /
• pp.201-204
• /
• 1997

최근 주소를 자동으로 인식하여 우편물 분류와 같은 업무를 효과적으로 수행하기 위한 연구가 진행되고 있다. 기존 연구들은 낱자 단위의 인식을 수행한 후 사전 형태의 간단한 DB를 통해 최종의 결과를 생성한다. 그러나 한글과 같은 복잡한 구조의 필기 문자에 대한 인식기의 성능은 아직도 미흡한 상태이다. 따라서 낱자 인식기의 성능에 의존하는 현재와 같은 방법으로는 만족할 만한 결과를 얻기가 힘들 것으로 생각된다. 본 논문에서는 낱자 인식 결과에 크게 의존하지 않고 주소에 나타나는 단어의 낱자들 사이간 연결 정보를 이용하여 단어를 인식할 수 있는 시스템을 제안한다. 본 시스템은 통계적 인식기를 사용하여 낱자를 인식하는 부분과 낱자 인식 결과를 조합하여 단어 수준의 인식과정을 통해 최종의 결과를 생성하는 부분으로 구성된다. 통계적 인식기는 Nearest neighborhood 방법을 사용하여 간단한 형태로 구현하였다. 단어인식 모듈은 단어에서 모든 문자간의 관계를 표현할 수 있도록 HMM 모형을 사용하여 어휘정보 네트워크를 구성하고 이를 이용하여 주소에 나타나는 단어를 인식하도록 하였다. PE92 한글 문자 데이터를 이용하여 실험을 수 璿\ulcorner 결과, 통계적 인식기의 성능이 저조함에도 불구하고 HMM을 이용한 어휘정보 네트워크가 이를 보완함으로써 좋은 결과를 얻었다. 이러한 단어 인식 방법을 주소 이외의 다른 단어 집합에 대해서도 쉽게 적용될 수 있을 것으로 예상된다.

• Journal of Korea Multimedia Society
• /
• v.14 no.5
• /
• 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.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.7
• /
• pp.1907-1928
• /
• 2024

In recent years, multi-focus image fusion has emerged as a prominent area of research, with transformers gaining recognition in the field of image processing. Current approaches encounter challenges such as boundary artifacts, loss of detailed information, and inaccurate localization of focused regions, leading to suboptimal fusion outcomes necessitating subsequent post-processing interventions. To address these issues, this paper introduces a novel multi-focus image fusion technique leveraging the Swin Transformer architecture. This method integrates a frequency layer utilizing Wavelet Transform, enhancing performance in comparison to conventional Swin Transformer configurations. Additionally, to mitigate the deficiency of local detail information within the attention mechanism, Convolutional Neural Networks (CNN) are incorporated to enhance region recognition accuracy. Comparative evaluations of various fusion methods across three datasets were conducted in the paper. The experimental findings demonstrate that the proposed model outperformed existing techniques, yielding superior quality in the resultant fused images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.751-770
• /
• 2019

Action recognition is an essential task in computer vision due to the variety of prospective applications, such as security surveillance, machine learning, and human-computer interaction. The availability of more video data than ever before and the lofty performance of deep convolutional neural networks also make it essential for action recognition in video. Unfortunately, limited crafted video features and the scarcity of benchmark datasets make it challenging to address the multi-person action recognition task in video data. In this work, we propose a deep convolutional neural network-based Effective Hybrid Learning (EHL) framework for two-person interaction classification in video data. Our approach exploits a pre-trained network model (the VGG16 from the University of Oxford Visual Geometry Group) and extends the Faster R-CNN (region-based convolutional neural network a state-of-the-art detector for image classification). We broaden a semi-supervised learning method combined with an active learning method to improve overall performance. Numerous types of two-person interactions exist in the real world, which makes this a challenging task. In our experiment, we consider a limited number of actions, such as hugging, fighting, linking arms, talking, and kidnapping in two environment such simple and complex. We show that our trained model with an active semi-supervised learning architecture gradually improves the performance. In a simple environment using an Intelligent Technology Laboratory (ITLab) dataset from Inha University, performance increased to 95.6% accuracy, and in a complex environment, performance reached 81% accuracy. Our method reduces data-labeling time, compared to supervised learning methods, for the ITLab dataset. We also conduct extensive experiment on Human Action Recognition benchmarks such as UT-Interaction dataset, HMDB51 dataset and obtain better performance than state-of-the-art approaches.

• Journal of the Korea Institute of Military Science and Technology
• /
• v.24 no.2
• /
• pp.175-186
• /
• 2021

In the field of automatic target recognition(ATR) with synthetic aperture radar(SAR) imagery, it is usually impractical to obtain SAR target images covering a full range of aspect views. When the database consists of SAR target images with limited angular diversity, it can lead to performance degradation of the SAR-ATR system. To address this problem, this paper proposes a deep learning-based method where channel attention modules(CAMs) are inserted to a convolutional neural network(CNN). Motivated by the idea of the squeeze-and-excitation(SE) network, the CAM is considered to help improve recognition performance by selectively emphasizing discriminative features and suppressing ones with less information. After testing various CAM types included in the ResNet18-type base network, the SE CAM and its modified forms are applied to SAR target recognition using MSTAR dataset with different reduction ratios in order to validate recognition performance improvement under the limited angular diversity condition.

• The Journal of the Acoustical Society of Korea
• /
• v.41 no.6
• /
• pp.647-654
• /
• 2022

Although the fields in which is utilized children's speech recognition is on the rise, the lack of quality data is an obstacle to improving children's speech recognition performance. This paper proposes a new method for improving children's speech recognition performance by additionally using adult speech data. The proposed method is a transformer based domain adversarial training using dynamically weighted loss to effectively address the data imbalance gap between age that grows as the amount of adult training data increases. Specifically, the degree of class imbalance in the mini-batch during training was quantified, and the loss function was defined and used so that the smaller the data, the greater the weight. Experiments validate the utility of proposed domain adversarial training following asymmetry between adults and children training data. Experiments show that the proposed method has higher children's speech recognition performance than traditional domain adversarial training method under all conditions in which asymmetry between age occurs in the training data.

• ETRI Journal
• /
• v.35 no.5
• /
• pp.869-879
• /
• 2013

This study is related to the acquisition of physiological signals of human emotions and the recognition of human emotions using such physiological signals. To acquire physiological signals, seven emotions are evoked through stimuli. Regarding the induced emotions, the results of skin temperature, photoplethysmography, electrodermal activity, and an electrocardiogram are recorded and analyzed as physiological signals. The suitability and effectiveness of the stimuli are evaluated by the subjects themselves. To address the problem of the emotions not being recognized, we introduce a methodology for a recognizer using prototype-based learning and particle swarm optimization (PSO). The design involves two main phases: i) PSO selects the P% of the patterns to be treated as prototypes of the seven emotions; ii) PSO is instrumental in the formation of the core set of features. The experiments show that a suitable selection of prototypes and a substantial reduction of the feature space can be accomplished, and the recognizer formed in this manner is characterized by high recognition accuracy for the seven emotions using physiological signals.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2015.11a
• /
• pp.64-65
• /
• 2015

As construction projects have increased in size and height recently, lifting accounts for increasingly greater portion and tower cranes are used more frequently. At present, the selection and arrangement of tower crane are depend on the experience of experts. However, since the number of experts is fairly limited and a database for tower cranes regarding lifting capacity, operation properties, rent, etc has not been widely employed, tower cranes are often not effectively selected and arranged which can cause cost overruns and delays in the lifting work. To address such issues, this study attempts to perform a basic study for development of automatic arrangement algorithm of tower crane using drawing recognition. If relevant database is established and the algorithm suggested in this study is refined more systematically, even beginning level engineers will be able to plan tower crane arrangement in a way comparable to experienced experts.

• International Journal of Contents
• /
• v.10 no.3
• /
• pp.26-34
• /
• 2014

Mobile learning is a practical learning method for busy adult learners because the mobility of digital devices can overcome the drawbacks of e-learning. However, research is strongly lacking in the theoretical exploration of mobile learning effects and functions and its empirical research. Moreover, the research of learning characteristics and learners' requirements must be considered before applying and disseminating mobile learning into the educational field. To address this shortcoming, this study conducted an online survey with 1,542 officers of the Ministry of Health and Welfare Affairs (MHWA) regarding learner recognition to mobile learning. The analysis of learners' attitudes toward mobile learning, based on age and position, indicated that subordinate workers appeared to place more value on mobile learning. Many participants preferred mobile learning because of its mobility and the effectiveness of anywhere and anytime. However, some participants continue to misunderstand mobile learning and its necessity. Therefore, consideration of learning effectiveness, the form of the content, and learner-centered learning must be reviewed in advance. This study could lead to practical implications of mobile learning.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.9
• /
• pp.4412-4428
• /
• 2018

The purpose of this work is to solve the problem of representing an entire video using Convolutional Neural Network (CNN) features for human action recognition. Recently, due to insufficient GPU memory, it has been difficult to take the whole video as the input of the CNN for end-to-end learning. A typical method is to use sampled video frames as inputs and corresponding labels as supervision. One major issue of this popular approach is that the local samples may not contain the information indicated by the global labels and sufficient motion information. To address this issue, we propose a binary hashing method to enhance the local feature extractors. First, we extract the local features and aggregate them into global features using maximum/minimum pooling. Second, we use the binary hashing method to capture the motion features. Finally, we concatenate the hashing features with global features using different normalization methods to train the classifier. Experimental results on the JHMDB and MPII-Cooking datasets show that, for these new local features, binary hashing mapping on the sparsely sampled features led to significant performance improvements.