• The Journal of the Convergence on Culture Technology
• /
• v.8 no.4
• /
• pp.339-345
• /
• 2022

Cardiovascular diseases is one of the leading causes of death in the world. The objectives of this study were to build various models using sociodemographic variables based on three variable selection methods and seven machine learning algorithms for the identification of hypertension and dyslipidemia and to evaluate predictive powers of the models. In experiments based on full variables and correlation-based feature subset selection methods, our results showed that performance of models using naive Bayes was better than those of models using other machine learning algorithms in both two diseases. In wrapper-based feature subset selection method, performance of models using logistic regression was higher than those of models using other algorithms. Our finding may provide basic data for public health and machine learning fields.

• Food Engineering Progress
• /
• v.15 no.2
• /
• pp.136-142
• /
• 2011

The objective of this research was to classify intact soybeans and foreign objects using near-infrared (NIR) spectroscopy. Intact soybeans and foreign objects were scanned using a NIR spectrometer equipped with scanning monochromator. NIR spectra of intact soybeans and foreign objects in the wavelength range from 900 to 1800 nm were collected. The classification of intact soybeans and foreign objects were conducted by using partial least-square discriminant analysis (PLS-DA) and soft independent modelling of class analogy (SIMCA) multivariate methods. Various types of data pretreatments were tested to develop the classification models. Intact soybeans and foreign objects were successfully classified by the PLS-DA prediction model with mean normalization pretreatment. These results showed the potential of NIR spectroscopy combined with multivariate analysis as a method for classifying intact soybeans and foreign objects.

• Journal of Korea Society of Industrial Information Systems
• /
• v.23 no.1
• /
• pp.87-96
• /
• 2018

In this paper, we propose a method to determine whether there is a defect by using the similarity between ROIs (Region of Interest) of the standard image and ROIs of the image which is corrected in position and rotation after capturing the vehicle headlight. The degree of similarity is determined by the template matching based on the histogram of image, which is a some modification of the method provided by OpenCV where template matching is performed on the raw image not the histogram. The proposed method is compared with the basic method of OpenCV for performance analysis. As a result of the analysis, it was found that the proposed method showed better performance than the OpenCV method, showing the accuracy close to 100%.

• The Journal of Korean Institute of Next Generation Computing
• /
• v.13 no.2
• /
• pp.18-25
• /
• 2017

In computer science, an encoding is a standardization of converting information to one format for audio, video or text. Therefore, the encoding information of the data should be known to open and read it and there are algorithms detecting encoder of the data. However, some informations of data could be disappeared by packet loss when transmitted on network, especially, if the data is snatched by packet sniffing or eavesdropping from wireless communications. In this paper, we improve the performance of encoding detection algorithm of 'uchardet' program for multi-byte encoded data with errors based on bit-shift algorithm. To simulate the performance, we generated Korean and Japanese text data with errors that is removed some random bits at random positions. Then the detection algorithm are tested using the data and 'uchardet-bitshift' showed better performance than 'uchardet'. When Korean texts are used, 'uchardet' could detect perfectly with ≤0.005% errors but it showed 0% detection rate with ≥1% errors while 'uchardet-bitshift' detected perfectly with ≤0.05% errors and it showed correct detection cases with ≥1% errors. Japanese texts with errors tend to report falsely as Chinese encoding because Japanese texts include lots of Chinese characters. As a results, we improved encoding detection algorithms by applying bit shift operation.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2023.05a
• /
• pp.513-514
• /
• 2023

초해상은 저해상도의 영상을 고해상도 영상으로 합성하는 기술이다. 이 기술에 딥러닝이 적용되어, 2014년에는 SRCNN(Super Resolution Convolutional Neural Network) 모델이 발표됐다. 이후에는 SRCAE(Super Resolution Convolutional Autoencoders)와 GAN(Generative Adversarial Networks)을 기반으로 한 SRGAN(Super Resolution Generative Adversarial Networks) 등, SRCNN의 성능을 능가하는 모델들이 발표됐다. ESRGAN(Enhanced Super Resolution Generative Adversarial Networks)은 SRGAN 모델의 성능을 개선했지만, 완벽한 성능을 내지 못하는 문제점이 있다. 이에 본 논문에서는 판별자(Discriminator) 구조를 변경하여 ESRGAN의 성능을 개선한다. 실험 결과, 제안하는 모델이 ESRGAN보다 더 높은 성능을 보일 것으로 기대된다.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.31 no.1
• /
• pp.49-55
• /
• 2013

In this paper, we proposed a good classifier to match different spatial data sets by applying evaluation of classifiers performance in data mining and biometrics. For this, we calculated distances between a pair of candidate features for matching criteria, and normalized the distances by Min-Max method and Tanh (TH) method. We defined classifiers that shape similarity is derived from fusion of these similarities by CRiteria Importance Through Intercriteria correlation (CRITIC) method, Matcher Weighting method and Simple Sum (SS) method. As results of evaluation of classifiers performance by Precision-Recall (PR) curve and area under the PR curve (AUC-PR), we confirmed that value of AUC-PR in a classifier of TH normalization and SS method is 0.893 and the value is the highest. Therefore, to match different spatial data sets, we thought that it is appropriate to a classifier that distances of matching criteria are normalized by TH method and shape similarity is calculated by SS method.

• Journal of Broadcast Engineering
• /
• v.19 no.2
• /
• pp.166-173
• /
• 2014

Recently, human action recognition have been developed for various broadcasting and video process. Since a video can consist of various scenes, keypoint approaches have been more attracted than template based methods for real application. Keypoint approahces tried to find regions having motion in video, and made 3-dimensional patches. Then, descriptors using histograms were computed from the patches, and a classifier based on machine learning method was applied to detect actions in video. However, a single classifier was difficult to handle various human actions. In order to improve this problem, approaches using multi classifiers were used to detect and to recognize objects. Thus, we propose a new human action recognition using decision-level fusion with support vector machine and sparse representation. The proposed method extracted descriptors based on keypoint approach from a video, and acquired results from each classifier for human action recognition. Then, we applied weights which were acquired by training stage to fuse each results from two classifiers. The experiment results in this paper show better result than a previous fusion method.

• Journal of Korea Society of Industrial Information Systems
• /
• v.19 no.2
• /
• pp.73-83
• /
• 2014

Thanks to the development of diverse audio editing Technology, audio file can be easily revised. As a result, diverse social problems like forgery may be caused. Digital forensic technology is actively studied to solve these problems. In this paper, a hand-held device identification method, an area of digital forensic technology is proposed. It uses the noise features of devices caused by the design and the integrated circuit of each device but cannot be identified by the audience. Wiener filter is used to get the noise sounds of devices and their acoustic features are extracted via MIRtoolbox and then they are trained by multi-layer neural network. To evaluate the proposed method, we use 5-fold cross-validation for the recorded data collected from 6 mobile devices. The experiments show the performance 99.9%. We also perform some experiments to observe the noise features of mobile devices are still useful after the data are uploaded to UCC. The experiments show the performance of 99.8% for UCC data.

• KIPS Transactions on Software and Data Engineering
• /
• v.8 no.1
• /
• pp.37-44
• /
• 2019

With the advance of digital IT technology, the performance of the printing and scanning devices is improved and their price becomes cheaper. As a result, the public can easily access these devices for crimes such as forgery of official and private documents. Therefore, if we can identify which printing device is used to print the documents, it would help to narrow the investigation and identify suspects. In this paper, we propose a deep learning model for printer identification. A convolutional neural network model based on local features which is widely used for identification in recent is presented. Then, another model including a step to calculate global features and hence improving the convergence speed and accuracy is presented. Using 8 printer models, the performance of the presented models was compared with previous feature-based identification methods. Experimental results show that the presented model using local feature and global feature achieved 97.23% and 99.98% accuracy respectively, which is much better than other previous methods in accuracy.

• The Journal of the Korea Contents Association
• /
• v.14 no.3
• /
• pp.40-51
• /
• 2014

Owing to importance of document copyright the need to detect document duplication and plagiarism is increasing. Many studies have sought to meet such need, but there are difficulties in document duplication detection due to technological limitations with the processing of natural language. This thesis designs and implements a discriminator of similar documents with natural language processing technique. This system discriminates similar documents using morphological analysis, syntactic analysis, and weight on low frequency and idiom. To evaluate the system, we analyze the correlation between human discrimination and term-based discrimination, and between human discrimination and proposed discrimination. This analysis shows that the proposed discrimination needs improving. Future research should work to define the document type and improve the processing technique appropriate for each type.