• The Korean Journal of Applied Statistics
• /
• v.32 no.1
• /
• pp.149-160
• /
• 2019

The electricity consumption time series data of 'A' University from July 2016 to June 2017 is analyzed via nonparametric functional data clustering since the time series data can be regarded as realization of continuous functions with dependency structure. We use a Bouveyron and Jacques (Advances in Data Analysis and Classification, 5, 4, 281-300, 2011) method based on model-based functional clustering with an FEM algorithm that assumes a Gaussian distribution on functional principal components. Clusterwise analysis is provided with cluster mean functions, densities and cluster profiles.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.13 no.4
• /
• pp.765-773
• /
• 2009

Currently Telematics traffic information services have been various because we can collect real-time traffic information through Intelligent Transport System. In this paper, we proposed and implemented a short-term traffic information prediction model for giving to guarantee the traffic information with high quality in the near future. A Short-term prediction model is for forecasting traffic flows of each segment in the near future. Our prediction model gives an average speed on the each segment from 5 minutes later to 60 minutes later. We designed a Bayesian network for each segment with some casual nodes which makes an impact to the road situation in the future and found out its joint probability density function on the supposition of GMM(Gaussian Mixture Model) using EM(Expectation Maximization) algorithm with training real-time traffic data. To validate the precision of our prediction model we had conducted various experiments with real-time traffic data and computed RMSE(Root Mean Square Error) between a real speed and its prediction speed. As the result, our model gave 4.5, 4.8, 5.2 as an average value of RMSE about 10, 30, 60 minutes later, respectively.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.46 no.12
• /
• pp.1049-1055
• /
• 2018

This paper proposes the Signal Quality Indicator (SQI) to certify the wired communication channels connected between sub-systems embedded in guided flight systems. The communication signals can be distorted due to the poor interconnections of wired channels, the interference from the signals passing through the adjacent routed channels, and additive white Gaussian noises. As the ways to find out the condition of the communication channels, we present the Hamming distance based SQI (H-SQI) and the Euclidean distance based SQI (E-SQI). Two SQIs are compared in terms of the SQI resolution performance and the required number of hardware resources for implementations. The E-SQI requires the 10 times FPGA resources and an additional analog-digital converter over the H-SQI in spite of its outstanding SQI resolution performance. Moreover, the H-SQI could have the enough SQI resolution performance to find out the channel condition by increasing the oversampling rate, so the H-SQI is the more adequate than the E-SQI for the SQI of the guided flight systems.

• The KIPS Transactions:PartB
• /
• v.16B no.5
• /
• pp.385-394
• /
• 2009

Face tracking and recognition are difficult problems because the face is a non-rigid object. The main reasons for the failure to track and recognize the faces are the changes of a face pose and environmental illumination. To solve these problems, we propose a nonlinear manifold framework for the face pose and the face illumination normalization processing. Specifically, to track and recognize a face on the video that has various pose variations, we approximate a face pose density to single Gaussian density by PCA(Principle Component Analysis) using images sampled from training video sequences and then construct the GMM(Gaussian Mixture Model) for each person. To solve the illumination problem for the face tracking and recognition, we decompose the face images into the reflectance and the illuminance using the SSR(Single Scale Retinex) model. To obtain the normalized reflectance, the reflectance is rescaled by histogram equalization on the defined range. We newly approximate the illuminance by the trained manifold since the illuminance has almost variations by illumination. By combining these two features into our manifold framework, we derived the efficient face tracking and recognition results on indoor and outdoor video. To improve the video based tracking results, we update the weights of each face pose density at each frame by the tracking result at the previous frame using EM algorithm. Our experimental results show that our method is more efficient than other methods.

• Journal of the Korean earth science society
• /
• v.32 no.3
• /
• pp.249-264
• /
• 2011

A multi-Gaussian kriging approach extended to space-time domain is presented for uncertainty modeling as well as time-series mapping of environmental variables. Within a multi-Gaussian framework, normal score transformed environmental variables are first decomposed into deterministic trend and stochastic residual components. After local temporal trend models are constructed, the parameters of the models are estimated and interpolated in space. Space-time correlation structures of stationary residual components are quantified using a product-sum space-time variogram model. The ccdf is modeled at all grid locations using this space-time variogram model and space-time kriging. Finally, e-type estimates and conditional variances are computed from the ccdf models for spatial mapping and uncertainty analysis, respectively. The proposed approach is illustrated through a case of time-series Particulate Matter 10 ($PM_{10}$) concentration mapping in Incheon Metropolitan city using monthly $PM_{10}$ concentrations at 13 stations for 3 years. It is shown that the proposed approach would generate reliable time-series $PM_{10}$ concentration maps with less mean bias and better prediction capability, compared to conventional spatial-only ordinary kriging. It is also demonstrated that the conditional variances and the probability exceeding a certain thresholding value would be useful information sources for interpretation.

• Journal of the Institute of Convergence Signal Processing
• /
• v.22 no.3
• /
• pp.141-148
• /
• 2021

In this paper, we propose an algorithm of welding bead detection and evaluation using geodesic active contour algorithm and high pass filter with image processing technique. The algorithm uses histogram equalization and high pass filter as gaussian filter to improve contrast. The image processing techniques smoothens the welding beads reduce the noise on an image. Then, the algorithm detects the welding bead area by applying the geodesic active contour algorithm and morphological ooperation. It also applies the balloon force that either inflates in, or deflates out the evolving contour for a better segmentation. After that, we propose a method for determining the quality of welding bead using effective length and width of the detected bead. In the experiments, our algorithm achieved the highest recall, precision, F-measure and IOU as 0.9894, 0.9668, 0.9780, and 0.8957 respectively. We compared the proposed algorithm with the conventional algorithms to evaluate the performance of the proposed algorithm. The proposed algorithm achieved better performance compared to the conventional ones with a maximum computational time of 0.6 seconds for segmenting and evaluating one welding bead.

• The Journal of the Acoustical Society of Korea
• /
• v.30 no.8
• /
• pp.446-460
• /
• 2011

The use of induction motors has been recently increasing with automation in aeronautical and automotive industries, and it playes a significant role. This has motivated that many researchers have studied on developing fault detection and classification systems of an induction motor in order to minimize economical damage caused by its fault. With this reason, this paper proposed feature vector extraction methods based on STE (short-time energy)+SVD (singular value decomposition) and DCT (discrete cosine transform)+SVD techniques to early detect and diagnose faults of induction motors, and classified faults of an induction motor into different types of them by using extracted features as inputs of BPNN (back propagation neural network) and multi-layer SVM (support vector machine). When BPNN and multi-lay SVM are used as classifiers for fault classification, there are many settings that affect classification performance: the number of input layers, the number of hidden layers and learning algorithms for BPNN, and standard deviation values of Gaussian radial basis function for multi-layer SVM. Therefore, this paper quantitatively simulated to find appropriate settings for those classifiers yielding higher classification performance than others.

• The Korean Journal of Nuclear Medicine
• /
• v.39 no.1
• /
• pp.49-56
• /
• 2005

Purpose: Dedicated animal PET is useful equipment for the study of new PET tracer. recently, microPET R4 was installed in the Korea institute of radiology and medical science. In this study, we measured the characteristics of scanner. Materials and methods: Resolution was measured using a line source (F-18:65 ${\mu}Ci$, inner diameter: 0.5 mm). The line source was put in the axial direction and was moved from the center of field of view to outside with 1 mm interval. PET images were reconstructed using a filtered back-protection and ordered subset expectation maximization. line source (16.5 ${\mu}Ci$, 78 mm) was put on the tenter of axial direction to measure the sensitivity when the deadtime was under 1%. Images were acquired during 4 minutes respectively from center to 39 mm outward. Delayed count was subtracted from total count and then decay was corrected for the calculation of sensitivity. Noise equivalent count ratio and scatter fraction were calculated using cylindrical phantom. Results: Spatial resolution of reconstructed image using filtered back-projection was 1.86 mm(radial), 1.95 mm(tangential), 1.95 mm(axial) in the tenter of field of view, and 2.54 mm, 2.8 mm, 1.61 mm in 2 cm away from the center respectively. Sensitivity was 2.36% at the center of transaxial field of view. Scatter fraction was 20%. Maximal noise equivalent count ratio was 66.4 kcps at 242 kBq/mL. Small animal images were acquired for confirmation of performance. Conclusion: Performance characteristics of microPET R4 were similar with reported value. So this will be a useful tool for small animal imaging.

• The Journal of Bigdata
• /
• v.4 no.2
• /
• pp.1-12
• /
• 2019

As Artificial Intelligence (AI) technology develops, it is applied to various fields such as image, voice, and text. AI has shown fine results in certain areas. Researchers have tried to predict the stock market by utilizing artificial intelligence as well. Predicting the stock market is known as one of the difficult problems since the stock market is affected by various factors such as economy and politics. In the field of AI, there are attempts to predict the ups and downs of stock price by studying stock price patterns using various machine learning techniques. This study suggest a way of predicting stock price patterns based on the Convolutional Neural Network(CNN) among machine learning techniques. CNN uses neural networks to classify images by extracting features from images through convolutional layers. Therefore, this study tries to classify candlestick images made by stock data in order to predict patterns. This study has two objectives. The first one referred as Case 1 is to predict the patterns with the images made by the same-day stock price data. The second one referred as Case 2 is to predict the next day stock price patterns with the images produced by the daily stock price data. In Case 1, data augmentation methods - random modification and Gaussian noise - are applied to generate more training data, and the generated images are put into the model to fit. Given that deep learning requires a large amount of data, this study suggests a method of data augmentation for candlestick images. Also, this study compares the accuracies of the images with Gaussian noise and different classification problems. All data in this study is collected through OpenAPI provided by DaiShin Securities. Case 1 has five different labels depending on patterns. The patterns are up with up closing, up with down closing, down with up closing, down with down closing, and staying. The images in Case 1 are created by removing the last candle(-1candle), the last two candles(-2candles), and the last three candles(-3candles) from 60 minutes, 30 minutes, 10 minutes, and 5 minutes candle charts. 60 minutes candle chart means one candle in the image has 60 minutes of information containing an open price, high price, low price, close price. Case 2 has two labels that are up and down. This study for Case 2 has generated for 60 minutes, 30 minutes, 10 minutes, and 5minutes candle charts without removing any candle. Considering the stock data, moving the candles in the images is suggested, instead of existing data augmentation techniques. How much the candles are moved is defined as the modified value. The average difference of closing prices between candles was 0.0029. Therefore, in this study, 0.003, 0.002, 0.001, 0.00025 are used for the modified value. The number of images was doubled after data augmentation. When it comes to Gaussian Noise, the mean value was 0, and the value of variance was 0.01. For both Case 1 and Case 2, the model is based on VGG-Net16 that has 16 layers. As a result, 10 minutes -1candle showed the best accuracy among 60 minutes, 30 minutes, 10 minutes, 5minutes candle charts. Thus, 10 minutes images were utilized for the rest of the experiment in Case 1. The three candles removed from the images were selected for data augmentation and application of Gaussian noise. 10 minutes -3candle resulted in 79.72% accuracy. The accuracy of the images with 0.00025 modified value and 100% changed candles was 79.92%. Applying Gaussian noise helped the accuracy to be 80.98%. According to the outcomes of Case 2, 60minutes candle charts could predict patterns of tomorrow by 82.60%. To sum up, this study is expected to contribute to further studies on the prediction of stock price patterns using images. This research provides a possible method for data augmentation of stock data.

• Journal of the Korean Magnetics Society
• /
• v.14 no.2
• /
• pp.65-70
• /
• 2004

Read out signal and temperature distribution of magnetic amplifying magnetooptical disk were studied. Temperature distribution of recording layer and adjacent layers were calculated when the disk was at rest. Mark size, length and location were simulated from a chain of recording beam pulses. In addition, signal amplitude depending on the shape of the marks, readout signals from the recording layer and amplified marks of the readout layer, were simulated. Simplified thermal conduction model was used to calculate the temperature distribution of recording and adjacent layers as a function of time as well as to calculate the mark size, length and location. Readout signal was calculated by the convolution of the disk reflectivity and the Gaussian beam intensity. Readout signal from the mark in the readout layer amplified to the size of the laser beam fumed out to be twice as large as the signal from the crescent shaped mark in the recording layer.