• Journal of Korea Multimedia Society
• /
• 제24권7호
• /
• pp.880-889
• /
• 2021

In everyday life, recognizing people's emotions from their frames is essential and is a popular research domain in the area of computer vision. Visual emotion has a severe class imbalance in which most of the data are distributed in specific categories. The existing methods do not consider class imbalance and used accuracy as the performance metric, which is not suitable for evaluating the performance of the imbalanced dataset. Therefore, we proposed a method for recognizing visual emotion using balanced data augmentation to address the class imbalance. The proposed method generates a balanced dataset by adopting the random over-sampling and image transformation methods. Also, the proposed method uses the Focal loss as a loss function, which can mitigate the class imbalance by down weighting the well-classified samples. EfficientNet, which is the state-of-the-art method for image classification is used to recognize visual emotion. We compare the performance of the proposed method with that of conventional methods by using a public dataset. The experimental results show that the proposed method increases the F1 score by 40% compared with the method without data augmentation, mitigating class imbalance without loss of classification accuracy.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• 제20권3호
• /
• pp.1-7
• /
• 2021

With the advent of industry 4.0, big data and machine learning techniques are being widely adopted in the maintenance domain. Inverters are widely used in many engineering applications. However, overloading and complex operation conditions may lead to various failures in inverters. In this study, failure mode effect analysis was performed on inverters and voltages collected to investigate the over-voltage effect on capacitors. Several features were extracted from the collected sensor data, which indicated the health state of the inverter. Based on this correlation, the best features were selected for classification. Moreover, random forest classifiers were used to classify the healthy and faulty states of inverters. Different performance metrics were computed, and the classifiers' performance was evaluated in terms of various health features.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권9호
• /
• pp.2927-2941
• /
• 2022

The primary task of machine fault diagnosis is to judge whether the current state is normal or damaged, so it is a typical binary classification problem with mutual exclusion. Mutually exclusive events and out-of-domain detection have one thing in common: there are two types of data and no intersection. We proposed a fusion model method to improve the accuracy of machine fault diagnosis, which is based on the mutual exclusivity of events and the commonality of out-of-distribution detection, and finally generalized to all binary classification problems. It is reported that the performance of a convolutional neural network (CNN) will decrease as the recognition type increases, so the variational auto-encoder (VAE) is used as the primary model. Two VAE models are used to train the machine's normal and fault sound data. Two reconstruction probabilities will be obtained during the test. The smaller value is transformed into a correction value of another value according to the mutually exclusive characteristics. Finally, the classification result is obtained according to the fusion algorithm. Filtering normal data features from fault data features is proposed, which shields the interference and makes the fault features more prominent. We confirm that good performance improvements have been achieved in the machine fault detection data set, and the results are better than most mainstream models.

• International Journal of Internet, Broadcasting and Communication
• /
• 제15권2호
• /
• pp.124-132
• /
• 2023

In this paper, we present a simple classification model based on statistical features and demonstrate the successful implementation of a brain-computer interface (BCI) based light on/off control system. This research shows study and development of light on/off control system based on BCI technology, which allows the users to control switching a lamp using electroencephalogram (EEG) signals. The logistic regression algorithm is used for classification of the EEG signal to convert it into light on, light off control commands. Training data were collected using 14-channel BCI system which records the brain signals of participants watching a screen with flickering lights and saves the data into .csv file for future analysis. After extracting a number of features from the data and performing classification using logistic regression, we created commands to switch on a physical lamp and tested it in a real environment. Logistic regression allowed us to quite accurately classify the EEG signals based on the user's mental state and we were able to classify the EEG signals with 82.5% accuracy, producing reliable commands for turning on and off the light.

• Investigative Magnetic Resonance Imaging
• /
• 제6권2호
• /
• pp.137-146
• /
• 2002

Purpose : This paper introduces an improved classification approach which adopts a deterministic relaxation method and an agglomerative clustering technique for the classification of MRI using neural network. The proposed approach can solve the problems of convergency to local optima and computational burden caused by a large number of input patterns when a neural network is used for image classification. Materials and methods : Application of Hopfield neural network has been solving various optimization problems. However, major problem of mapping an image classification problem into a neural network is that network is opt to converge to local optima and its convergency toward the global solution with a standard stochastic relaxation spends much time. Therefore, to avoid local solutions and to achieve fast convergency toward a global optimization, we adopt MFA to a Hopfield network during the classification. MFA replaces the stochastic nature of simulated annealing method with a set of deterministic update rules that act on the average value of the variable. By minimizing averages, it is possible to converge to an equilibrium state considerably faster than standard simulated annealing method. Moreover, the proposed agglomerative clustering algorithm which determines the underlying clusters of the image provides initial input values of Hopfield neural network. Results : The proposed approach which uses agglomerative clustering and deterministic relaxation approach resolves the problem of local optimization and achieves fast convergency toward a global optimization when a neural network is used for MRI classification. Conclusion : In this paper, we introduce a new paradigm to classify MRI using clustering analysis and deterministic relaxation for neural network to improve the classification results.

• Journal of the Korea Society for Simulation
• /
• 제19권1호
• /
• pp.143-151
• /
• 2010

Fingerprint classification is a step to increase the efficiency of an 1:N fingerprint recognition system and plays a role to reduce the matching time of fingerprint and to increase accuracy of recognition. It is difficult to classify fingerprints, because the ridge pattern of each fingerprint class has an overlapping characteristic with more than one class, fingerprint images may include a lot of noise and an input condition is an exceptional case. In this paper, we propose a novel approach to design a stochastic model and to accomplish fingerprint classification using a directional characteristic of fingerprints for an effective classification of various qualities. We compute the directional value by searching a fingerprint ridge pixel by pixel and extract a directional characteristic by merging a computed directional value by fixed pixels unit. The modified Markov model of each fingerprint class is generated using Markov model which is a stochastic information extraction and a recognition method by extracted directional characteristic. The weight list of classification model of each class is decided by analyzing the state transition matrixes of the generated Markov model of each class and the optimized value which improves the performance of fingerprint classification using GA (Genetic Algorithm) is estimated. The performance of the optimized classification model by GA is superior to the model before the optimization by the experiment result of applying the fingerprint database of various qualities to the optimized model by GA. And the proposed method effectively achieved fingerprint classification to exceptional input conditions because this approach is independent of the existence and nonexistence of singular points by the result of analyzing the fingerprint database which is used to the experiments.

• Journal of Korea Water Resources Association
• /
• 제55권4호
• /
• pp.267-278
• /
• 2022

Recently Flood damage volume has increased as heavy rain has frequently occurred. Especially urban areas are a vulnerability to flooding damage because of densely concentrated population and property. A local government is preparing to mitigate flood damage through the heavy rain warning issued by Korea Meteorological Administration. This warning classification is identical for a national scale. However, Seoul has 25 administrative districts with different regional characteristics such as climate, topography, disaster prevention state, and flood damage severity. This study considered the regional characteristics of 25 administrative districts to analyze the flood vulnerability using entropy weight and Euclidean distance. The rainfall classification was derived based on probability rainfall and flood damage rainfall that occurred in the past. The result shows the step 2 and step 4 of rainfall classification was not significantly different from the heavy rain classification of the Korea Meteorological Administration. The flood vulnerability is high with high climate exposure and low adaptability to climate change, and the rainfall classification is low in the northern region of Seoul. It is possible to preemptively respond to floods in the northern region of Seoul based on relatively low rainfall classification. In the future, we plan to review the applicability of rainfall forecast data using the rainfall classification of results from this study. These results will contribute to research for preemptive flood response measures.

• Asian Pacific Journal of Cancer Prevention
• /
• 제13권12호
• /
• pp.6527-6531
• /
• 2012

Purpose: Complement receptor 1 (CR1) is induced by Epstein-Barr virus (EBV) and may be a potential biomarker of nasopharyngeal carcinoma (NPC). We conducted the present study to evaluate the association of CR1 expression with clinicopathological features and prognosis of NPC. Methods: We enrolled 145 NPC patients and 110 controls. Expression levels of CR1 in peripheral blood mononuclear cells (PBMCs) were detected using quantitative real-time PCR and associations with clinicopathological features and prognosis were examined. Results: CR1 levels in the NPC group [3.54 (3.34, 3.79)] were slightly higher than those in the controls [3.33 (3.20, 3.47)] (P<0.001). Increased CR1 expression was associated with histology classification (type III vs. type II, P=0.002), advanced clinical stage (P=0.003), high T stage (P=0.017), and poor overall survival (HR, 4.89; 95% CI, 1.23-19.42; P=0.024). However, there were no statistically significant differences in CR1 expression among N or M stages. Conclusion: These findings indicate that CR1 expression in PBMCs may be a new biomarker for prognosis of NPC and a potential therapeutic target.

• Journal of the Ergonomics Society of Korea
• /
• 제31권3호
• /
• pp.427-435
• /
• 2012

The aim of this study is to investigate facial temperature changes induced by facial expression and emotional state in order to recognize a persons emotion using facial thermal images. Background: Facial thermal images have two advantages compared to visual images. Firstly, facial temperature measured by thermal camera does not depend on skin color, darkness, and lighting condition. Secondly, facial thermal images are changed not only by facial expression but also emotional state. To our knowledge, there is no study to concurrently investigate these two sources of facial temperature changes. Method: 231 students participated in the experiment. Four kinds of stimuli inducing anger, fear, boredom, and neutral were presented to participants and the facial temperatures were measured by an infrared camera. Each stimulus consisted of baseline and emotion period. Baseline period lasted during 1min and emotion period 1~3min. In the data analysis, the temperature differences between the baseline and emotion state were analyzed. Eyes, mouth, and glabella were selected for facial expression features, and forehead, nose, cheeks were selected for emotional state features. Results: The temperatures of eyes, mouth, glanella, forehead, and nose area were significantly decreased during the emotional experience and the changes were significantly different by the kind of emotion. The result of linear discriminant analysis for emotion recognition showed that the correct classification percentage in four emotions was 62.7% when using both facial expression features and emotional state features. The accuracy was slightly but significantly decreased at 56.7% when using only facial expression features, and the accuracy was 40.2% when using only emotional state features. Conclusion: Facial expression features are essential in emotion recognition, but emotion state features are also important to classify the emotion. Application: The results of this study can be applied to human-computer interaction system in the work places or the automobiles.

• Annals of Hepato-Biliary-Pancreatic Surgery
• /
• 제26권4호
• /
• pp.325-332
• /
• 2022

Backgrounds/Aims: To analyze relationships of hepatic histopathological findings and bile microbiological profiles with perioperative outcomes and risk of late biliary stricture in individuals undergoing surgical bile duct injury (BDI) repair. Methods: A historical cohort study was carried out at a tertiary university hospital. Fifty-six individuals who underwent surgical BDI repair from 2014-2018 with a minimal follow-up of 24 months were enrolled. Liver biopsies were performed to analyze histopathology. Bile samples were collected during repair procedures. Hepatic histopathological findings and bile microbiological profiles were then correlated with perioperative and late outcomes through uni- and multi-variate analyses. Results: Forty-three individuals (76.8%) were females and average age was 47.2 ± 13.2 years; mean follow-up was 38.1 ± 18.6 months. The commonest histopathological finding was hepatic fibrosis (87.5%). Bile cultures were positive in 53.5%. The main surgical technique was Roux-en-Y hepaticojejunostomy (96.4%). Overall morbidity was 35.7%. In univariate analysis, liver fibrosis correlated with the duration of the operation (R = 0.3; p = 0.02). In multivariate analysis, fibrosis (R = 0.36; p = 0.02) and cholestasis (R = 0.34; p = 0.02) independently correlated with operative time. Strasberg classification independently correlated with estimated bleeding (R = 0.31; p = 0.049). The time elapsed between primary cholecystectomy and BDI repair correlated with hepatic fibrosis (R = 0.4; p = 0.01). Conclusions: Bacterial contamination of bile was observed in most cases. The degree of fibrosis and cholestasis correlated with operative time. The waiting time for definitive repair correlated with the severity of liver fibrosis.