• Journal of Intelligence and Information Systems
• /
• v.24 no.1
• /
• pp.205-225
• /
• 2018

Convolutional Neural Network (ConvNet) is one class of the powerful Deep Neural Network that can analyze and learn hierarchies of visual features. Originally, first neural network (Neocognitron) was introduced in the 80s. At that time, the neural network was not broadly used in both industry and academic field by cause of large-scale dataset shortage and low computational power. However, after a few decades later in 2012, Krizhevsky made a breakthrough on ILSVRC-12 visual recognition competition using Convolutional Neural Network. That breakthrough revived people interest in the neural network. The success of Convolutional Neural Network is achieved with two main factors. First of them is the emergence of advanced hardware (GPUs) for sufficient parallel computation. Second is the availability of large-scale datasets such as ImageNet (ILSVRC) dataset for training. Unfortunately, many new domains are bottlenecked by these factors. For most domains, it is difficult and requires lots of effort to gather large-scale dataset to train a ConvNet. Moreover, even if we have a large-scale dataset, training ConvNet from scratch is required expensive resource and time-consuming. These two obstacles can be solved by using transfer learning. Transfer learning is a method for transferring the knowledge from a source domain to new domain. There are two major Transfer learning cases. First one is ConvNet as fixed feature extractor, and the second one is Fine-tune the ConvNet on a new dataset. In the first case, using pre-trained ConvNet (such as on ImageNet) to compute feed-forward activations of the image into the ConvNet and extract activation features from specific layers. In the second case, replacing and retraining the ConvNet classifier on the new dataset, then fine-tune the weights of the pre-trained network with the backpropagation. In this paper, we focus on using multiple ConvNet layers as a fixed feature extractor only. However, applying features with high dimensional complexity that is directly extracted from multiple ConvNet layers is still a challenging problem. We observe that features extracted from multiple ConvNet layers address the different characteristics of the image which means better representation could be obtained by finding the optimal combination of multiple ConvNet layers. Based on that observation, we propose to employ multiple ConvNet layer representations for transfer learning instead of a single ConvNet layer representation. Overall, our primary pipeline has three steps. Firstly, images from target task are given as input to ConvNet, then that image will be feed-forwarded into pre-trained AlexNet, and the activation features from three fully connected convolutional layers are extracted. Secondly, activation features of three ConvNet layers are concatenated to obtain multiple ConvNet layers representation because it will gain more information about an image. When three fully connected layer features concatenated, the occurring image representation would have 9192 (4096+4096+1000) dimension features. However, features extracted from multiple ConvNet layers are redundant and noisy since they are extracted from the same ConvNet. Thus, a third step, we will use Principal Component Analysis (PCA) to select salient features before the training phase. When salient features are obtained, the classifier can classify image more accurately, and the performance of transfer learning can be improved. To evaluate proposed method, experiments are conducted in three standard datasets (Caltech-256, VOC07, and SUN397) to compare multiple ConvNet layer representations against single ConvNet layer representation by using PCA for feature selection and dimension reduction. Our experiments demonstrated the importance of feature selection for multiple ConvNet layer representation. Moreover, our proposed approach achieved 75.6% accuracy compared to 73.9% accuracy achieved by FC7 layer on the Caltech-256 dataset, 73.1% accuracy compared to 69.2% accuracy achieved by FC8 layer on the VOC07 dataset, 52.2% accuracy compared to 48.7% accuracy achieved by FC7 layer on the SUN397 dataset. We also showed that our proposed approach achieved superior performance, 2.8%, 2.1% and 3.1% accuracy improvement on Caltech-256, VOC07, and SUN397 dataset respectively compare to existing work.

• Journal of Digital Convergence
• /
• v.12 no.9
• /
• pp.219-223
• /
• 2014

In line with the advanced wireless communication technology, M2M (Machine-to-Machine) communication has drawn attention in industry. M2M communication features are installed and operated in the fields where human accessibility is highly limited such as disaster, safety, construction, health and welfare, climate, environment, logistics, culture, defense, medical care, agriculture and stockbreeding. In M2M communication, machine replaces people for automatic communication and countermeasures as part of unmanned information management and machine operation. Wireless M2M inter-device communication is likely to be exposed to intruders' attacks, causing security issues, which warrants proper security measures including cross-authentication of whether devices are legitimate. Therefore, research on multiple security protocols has been conducted. The present study applied SessionKey, HashFunction and Nonce to address security issues in M2M communication and proposed a safe protocol with reinforced security properties. Notably, unlike most previous studies arguing for the security of certain protocols based on mathematical theorem proving, the present study used the formal verification with Casper/FDR to prove the safety of the proposed protocol. In short, the proposed protocol was found to be safe and secure.

• Journal of Digital Convergence
• /
• v.12 no.9
• /
• pp.417-422
• /
• 2014

This paper was to explore the effect of aging anxiety of the middle-aged on depression in aging society. This survey was conducted from September, 2013 to December, 2013 and 302responses from the survey were used for the analysis. The data was analyzed by descriptive statistics, t-test, ANOVA, pearson correlation and Multiple Regression. The research findings could be summarized as the followings. First, the mean scores of aging anxiety and depression were lower than average. Second, aging anxiety was the significant predictor to explain depression. Third, income also was the significant predictors which have the effectiveness to the depression.

• Journal of the Korean Society of Clothing and Textiles
• /
• v.22 no.4
• /
• pp.503-514
• /
• 1998

The purpose of this study was to develop a large representative data base for jeans marketing strategy This study was to survey brand features(launching year, launching company, design concept, sales volume, and price zone) in the current market, and was to examine brand awareness and it's relationship to segmented distribution regions, demo- graphic variables(sex, age, monthly household income, and seasonal clothing expenditure). This study was also to analyze brand image and consumer's seeking image. The 660 questionnaires were distributed and 618 reliable ones were used for statistical analysis. A SAS statistical package including frequency table, Chi-square test, factor analysis, analysis of variance(ANOVA), Duncan's multiple range test and paired-t test was used. The results are as follows: 1. Brand awareness involves "brand recall" based on asking a person to name the brand recalled first, and "brand recognition" based on asking to identify brand name from 30 given brands. The result of recall brand test indicated that Levi's was dominant brand. People recognized about 70.8% of brands on the average. Brand recognition was influenced by segmented distribution region and demographic variables. 2. There was significantly positive relationship between brand recognition and purchasing behavior. 3. National brands were more recognized than Licensed brands. 4. The result showed that "Nix" was best represented for sophisticated brand image, "Strom" for characteristic, "Jambangee" for resonable price, and "Levi's" for classic '||'&'||' comfortable brand image. 5. As a result of factor analysis on consumer's seeking image, six factors(characteristic, young, intelligent/sexy, comfortable, exotic and popular) were found. Several factors had a relationship with preferred design, demographic variables, fashion interest, and brand recognition. variables, fashion interest, and brand recognition.

• Journal of Families and Better Life
• /
• v.27 no.3
• /
• pp.239-251
• /
• 2009

This study aims to examine how much Korean's family value orientations and gender role attitudes are different from those of U.S.A., Sweden, and Japan, and how demographic variables influence family value orientations and gender role attitudes across the countries. By using 2004 Korea General Social Survey data and 2002 International Social Survey Program family module, multiple regression analyses showed that Korean's family value orientations and gender role attitudes were much more traditional than those of U.S.A., Sweden, and Japan, even after controlling demographic variables. Furthermore, each country showed a distinct pattern in the impact of demographic variables on family value orientations and gender role attitudes. Among the demographic variables, age and marital status were statistically significant indicators of family value orientations for all the countries. However, gender, the year of education, and employment status effected on family value orientations only in some countries. The findings of this study showed that Korea was still traditional in terms of family value orientations and gender role attitudes, compared with U.S.A, Sweden, and Japan. Although family value orientations were more traditional in Korea than in the other countries, all the countries showed similar patterns of explaining mechanism in the effect of demographic variables on family value orientations. People who were men and married were likely to be more traditional than those who were women and unmarried. However, gender role attitudes showed interesting results. All the demographic variables were significant predictors of gender role attitudes for Korea, whereas only some of demographic variables were statistically significant indicators of gender role attitudes for other countries. That is, Korean society showed strong attitudinal differences on the basis of demographic variables. The implication of these differences was discussed.

• Journal of Korean Academy of Fundamentals of Nursing
• /
• v.10 no.2
• /
• pp.163-170
• /
• 2003

Purpose: This study was done to identify stress, fatigue, and self efficacy in patients on dialysis and identify their effect on quality of life. Method: The participants (n=149) who agreed to participate in the research were selected from among those who were on dialysis treatment at C University Medical Centers in Seoul and Inchon, and at C University Medical Center in Jeju. The data were collected from July, 2001 to March, 2002 using interviews. Data analysis was done with t-test, ANOVA, and multiple stepwise regression using the SAS program. Result: Quality of life was significantly different according to economic status, and type of dialysis. Quality of life for the group with higher economic status was better than for the group with lower economic status, and the group on peritoneal dialysis had higher quality of lift scores than the group on hemodialysis. As for self-efficacy, there were significant difference according to duration of disease, type of dialysis, and duration of dialysis. Stress was significantly different according to marital status. Quality of life was significantly predicted by stress (40.02%) and fatigue (3.85%). Conclusion: To improve quality of life for people on dialysis, their stress should be treated positively, also multi-dimensional nursing interventions to provide emotional support and nursing interventions to diminish fatigue are required.

• Industrial Engineering and Management Systems
• /
• v.11 no.2
• /
• pp.142-147
• /
• 2012

Individuals with autistic spectrum disorders (ASD) have difficulty inferring other people's feelings from their facial expressions and/or from situational cues, and therefore, they are less able to respond with prosocial behavior. We developed a computer-based training program to help teach the connection between facial-based or situation-based emotions and prosocial behavioral responses. An 8-year-old male school child with ASD participated in the study. In this program, he was trained to identify persons in need of help and appropriate prosocial responses using novel photo-based scenarios. When he misidentified emotions from photographs of another's face, the program highlighted those parts of the face which effectively communicate emotion. To increase the likelihood that he would learn a generalized repertoire of emotional understanding, multiple examples of emotional expressions and situations were provided. When he misidentified persons expressing a need for help, or failed to identify appropriate helping behaviors, role playing was used to help him appreciate the state of mind of a person in need of help. The results of the training indicated increases in prosocial behaviors during a laboratory task that required collaborative work. His homeroom teacher, using a behavioral rating scale, reported that he now understood another's emotion or situation better than before training. These findings indicate the effects of the training are not limited to the artificial experiment situation, but also carried over to his school life.

• Journal of Biomedical Engineering Research
• /
• v.33 no.2
• /
• pp.59-64
• /
• 2012

With rapid increasing usage of smart phones, social concerns have arisen about the possible effects of electromagnetic fields emitted from wideband code division multiple access(WCDMA) mobile phones on human health. The number of people with self-reported electromagnetic hypersensitivity(EHS) who complain of various subjective symptoms such as headache, insomnia etc. has also recently increased. However, it is unclear whether EHS results from physiological or other origins. In this double-blinded study, we investigated physiological changes such as heart rate, respiration rate, and heart rate variability with real and sham exposures for 15 EHS and 17 non-EHS persons using a module inside a dummy phone. Experiment was conducted using a WCDMA module with average power of 24 dBm at 1950 MHz with the specific absorption rate of 1.57 W/kg using a headphone for 32 min. As a conclusion, WCDMA RF exposure did not have any effects on the physiological variables in either group.

• Proceedings of the KSR Conference
• /
• 2011.10a
• /
• pp.165-171
• /
• 2011

A Consider the dead man's switch installed in each and every locomotive cab, which support operational safety on railways around the world. The concept is very simple - every 150 to 180 seconds an illuminated push-button demands to be acknowledged so as to know that the Train Driver is alive and active. In the absence of a response over a period of minutes, the vigilance control will automatically apply the train brakes and bring the train to a stand. If we multiply the resetting of the vigilance control 60 times per hour by a 10-hour shift it equals 600 presses of the button during the shift that a Train Driver must pay attention to and acknowledge. This adds a fair bit of pressure on the train driver's job, particularly when he/she is driving through stations, with passengers moving about on platforms in an environment of complex signaling arrangements - all the while looking out for restricting signals. From this perspective, the Vigilance System's demand to be acknowledged every 150/180 seconds is disturbing and can unnecessarily take a driver's attention away from what is happening outside the confines of the cab. A much more dramatic situation can happen when a train driver is driving hour after hour at night when, by Mother's Nature request - people need to sleep. Experience and research shows that the the dead man's switch can be pressed by train driver in a state of deep relaxation and 'micro-sleep'. The vigilance control system which is applied to reduce the drive load considering bio-response multiple unit train is proposed.

• Journal of Family Resource Management and Policy Review
• /
• v.20 no.1
• /
• pp.89-108
• /
• 2016

The purpose of this study is to examine the demographic, economic, and psychological factors that influence the quality of life (QoL) of the elderly. The data in the 'Survey on the Elderly in 2011', were collected from The Ministry of Health & Welfare and the Korea Institute for Health and Social Affairs. The samples included 3,312 out of the total of 15,146 elderly people aged from 65 to 70 years. Multiple regression analysis and path analysis were used to analyze the research model. The findings are as follows. First, the most critical factor that affected the QoL of the elderly was "subjective standard of living" with other factors including "subjective physical condition level", "depression", "total annual income", "the number of chronic disease", "self-esteem", "economic supports from their adult children", "education level", "living with their adult children or not", "sex", and "home ownership", in that order. Second, based on the path analysis results, the QoL of the elderly was influenced by "subjective physical condition level", "subjective standard of living", "depression", "self-esteem", "total annual income", "education level", and "the number of chronic disease", in that order. In particular, "depression" had a negative direct effect and an indirect effect on the QoL of the elderly, whereas "self-esteem" had a positive direct effect and indirect effect on their QoL. Third, "labor force participation status" did not affect the QoL of the elderly. However, it implied the mediating effect of "self-esteem" and "depression" on their QoL. These results show that demographic, economic, and psychological factors heavily influence the QoL of the elderly. Specifically, "depression" and "self-esteem" turn out to be mediating variables that strongly influence the QoL of the elderly. Also, "labor force participation status" does not have a positive direct effect on the QoL of the elderly but has a positive mediating effect of "depression" and "self-esteem" on their QoL.