• Journal of Intelligence and Information Systems
• /
• v.25 no.1
• /
• pp.163-177
• /
• 2019

As smartphones are getting widely used, human activity recognition (HAR) tasks for recognizing personal activities of smartphone users with multimodal data have been actively studied recently. The research area is expanding from the recognition of the simple body movement of an individual user to the recognition of low-level behavior and high-level behavior. However, HAR tasks for recognizing interaction behavior with other people, such as whether the user is accompanying or communicating with someone else, have gotten less attention so far. And previous research for recognizing interaction behavior has usually depended on audio, Bluetooth, and Wi-Fi sensors, which are vulnerable to privacy issues and require much time to collect enough data. Whereas physical sensors including accelerometer, magnetic field and gyroscope sensors are less vulnerable to privacy issues and can collect a large amount of data within a short time. In this paper, a method for detecting accompanying status based on deep learning model by only using multimodal physical sensor data, such as an accelerometer, magnetic field and gyroscope, was proposed. The accompanying status was defined as a redefinition of a part of the user interaction behavior, including whether the user is accompanying with an acquaintance at a close distance and the user is actively communicating with the acquaintance. A framework based on convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent networks for classifying accompanying and conversation was proposed. First, a data preprocessing method which consists of time synchronization of multimodal data from different physical sensors, data normalization and sequence data generation was introduced. We applied the nearest interpolation to synchronize the time of collected data from different sensors. Normalization was performed for each x, y, z axis value of the sensor data, and the sequence data was generated according to the sliding window method. Then, the sequence data became the input for CNN, where feature maps representing local dependencies of the original sequence are extracted. The CNN consisted of 3 convolutional layers and did not have a pooling layer to maintain the temporal information of the sequence data. Next, LSTM recurrent networks received the feature maps, learned long-term dependencies from them and extracted features. The LSTM recurrent networks consisted of two layers, each with 128 cells. Finally, the extracted features were used for classification by softmax classifier. The loss function of the model was cross entropy function and the weights of the model were randomly initialized on a normal distribution with an average of 0 and a standard deviation of 0.1. The model was trained using adaptive moment estimation (ADAM) optimization algorithm and the mini batch size was set to 128. We applied dropout to input values of the LSTM recurrent networks to prevent overfitting. The initial learning rate was set to 0.001, and it decreased exponentially by 0.99 at the end of each epoch training. An Android smartphone application was developed and released to collect data. We collected smartphone data for a total of 18 subjects. Using the data, the model classified accompanying and conversation by 98.74% and 98.83% accuracy each. Both the F1 score and accuracy of the model were higher than the F1 score and accuracy of the majority vote classifier, support vector machine, and deep recurrent neural network. In the future research, we will focus on more rigorous multimodal sensor data synchronization methods that minimize the time stamp differences. In addition, we will further study transfer learning method that enables transfer of trained models tailored to the training data to the evaluation data that follows a different distribution. It is expected that a model capable of exhibiting robust recognition performance against changes in data that is not considered in the model learning stage will be obtained.

• Journal of Korea Entertainment Industry Association
• /
• v.13 no.5
• /
• pp.263-271
• /
• 2019

The recent phenomenon of job "Miss Match", which is inconsistent with knowledge in the demand of educational training institutes and industries, has spread to an increase in private education costs for reeducation and employment of new hires, resulting in weak individual job competency and poor employment capability, as well as economic and material waste at the national level. To compensate for these problems, the National Competency Standards(NCS), which are available immediately in practice and look for a standard point of national job competency with the aim of fostering human resources sought by industries, were developed, and even the NCS-based qualification system was launched in line with the stream of times. This study is intended to look into the importance and priority of competency units and competency unit elements at the NCS-based qualification L3 level in the skin care field for an overall check of the NCS-based qualification level at a time when educational institutes are organizing and operating the school curriculums according to the NCS and NCS-based qualification level. And it is attempted to provide basic data for the development of curriculum in fostering professional human resources required by industries. To analyze the needs for competency units and competency unit elements at the L3 level, a survey using AHP method was carried out to a group of field experts and a group of education experts. In addition, the SPSS(Statistical Package for Social Science) ver. 21.0 and Expert Choice 2000, an AHP-only solution was used to do statistical processing through the processes of data coding and data cleaning. The findings showed that there was a partial difference of opinion between a group of field experts and a group of education experts. This indicates that the inconsistencies between educational training institutes and industrial sites should be resolved at this time of change with the aim of fostering field customized human resources with professional skills. Consequently, the solution is to combine jobs at industrial sites and standardized educations of educational institutes with human resources required at industrial sites.

• Journal of the Korean Society of Safety
• /
• v.22 no.4
• /
• pp.110-116
• /
• 2007

There have been 3E problems of energy, economy and environment since the earth has its history. Especially, as the industrial society is highly developing, human need in daily life has also changed drastically. With the introduction of 40 hour working week system, more households enjoy picnics on weekends. More gas accidents take place on Saturdays and on Sundays than any other days of week. Consequently, this study tries to find out the influence of flame caused by the explosion of butane canister on the adjacent combustibles and people by simulating relevant quantity of TNT. In addition, the damage estimation was conducted by using API regulations. If the scale of the radiation heat is known by calculating the distance of flame influence from the explosion site, the damage from the site can be easily estimated. And the accident damage was estimated by applying the influence on the adjacent structures and people into the PROBIT model. According to the pro bit analyze, the spot which is 50cm away from the flame has 97% of the damage probability by the first-degree burn, 8% of the damage probability by the second-degree burn and 4% of the death probability by the fire.

• Journal of the Society of Disaster Information
• /
• v.16 no.3
• /
• pp.505-512
• /
• 2020

Purpose: The purpose of this study is to identify the contents and problems of present education for safety education specialist and provide suggestions for improvement in the future by examining the current status of capacity building education for safe education specialist. Method: The analysis of current status was conducted based on the results of two training sessions on safety education specialists and the research on the actual condition was carried on targeting safety education specialists. Result: As a result of the research, most of the contents of education for safety education specialist are biased toward theory and lack of a systematic education system, resulting in that safety education tends to be field education and focuses on theory and audio-visual education. Safety education is not solved by theory, but can be maximized the effect through experience and practice to experience the real situation. Conclusion: Therefore, based on the results of this study, a method of ICT is proposed to utilize for safety education, so that safety education specialists can be provided with practical and effective safety education, which can be used at real safety field.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.42 no.1
• /
• pp.19-29
• /
• 2006

To investigate the fishery and fishing ground environment of red horsehead (Branchiostegus japonicus), the author analyzed the fishery data and examined the amount of catches and oceanic environment on the adjacent seas of Jeju island and East China Sea. It was turned out that the favourable season of the red horsehead fishery is the month from March to June, the main fishing ground is located in 60 mile radius from the position $32.5^{\circ}N,\;125.7^{\circ}E$. The bottom seawater temperature in fishing ground was shown between $l3^{\circ}C\;and\;16^{\circ}C$, the salinity was appeared between 33.5 and 34.0psu without the seasonal variation of the year. Concentrations of materials(e.g, $NO_3^-\;and\;NO_2^-$) in spring and summer time in main fishing ground were higher than any other seasons, but that of phospheric materials were lower than any other seasons. Concentrations of $chlorophyll\;-\;{\alpha}$ in the main fishing ground was the highest in spring and summer at the surface layer, but the vertical profile of the $chlorophyll\;-\;{\alpha}$ concentrations in all seasons were not variable at bottom layer. Mean density of zooplankton abundance according to the vertical structure was higher and much stable in summer and autumn than spring and winter.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.25 no.12
• /
• pp.807-814
• /
• 2015

A control rod drive mechanism(CRDM) is a reactor regulating system, which inserts, withdraws or maintains a control rod containing a neutron absorbing material within a reactor core to control the reactivity of the core. The top-mounted CRDM for Jordan Research and Training Reactor(JRTR) with 5 MW power has been designed and fabricated based on the HANARO's experience through KAERI and DAEWOO consortium project. This paper describes the performance qualification test results to demonstrate the operability of a prototype and four production CRDMs during the reactor lifetime. The driving performance, the drop performance and the endurance tests for CRDM are carried out at a test rig simulating the actual reactor conditions. A vibration of internal components due to the coolant flow is also measured using a laser vibrometer. As a result, the CRDMs are driven having a good driving performance without a malfunction between command and output signals for the stepping motor. Also, the pure drop time and the impact acceleration are within 0.72 s and 4.2 g to meet the design requirements, and the vibrational displacement of control rod is measured as maximum $5.2{\mu}m$.

• Korean Journal of Remote Sensing
• /
• v.27 no.6
• /
• pp.637-651
• /
• 2011

This study presents a rule-based urban image classification method for time series analysis of changes in the vicinity of Asan-si and Cheonan-si in Chungcheongnam-do, using Landsat satellite images (1991-2006). The area has been highly developed through the relocation of industrial facilities, land development, construction of a high-speed railroad, and an extension of the subway. To determine the yearly changing pattern of the urban area, eleven classes were made depending on the trend of development. An algorithm was generalized for the rules to be applied as an unsupervised classification, without the need of training area. The analysis results show that the urban zone of the research area has increased by about 1.53 times, and each correlation graph confirmed the distribution of the Built Up Index (BUI) values for each class. To evaluate the rule-based classification, coverage and accuracy were assessed. When Optimal allowable factor=0.36, the coverage of the rule was 98.4%, and for the test using ground data from 1991 to 2006, overall accuracy was 99.49%. It was confirmed that the method suggested to determine the maximum allowable factor correlates to the accuracy test results using ground data. Among the multiple images, available data was used as best as possible and classification accuracy could be improved since optimal classification to suit objectives was possible. The rule-based urban image classification method is expected to be applied to time series image analyses such as thematic mapping for urban development, urban development, and monitoring of environmental changes.

• The Journal of Industrial Distribution & Business
• /
• v.5 no.2
• /
• pp.29-35
• /
• 2014

Purpose - The study aimed to evaluate the market opportunities and constraints confronting resource-poor pig farmers in South Africa. Research design, data, and methodology - Information was collected from 292 households in three municipalities through interviews with key informants. The data collected included socio-economic characteristics, major market channels, prices for different pig classes, average weight of the pigs on sale, number of pigs sold annually, and preferred meat quality attributes. Results - In Ngqushwa, 96% of respondents sold pigs as compared to Elundini (81%) and Ntabankulu (65%). Less resource-poor households and those with market-oriented production had large herdsizes (P < 0.05) when compared to more resource-poor farmers. The probability of selling pigs was high for the backyard production system and educated farmers. For all farmers, opportunities included high pork demand, good prices, employment creation, and a niche market for organically produced indigenous pork. Constraints include disease, feed shortages for large herds, distances to formal markets, lack of training, and drugs. Conclusions - Constraints outnumbered opportunities for the resource-poor pig farmers.

• Journal of Distribution Research
• /
• v.17 no.5
• /
• pp.25-38
• /
• 2012

In general, commercial franchisors prefer franchisees with high entrepreneurship and business capability. However, these entrepreneurial franchisees with high capability tend to depend less on franchisors. Although franchisees with less entrepreneurship and low capability need intensive care from franchisors, some of them result in business failure due to improper supports from incapable franchisors. This paper suggests several propositions regarding social franchising's role in supporting the low-income urban people on the premise that social franchising or micro-franchising provides implications for franchisor strategic orientations for franchisees with little capital. Through literature review about social franchising and micro franchising, some implications are drawn. Many social enterprises use franchising to get growth and sustainability, because franchising allow social enterprise to expand its scale and to achieve economies of scale despite of its non-commercial and social purpose. And continuous support and training undertake the most important role to achieve its social purpose. In commercial franchising, especially small business format franchising sector, franchisor have to consider not only commercial purpose but also social responsibility because of low capability and less entrepreneurship of franchisee. If franchisor pursue only own profit, this can increase the conflict and franchise system failure. So far many franchisors are concerned with own profit and external growth. But it is necessary to consider symbiotic relationship, social responsibility and sustainability more for the sake of desirable industrial growth in the future.

• Journal of Advanced Navigation Technology
• /
• v.21 no.6
• /
• pp.674-679
• /
• 2017

As the importance of the 4th industrial revolution and ICT technology increased, it became a software centered society. Existing software training was limited to the composition of the learning environment, and a lot of costs were incurred early. In order to solve these problems, a learning method using a web compiler was developed. The web compiler supports various software languages and shows compilation results to the user via the web. However, Web compilers that support artificial intelligence technology are missing. In this paper, we designed and implemented a tensor flow based web compiler, Google's artificial intelligence library. We implemented a system for learning artificial intelligence by building a meteorJS based web server, implementing tensor flow and tensor flow serving, Python Jupyter on a nodeJS based server. It is expected that it can be utilized as a tool for learning artificial intelligence in software centered society.