• Journal of Computing Science and Engineering
• /
• v.12 no.1
• /
• pp.24-35
• /
• 2018

Deep convolutional neural network (DCNN) is an advanced technology in image recognition. Because of extreme computing resource requirements, DCNN implementation with software alone cannot achieve real-time requirement. Therefore, the need to implement DCNN accelerator hardware is increasing. In this paper, we present a field programmable gate array (FPGA)-based hardware accelerator design of DCNN targeting handwritten Hangul character recognition application. Also, we present design optimization techniques in SDAccel environments for searching the optimal FPGA design space. The techniques we used include memory access optimization and computing unit parallelism, and data conversion. We achieved about 11.19 ms recognition time per character with Xilinx FPGA accelerator. Our design optimization was performed with Xilinx HLS and SDAccel environment targeting Kintex XCKU115 FPGA from Xilinx. Our design outperforms CPU in terms of energy efficiency (the number of samples per unit energy) by 5.88 times, and GPGPU in terms of energy efficiency by 5 times. We expect the research results will be an alternative to GPGPU solution for real-time applications, especially in data centers or server farms where energy consumption is a critical problem.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.6
• /
• pp.648-653
• /
• 2010

The non-metric multidimensional scaling (nMDS) is a method for analyzing the relation among objects by mapping them onto the Euclidean space. The nMDS is useful when it is difficult to use the concept of distance between pairs of objects due to non-metric dissimilarities between objects. The nMDS can be regarded as an optimization problem in which there are many local optima. Since the conventional nMDS algorithm utilizes the steepest descent method, it has a drawback in that the method can hardly find a better solution once it falls into a local optimum. To remedy this problem, in this paper, we applied the simulated annealing to the nMDS and proposed a new optimization algorithm which could search for a global optimum more effectively. We examined the algorithm using benchmarking problems and found that improvement rate of the proposed algorithm against the conventional algorithm ranged from 0.7% to 3.2%. In addition, the statistical hypothesis test also showed that the proposed algorithm outperformed the conventional one.

• Atmosphere
• /
• v.18 no.4
• /
• pp.267-277
• /
• 2008

The occurrence and strength of the Asian Dust over the Korea Peninsular have been increased by the expansion of the desert area. For the continuous monitoring of the Asian Dust event, the geostationary satellites provide useful information by detecting the outbreak of the event as well as the long-range transportation of dust. The Infrared Optical Depth Index (IODI) derived from the MTSAT-1R data, indicating a quantitative index of the dust intensity, has been produced in real-time at Korea Meteorological Administration (KMA) since spring of 2007 for the forecast of Asian dust. The data processing algorithm for IODI consists of mainly two steps. The first step is to detect dust area by using brightness temperature difference between two thermal window channels which are influenced with different extinction coefficients by dust. Here we use dynamic threshold values based on the change of surface temperature. In the second step, the IODI is calculated using the ratio between current IR1 brightness temperature and the maximum brightness temperature of the last 10 days which we assume the clear sky. Validation with AOD retrieved from MODIS shows a good agreement over the ocean. Comparison of IODI with the ground based PM10 observation network in Korea shows distinct characteristics depending on the altitude of dust layer estimated from the Lidar data. In the case that the altitude of dust layer is relatively high, the intensity of IODI is larger than that of PM10. On the other hand, when the altitude of dust layer is lower, IODI seems to be relatively small comparing with PM10 measurement.

• Tunnel and Underground Space
• /
• v.29 no.6
• /
• pp.508-522
• /
• 2019

Because of the slope failure, not only property damage but also human damage can occur, slope stability analysis should be conducted to predict and reinforce of the slope. This paper, defines the ground areas that can be characterized in terms of slope failure such as Rockmass jointset, Rockmass fault, Soil, Leakage water and Crush zone in sloped images. As a result, it was shown that the deep learning instance segmentation network can be used to recognize and automatically segment the precise shape of the ground region with different characteristics shown in the image. It showed the possibility of supporting the slope mapping work and automatically calculating the ground characteristics information of slopes necessary for decision making such as slope reinforcement.

• Journal of the Korean Society for Precision Engineering
• /
• v.28 no.4
• /
• pp.482-489
• /
• 2011

In practical design process, designer needs to find an optimal solution by using full factorial discrete combination, rather than by using optimization algorithm considering continuous design variables. So, ANOVA(Analysis of Variance) based on an orthogonal array, i.e. Taguchi method, has been widely used in most parts of industry area. However, the Taguchi method is limited for the shape optimization by using CAE, because the multi-level and multi-objective optimization can't be carried out simultaneously. In this study, a combined method was proposed taking into account of multi-level computational orthogonal array and TOPSIS(Technique for Order preference by Similarity to Ideal Solution), which is known as a classical method of multiple attribute decision making and enables to solve various decision making or selection problems in an aspect of multi-objective optimization. The proposed method was applied to a case study of the multi-level shape optimization of lower arm used to automobile parts, and the design space was explored via an efficient application of the related CAE tools. The multi-level shape optimization was performed sequentially by applying both of the neural network model generated from seven-level four-factor computational orthogonal array and the TOPSIS. The weight and maximum stress of the lower arm, as the objective functions for the multi-level shape optimization, showed an improvement of 0.07% and 17.89%, respectively. In addition, the number of CAE carried out for the shape optimization was only 55 times in comparison to full factorial method necessary to 2,401 times.

• Journal of Internet Computing and Services
• /
• v.12 no.2
• /
• pp.37-46
• /
• 2011

The importance of data service with mobile is becoming more and more significant with the development of wireless network. However, due to the several type of mobile devices and contents, there is a problem that mobile contents should be edited according to the features of the mobile devices and the service providers. Fortunately, there is some solutions that convert mobile contents automatically according to mobile operator. Nevertheless, it has some issues that the response time of content service is delayed when converting mobile contents in real-time. In this paper, we propose an image cache replacement algorithm for minizing the volume of the real-time image transformation. Also, using both of the pre-conversion method and the real-time conversion method to improve storage space and response time.

• Convergence Security Journal
• /
• v.17 no.1
• /
• pp.101-109
• /
• 2017

The Internet of things is a system that connects and communicates all sorts of things such as people, objects, and data. It's to create and share information by its own each other. It can be used to enhance the function of private security and has brought about innovative development of private security. The Internet of things is a system that allows devices connected to the Internet to communicate independently of people-objects, objects-objects connected to the Internet. That and can be used in many industries, especially in the private security sector, its value is high. The use of the Internet of things to private security sector can reinforce security zones with always-on surveillance systems, also be enhanced by its own preparedness and response to the situation. However, this study will discusse the application and development of private security in the Internet of things. The practical application of the virtual space is an immediate task and it is also an essential factor in securing security.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.42 no.2 s.302
• /
• pp.39-48
• /
• 2005

This paper proposes a novel approach to recognize nine kinds of motion for a multifunction myoelectric hand, acquiring four channel EMG signals from electrodes placed on the forearm. To analyze EMG with properties of nonstationary signal, time-frequency features are extracted by wavelet packet transform. For dimensionality reduction and nonlinear mapping of the features, we also propose a feature projection composed of PCA and SOFM. The dimensionality reduction by PCA simplifies the structure of the classifier, and reduces processing time for the pattern recognition. The nonlinear mapping by SOFM transforms the PCA-reduced features to a new feature space with high class separability. Finally a multilayer neural network is employed as the pattern classifier. From experimental results, we show that the proposed method enhances the recognition accuracy, and makes it possible to implement a real-time pattern recognition.

• The KIPS Transactions:PartA
• /
• v.10A no.2
• /
• pp.123-130
• /
• 2003

This paper presents a new striped checkpointing scheme for serverless cluster computers, where the local disks are attached to the cluster nodes collectively form a distributed RAID with a single I/O space. Striping enables parallel I/O on the distributed disks and staggering avoids network bottleneck in the distributed RAID. We demonstrate how to reduce the checkpointing overhead and increase the availability by striping and staggering dynamically for communication intensive applications. Linpack HPC Benchamark and MPI programs are applied to these checkpointing schemes for performance evaluation on the 16-nodes cluster system. Benchmark results prove the benefits of the striped checkpointing scheme compare to the existing schemes, and these results are useful to design the efficient checkpointing scheme for fast rollback recovery from any single node failure in a cluster system.

• Journal of Digital Convergence
• /
• v.11 no.7
• /
• pp.185-192
• /
• 2013

The purpose of this study is to set the relationship between each parts that forms the engine of an automobile based on the 3D visualization technology which is able to be learned according to the skill of the operator in the industry field and to recommend the auto parts using a task ontology. A visualization method was proposed by structuring the complex knowledge by signifying the link and the node in forms of a network and using SOM which can be shown in the form of 3 dimension. In addition, by using is-a Relationship-based hierarchical Taxonomy setting the relationship between each of the parts that forms the engine of an automobile, to allow a recommendation using a weighted value possible. By providing and placing the complex knowledge in the 3D space to the user for an opportunity of more realistic and intuitive navigation, when randomly selecting the automobile parts, it allows the recommendation of the parts having a close relationship with the corresponding parts for easy assembly and to know the importance of usage for the automobile parts without any special expertise.