• Journal of Internet Computing and Services
• /
• v.22 no.2
• /
• pp.59-68
• /
• 2021

As artificial intelligence(AI) technologies, which have made rapid growth recently, began to be applied to the marine environment such as ships, there have been active researches on the application of CNN-based models specialized for digital videos. In E-Navigation service, which is combined with various technologies to detect floating objects of clash risk to reduce human errors and prevent fires inside ships, real-time processing is of huge importance. More functions added, however, mean a need for high-performance processes, which raises prices and poses a cost burden on shipowners. This study thus set out to propose a method capable of processing information at a high rate while maintaining the accuracy by applying Quantization techniques of a deep learning model. First, videos were pre-processed fit for the detection of floating matters in the sea to ensure the efficient transmission of video data to the deep learning entry. Secondly, the quantization technique, one of lightweight techniques for a deep learning model, was applied to reduce the usage rate of memory and increase the processing speed. Finally, the proposed deep learning model to which video pre-processing and quantization were applied was applied to various embedded boards to measure its accuracy and processing speed and test its performance. The proposed method was able to reduce the usage of memory capacity four times and improve the processing speed about four to five times while maintaining the old accuracy of recognition.

• The KIPS Transactions:PartD
• /
• v.10D no.4
• /
• pp.567-576
• /
• 2003

Spatial data mining, i.e., discovery of interesting characteristics and patterns that may implicitly exists in spatial databases, is a challenging task due to the huge amounts of spatial data. Clustering algorithms are attractive for the task of class identification in spatial databases. Several methods for spatial clustering have been presented in recent years, but have the following several drawbacks increase costs due to computing distance among objects and process only memory-resident data. In this paper, we propose an efficient grid cell based spatial clustering method for spatial data mining. It focuses on resolving disadvantages of existing clustering algorithms. In details, it aims to reduce cost further for good efficiency on large databases. To do this, we devise a spatial clustering algorithm based on grid ceil structures including cell relationships.

• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.444-448
• /
• 2005

This paper presents a developing procedure of the CAN-based real-time simulator for car body control, aiming at replacing the actual W/H (Wiring Harness) and J/B(Junction Box) couple eventually. The CAN protocol, as one kind of field-bus communication, defines the lowest 2 layers of the ISO/OSI standard, namely, the physical layer(PL) and the data link layer(DLL), for which the CSMA/NBA protocol is generally adopted. For CPU, two PIC18Fxx8x's are used because of their built-in integration of CAN controller, large internal FLASH memory (48K or 64K), and their costs. To control J/B's and actuators, 2 controller boards are separately implemented, between which CAN lines communicate through CAN transceivers MCP255. A power motor for washing windshield, 1 door lock motor, and 6 blink lamps are chosen for actuators of the simulator for the first stage. For the software architecture, a polling method is used for the fast global response time despite its slow individual response time. To improve the individual response time and to escape from some eventual trapped-function loops, High/Low ports of the CPU are simply used, which increases the stability of the actuator modules. The experimental test shows generally satisfactory results in normal transmitting / receiving function and message trace function. This simulator based on CAN shows a promising usefulness of lighter, more reliable and intelligent distributed body control approach than the conventional W/H and J/B couple. Another advantage of this approach lies in the distributed control itself, which gives better performance in hard real-time computing than centralized one, and in the ability of integrating different modules through CAN.

• Journal of Internet Computing and Services
• /
• v.18 no.1
• /
• pp.27-35
• /
• 2017

In near future, IoT data is expected to be a major portion of Big Data. Moreover, sensor data is expected to be major portion of IoT data, and its' research is actively carried out currently. However, processed results may not be trusted and used if outlier data is included in the processing of sensor data. Therefore, method for detection and deletion of those outlier data before processing is studied in this paper. Moreover, we used Spark which is memory based distributed processing environment for fast processing of big sensor data. The detection and deletion of outlier data consist of four stages, and each stage is implemented with Mapper and Reducer operation. The proposed method is compared in three different processing environments, and it is expected that the outlier detection and deletion performance is best in the distributed Spark environment as data volume is increasing.

• Ceramist
• /
• v.21 no.2
• /
• pp.4-18
• /
• 2018

Lastly, neuromorphic computing chip has been extensively studied as the technology that directly mimics efficient calculation algorithm of human brain, enabling a next-generation intelligent hardware system with high speed and low power consumption. Three-terminal based synaptic transistor has relatively low integration density compared to the two-terminal type memristor, while its power consumption can be realized as being so low and its spike plasticity from synapse can be reliably implemented. Also, the strong electrical interaction between two or more synaptic spikes offers the advantage of more precise control of synaptic weights. In this review paper, the results of synaptic transistor mimicking synaptic behavior of the brain are classified according to the channel material, in order of silicon, organic semiconductor, oxide semiconductor, 1D CNT(carbon nanotube) and 2D van der Waals atomic layer present. At the same time, key technologies related to dielectrics and electrolytes introduced to express hysteresis and plasticity are discussed. In addition, we compared the essential electrical characteristics (EPSC, IPSC, PPF, STM, LTM, and STDP) required to implement synaptic transistors in common and the power consumption required for unit synapse operation. Generally, synaptic devices should be integrated with other peripheral circuits such as neurons. Demonstration of this neuromorphic system level needs the linearity of synapse resistance change, the symmetry between potentiation and depression, and multi-level resistance states. Finally, in order to be used as a practical neuromorphic applications, the long-term stability and reliability of the synapse device have to be essentially secured through the retention and the endurance cycling test related to the long-term memory characteristics.

• The Journal of the Korea Contents Association
• /
• v.10 no.1
• /
• pp.94-102
• /
• 2010

In recent years, embedded systems such as cellular phones, Portable Multimedia Player, intelligent appliance, automobile engine control are reshaping the way people live, work, and play. Thereby, services application to guarantee various requirements of users become increasingly sophisticated and complicated, such embedded computing platforms use real-time operating systems (RTOSs) with time determinism. These RTOSs must not only provide predictable services but must also be efficient and small in size. Kernel services should also be deterministic by specifying how long each service call will take to execute. Having this information allows the application designers to better plan their real-time application software so as not to miss the deadline of each task. In this paper, we present the complete generalized real-time scheduling algorithm using multi-dimensional methodology to determine the highest priority in the ready list with 2r levels of priorities in a constant time without additional memory overhead.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.33 no.12C
• /
• pp.1055-1061
• /
• 2008

H.264 is the prominent video coding standard in various applications such as real-time streaming and digital multimedia broadcasting, since it provides enhanced compression performance, error resilience tools, and network adaptation. Compression efficiency of H.264 has been improved, however, it requires more computing and memory access than traditional methods. In this paper we proposed adaptive intra fast algorithm for real-time video surveillance system reducing the encoding complexity of H264/A VC. For this aim, temporal interrelationship between macroblock in the previous and the current frame is used to decide the encoding mode of macroblock fast. As a result, though video quality was deteriorated a little, less than 0.04dB, and bit rate was somewhat increased in suggested method, however, proposed method improved encoding time significantly and, in particular, encoding time of an image with little changes of neighboring background such as surveillance video was more shortened than traditional methods.

• Journal of KIISE:Computing Practices and Letters
• /
• v.16 no.11
• /
• pp.1071-1075
• /
• 2010

Because a sensor node in wireless sensor networks(WSNs) has limited resources, such as battery capacity and memory, data aggregation techniques have been studied to manage the limited resources efficiently. Because sensor network uses wireless communication, a data can be disclosed by attacker. Thus, the study on data protection schemes for data aggregation is essential in WSNs. But the existing data aggregation methods require both a large number of computation and communication, in case of network construction and data aggregation processing. To solve the problem, we propose a data protection scheme based on Hilbert-curve for data aggregation. Our scheme can minimizes communications among neighboring sensor nodes by using tree-based routing. Moreover, it can protect the data from attacker by doing encryption through a Hilbert-curve technique based on a private seed, Finally, we show that our scheme outperforms the existing methods in terms of message transmission and average sensor node lifetime.

• KIPS Transactions on Software and Data Engineering
• /
• v.2 no.2
• /
• pp.137-144
• /
• 2013

A sequential pattern mining is finding out frequent patterns from the data set in time order. In this field, a dynamic weighted sequential pattern mining is applied to a computing environment that changes depending on the time and it can be utilized in a variety of environments applying changes of dynamic weight. In this paper, we propose a new sequence data mining method to explore the stream data by applying the dynamic weight. This method reduces the candidate patterns that must be navigated by using the dynamic weight according to the relative time sequence, and it can find out frequent sequence patterns quickly as the data input and output using a hash structure. Using this method reduces the memory usage and processing time more than applying the existing methods. We show the importance of dynamic weighted mining through the comparison of different weighting sequential pattern mining techniques.

• Journal of Internet Computing and Services
• /
• v.5 no.2
• /
• pp.69-73
• /
• 2004

A speaker model In speaker recognition system is to be trained from a large data set gathered in multiple sessions. Large data set requires large amount of memory and computation, and moreover it's practically hard to make users utter the data inseveral sessions. Recently the incremental adaptation methods are proposed to cover the problems, However, the data set gathered from multiple sessions is vulnerable to the outliers from the irregular utterance variations and the presence of noise, which result in inaccurate speaker model. In this paper, we propose an incremental robust adaptation method to minimize the influence of outliers on Gaussian Mixture Madel based speaker model. The robust adaptation is obtained from an incremental version of M-estimation. Speaker model is initially trained from small amount of data and it is adapted recursively with the data available in each session, Experimental results from the data set gathered over seven months show that the proposed method is robust against outliers.