• IEMEK Journal of Embedded Systems and Applications
• /
• v.16 no.5
• /
• pp.179-186
• /
• 2021

For stereo matching based on deep learning, the design of network structure is crucial to the calculation of matching cost, and the time-consuming problem of convolutional neural network in image processing also needs to be solved urgently. In this paper, a method of stereo matching using sparse loss volume in parallax dimension is proposed. A sparse 3D loss volume is constructed by using a wide step length translation of the right view feature map, which reduces the video memory and computing resources required by the 3D convolution module by several times. In order to improve the accuracy of the algorithm, the nonlinear up-sampling of the matching loss in the parallax dimension is carried out by using the method of multi-category output, and the training model is combined with two kinds of loss functions. Compared with the benchmark algorithm, the proposed algorithm not only improves the accuracy but also shortens the running time by about 30%.

• Journal of Information Processing Systems
• /
• v.17 no.2
• /
• pp.227-241
• /
• 2021

Object detection techniques based on deep learning such as YOLO have high detection performance and precision in a single channel video stream. In order to expand to multiple channel object detection in real-time, however, high-performance hardware is required. In this paper, we propose a novel back-end server framework, a real-time AI vision platform (RAVIP), which can extend the object detection function from single channel to simultaneous multi-channels, which can work well even in low-end server hardware. RAVIP assembles appropriate component modules from the RODEM (real-time object detection module) Base to create per-channel instances for each channel, enabling efficient parallelization of object detection instances on limited hardware resources through continuous monitoring with respect to resource utilization. Through practical experiments, RAVIP shows that it is possible to optimize CPU, GPU, and memory utilization while performing object detection service in a multi-channel situation. In addition, it has been proven that RAVIP can provide object detection services with 25 FPS for all 16 channels at the same time.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.2
• /
• pp.63-69
• /
• 2019

Recently, ultra-low latency flash storage devices such as Z-SSD and Optane SSD were introduced with the significant technological improvement in the storage devices which provide much faster response time than today's other NVMe SSDs. With such ultra-low latency, $10{\mu}s$, storage devices the cost of context switch could be an overhead during interrupt-driven I/O completion process. As an interrupt-driven I/O completion process could bring an interrupt handling overhead, polling or hybrid-polling for the I/O completion is known to perform better. In this paper, we analyze tail latency problem in a polling process caused by process scheduling in data center environment where multiple applications run simultaneously under one system and we introduce our adaptive polling selection technique which dynamically selects efficient processing method between two techniques according to the system's conditions.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.1
• /
• pp.11-18
• /
• 2019

In this paper, we consider interferences from other automobile pulse radars using same frequency spectrum. In order to eliminate the interference, we propose the PN code modulation method. This method uses the cross-correlation between PN codes with different seed. The ROC performance is used for comparing the proposed detector to conventional method. And the proposed detector can decide the present or absent of targets and measure the range of the targets by using memory buffer of range gate. Especially, we use false alarm probability for all range gates. That is the false alarm if in any one range gate the false alarm occurs. From the simulation result, we can see that the proposed detector with using PN code is not influenced by interferences.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.15 no.6
• /
• pp.289-297
• /
• 2020

Although flash disks are being used widely instead of hard disks, it is difficult to optimize for effective utilization of flash disks because overwrite in place is impossible and the power consumption and time required for read, write, and erase operations are all different. One of these optimization issues is a cache management strategy to minimize write operations. The cache operates at two levels: an operating system equipped with flash disks and a translation layer within the flash disk. Most studies deal with the operating system-level cache strategy. In this study, we implement and analyse the DPW-LRU algorithm which is one of the recently proposed operating system cache replacement algorithms to apply to FTL, and grope with some improvements. As a result of the experiment, the DPW-LRU algorithm maintained superiority even in the FTL environment, and showed better performance with a slight improvement.

• Journal of the Korea Society of Computer and Information
• /
• v.12 no.5
• /
• pp.103-112
• /
• 2007

In this paper, we implement a hypervisor that runs multiple uC/OS-II real-time kernels on one microprocessor. The hypervisor virtualizes microprocessor and memory that are main resources managed by uC/OS-II kernel. Microprocessor is virtualized by controlling interrupts that uC/OS-II real-time kernel handles and memory is virtualized by partitioning physical memory. The hypervisor consists of three components: interrupt control routines that virtualize timer interrupt and software interrupt, a startup code that initializes the hypervisor and uC/OS-II kernels, and an API that provides communication between two kernels. The original uC/OS-II kernel needs to be modified slightly in source-code level to run on the hypervisor. We performed a real-time test and an independent computation test on Jupiter 32-bit EISC microprocessor and showed that the virtualized kernels run without problem. The result of our research can reduce the hardware cost, the system space and weight, and system power consumption when the hypervisor is applied in embedded applications that require many embedded microprocessors.

• ETRI Journal
• /
• v.29 no.3
• /
• pp.259-269
• /
• 2007

The ubiquitous flexible operating system (UbiFOS) is a real-time operating system designed for cost-conscious, low-power, small to medium-sized embedded systems such as cellular phones, MP3 players, and wearable computers. It offers efficient real-time operating system services like multi-task scheduling, memory management, inter-task communication and synchronization, and timers while keeping the kernel size to just a few to tens of kilobytes. For flexibility, UbiFOS uses various task scheduling policies such as cyclic time-slice (round-robin), priority-based preemption with round-robin, priority-based preemptive, and bitmap. When there are less than 64 tasks, bitmap scheduling is the best policy. The scheduling overhead is under 9 ${\mu}s$ on the ARM926EJ processor. UbiFOS also provides the flexibility for user to select from several inter-task communication techniques according to their applications. We ported UbiFOS on the ARM9-based DVD player (20 kB), the Calm16-based MP3 player (under 7 kB), and the ATmega128-based ubiquitous sensor node (under 6 kB). Also, we adopted the dynamic power management (DPM) scheme. Comparative experimental results show that UbiFOS could save energy up to 30% using DPM.

• The KIPS Transactions:PartA
• /
• v.17A no.4
• /
• pp.181-188
• /
• 2010

In this study, we suggest the system to analyze network users' action patterns by using Data-Mining Technique. We installed Network Tap to implement the analysis system of network action and copied the network packet. The copied packet is stored at the database through MainMemoryDB(MMDB) of the high-speed. The stored data analyze the users' action patterns by using Data-Mining Technique and then report the results to the network manager on real-time. Also, we applied the standard XML document exchange method to share the data between different systems. We propose this action pattern analysis system operated embedded type of SetToBox to install easily and support low price.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.64 no.3
• /
• pp.138-142
• /
• 2015

Most of the existing studies on functional safety testing for the railway signaling system software have focused on verifying the functional safety through the monitoring of internal memory embedded railway signaling system. However, the railway signaling system is one of the typical embedded control system in the railway sector, and the embedded software has a characteristic of generating an appropriate outputs through the combination of internal processing in consideration of the current internal status and external input. Therefore, the test approach of using the interface communication channel can be effective way for the functional testing for railway signaling system software in consideration of these characteristic. Since a communication interface specification of the railway signal system has a the properties of the sequence input and output signals, test-case for software testing is the most effective methodology by MSC (Message Sequence Chart) language, one of the graphic language. The MSC-based test-case generating methodology for signaling system software was proposed in this paper.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 1997.10a
• /
• pp.307-310
• /
• 1997

In case of environment requiring safety such as human body and requiring flexible shape, a conventional mechanical actuator system does not satisfy requirement. Therefore, in order to solve this problem, a research of various smart material such as EAP (Electro Active Polymer), EAC (Electro Active Ceramic) and SMA (Shape Memory Alloy) is in progress. Recently, the highest preferring material among various smart material is EP (Electrostictive Polymer), because it has very fast response time, poerful force and large displacement. The previous researches have been studied properties of polymer and simple control, but present researches are studied a polymer actuator. An EP (Electostrictive Polymer) actuator has properties which change variably as shape and environmental condition. Therefore, in order to coincide with a user's purpose, it is important not only to decide a shape of actuator and mechanical design but also to investigate a efficient controller. In this paper, we constructed the control logic with an adaptive fuvy algorithm which depends on the physical properties of EP that has a dielectric constant depending on time.