• Journal of IKEEE
• /
• v.21 no.4
• /
• pp.388-396
• /
• 2017

Since there are many variables such as various poses, illuminations and occlusions in a face detection problem, a high performance detection system is required. Although CNN is excellent in image classification, CNN operatioin requires high-performance hardware resources. But low cost low power environments are essential for small and mobile systems. So in this paper, the CPU-FPGA integrated system is designed based on 3-stage cascade CNN architecture using small size FPGA. Adaptive Region of Interest (ROI) is applied to reduce the number of CNN operations using face information of the previous frame. We use a Field Programmable Gate Array(FPGA) to accelerate the CNN computations. The accelerator reads multiple featuremap at once on the FPGA and performs a Multiply-Accumulate (MAC) operation in parallel for convolution operation. The system is implemented on Altera Cyclone V FPGA in which ARM Cortex A-9 and on-chip SRAM are embedded. The system runs at 30FPS with HD resolution input images. The CPU-FPGA integrated system showed 8.5 times of the power efficiency compared to systems using CPU only.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.10 no.10
• /
• pp.1123-1130
• /
• 2015

Network on a chip(NoC) is a communication subsystem between intellectual property(IP) cores in a SoC and improves high performance in the scalability and the power efficiency compared with conventional buses and crossbar switches. NoC needs a synchronizer to overcome the metastability problem between data links. This paper presents a new mesochronous synchronizer(MS) which is composed of selection window generator, selection signal generator, and data buffer. A delay line circuit is used to build selection window in selection window generator based on the delayed clock cycle of transmitted clock and the transmitted clock is compared with local clock to generate a selection signal in the SW(selection window). This MS gets rid of the restriction of metastability by choosing a rising edge or a falling edge of local clock according to the value of selection signal. The simulation results show that the proposed MS operates correctly for all phase differences between a transmitted clock and a local clock.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.20 no.2
• /
• pp.38-44
• /
• 2006

In this paper, a new interference-free Home CDMA-PLC(Code Division Multiple Access-Power Line Communication) system based on the binary ZCD(Zero Correlation Duration) spreading code is proposed as a key solution to overcome the previous problems. Binary ZCD spreading code sets with enlarged family sizes are generated by carrying out a chip-shift operation of the preferred pairs. The properties or the proposed ZCD-PLC systems are effective for MPI(Multi-Path Interference) and MAI (Multiple Access Interference) cancellation in the CDMA-PLC systems. By BER performance simulation, we certified the availability of proposed ZCD-CDMA-PLC system.

• Smart Structures and Systems
• /
• v.11 no.3
• /
• pp.261-282
• /
• 2013

Structural health monitoring (SHM) is an application area of Wireless Sensor Networks (WSNs) which usually needs high data communication rate to transfer a large amount of monitoring data. Traditional sink node can only process data from one communication channel at the same time because of the single radio chip structure. The sink node constitutes a bottleneck for constructing a high data rate SHM application giving rise to a long data transfer time. Multi-channel communication has been proved to be an efficient method to improve the data throughput by enabling parallel transmissions among different frequency channels. This paper proposes an 8-radio integrated sink node design method based on Field Programmable Gate Array (FPGA) and the time synchronization mechanism for the multi-channel network based on the proposed sink node. Three experiments have been performed to evaluate the data transfer ability of the developed multi-radio sink node and the performance of the time synchronization mechanism. A high data throughput of 1020Kbps of the developed sink node has been proved by experiments using IEEE.805.15.4.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.15 no.10
• /
• pp.889-895
• /
• 2002

Wafer level packaging is gain mote momentum as a low cost, high performance solution for RF-MEMS devices. In this work, the flip-chip method was used for the wafer level packaging of RF-MEMS devices on the quartz substrate with low losses. For analyzing the EM (electromagnetic) characteristic of proposed packaging structure, we got the 3D structure simulation using FEM (finite element method). The electric field distribution of CPW and hole feed-through at 3 GHz were concentrated on the hole and the CPW. The reflection loss of the package was totally below 23 dB and the insertion loss that presents the signal transmission characteristic is above 0.06 dB. The 4-inch Pyrex glass was used as a package substrate and it was punched with air-blast with 250${\mu}{\textrm}{m}$ diameter holes. We made the vortical feed-throughs to reduce the electric path length and parasitic parameters. The vias were filled with plating gold. The package substrate was bonded with the silicon substrate with the B-stage epoxy. The loss of the overall package structure was tested with a network analyzer and was within 0.05 dB. This structure can be used for wafer level packaging of not only the RF-MEMS devices but also the MEMS devices.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.6
• /
• pp.571-577
• /
• 2021

This study is a study on System On Chip (SOC) that implements virtual engine sound in electric vehicles without engines, and realizes vivid engine sound by combining Adaptive Difference PCM (ADPCM) method and frequency modulation method for satisfaction of driver's needs and safety of pedestrians. In addition, by proposing an electronic sound synthesis algorithm applying Musical Instrument Didital Interface (MIDI), an engine sound synthesis method and a constitutive model of an engine sound generation system are presented. In order to satisfy both drivers and pedestrians, this study uses Controller Area Network (CAN) communication to receive information such as Revolution Per Minute (RPM), vehicle speed, accelerator pedal depressed amount, torque, etc., transmitted according to the driver's driving habits, and then modulates the frequency according to the appropriate preset parameters We implemented an interaction algorithm that accurately reflects the intention of the system and driver by using interpolation for the system, ADPCM algorithm for reducing the amount of information, and MIDI format information for making engine sound easier.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.48 no.4
• /
• pp.35-42
• /
• 2011

Link quality assessment is a crucial part of sensor network formation to stably operate large-scale wireless sensor networks (WSNs). A stability of path consisting of several nodes strongly depends on all link quality between pair of consecutive nodes. Thus it is very important to assess the link quality on the stage of building a routing path. In this paper, we present a link quality assessment method, Hybrid Link Quality Metric (HQLM), which uses both of LQI and RSSI from RF chip of sensor nodes to minimize set-up time and energy consumption for network formation. The HQLM not only reduces the time and energy consumption, but also provides complementary cooperation of LQI and RSSI. In order to evaluate the validity and efficiency of the proposed method, we measure PDR (Packet Delivery Rate) by exchanging multiple messages and then, compare PDR to the result of HQLM for evaluation. From the research being carried out, we can conclude that the HQLM performs better than either LQI- or RSSI-based metric in terms of recall, precision, and matching on link quality.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.44 no.8
• /
• pp.52-59
• /
• 2007

In this paper, we analyzed and measured implant isolation characteristics for a 1.25 Gbps monolithic integrated hi-directional (M-BiDi) optoelectronic system-on-a-chip, which is a key component to constitute gigabit passive optical networks (PONs) for a fiber-to-the-home (FTTH). Also, we derived an equivalent circuit of the implant structure under various DC bias conditions. The 1.25 Gbps M-BiDi transmit-receive SoC consists of a laser diode with a monitor photodiode as a transmitter and a digital photodiode as a digital data receiver on the same InP wafer According to IEEE 802.3ah and ITU-T G.983.3 standards, a receiver sensitivity of the digital receiver has to satisfy under -24 dBm @ BER=10-12. Therefore, the electrical crosstalk levels have to maintain less than -86 dB from DC to 3 GHz. From analysed and measured results of the implant structure, the M-BiDi SoC with the implant area of 20 mm width and more than 200 mm distance between the laser diode and monitor photodiode, and between the monitor photodiode and digital photodiode, satisfies the electrical crosstalk level. These implant characteristics can be used for the design and fabrication of an optoelectronic SoC design, and expended to a mixed-mode SoC field.

• The Korean Journal of Air & Space Law and Policy
• /
• v.32 no.2
• /
• pp.155-188
• /
• 2017

On the one hand, the 4th Industrial Revolution provides a positive opportunity to build a new civilization paradigm for mankind. However, on the other hand, due to the 4th Industrial Revolution, artificial intelligence such as 'Goggle Alpha Go' revolutionized and even the human ability was replaced with a 'Silicon Chip' as the opportunity to communicate decreases, the existence of human beings is weakened. And there is a growing concern that the number of violent crimes, such as psychopath, which hunts humans as games, will increase. Moreover, recent international terrorism is being developed in a form similar to 'Psychopathic Violent-Crime' that indiscriminately attacks innocent people. So, the probability that terrorist organizations abuse the positive effects provided by the Fourth Industrial Revolution as means of terrorism is increasing. Therefore, the paradigm of aviation terrorism is expected to change in a way that attacks airport facilities and users rather than aircraft. Because airport facilities are crowded, and psychopathic terrorists are easily accessible. From this point of view, our counter terrorism system of aviation has many weak points in various aspects such as: (1) limitations of counter-terrorism center (2) inefficient on-site command and control system (3) separated organization for aviation security consultation (4) dispersed information collection function in government (5) vulnerable to cyber attack (6) lack of international cooperation network for aviation terrorism. Consequently, it is necessary to improve the domestic counter terrorism system of aviation so as to preemptively respond to the international terrorism. This study propose the following measures to improve the aviation security system by (1) create 'Aviation Special Judicial Police' (2) revise the anti-terrorism law and aviation security law (3) Strengthening the ability respond to terrorism in cyberspace (4) building an international cooperation network for aviation terrorism.

• Journal of Broadcast Engineering
• /
• v.25 no.2
• /
• pp.200-207
• /
• 2020

Recently, deep neural networks such as CNN are showing excellent performance in various fields such as image classification, object recognition, visual quality enhancement, etc. However, as the model size and computational complexity of deep learning models for most applications increases, it is hard to apply neural networks to IoT and mobile environments. Therefore, neural network compression algorithms for reducing the model size while keeping the performance have been being studied. In this paper, we apply few compression methods to CNN models and evaluate their performances in the embedded environment. For evaluate the performance, the classification performance and inference time of the original CNN models and the compressed CNN models on the image inputted by the camera are evaluated in the embedded board equipped with QCS605, which is a customized AI chip. In this paper, a few CNN models of MobileNetV2, ResNet50, and VGG-16 are compressed by applying the methods of pruning and matrix decomposition. The experimental results show that the compressed models give not only the model size reduction of 1.3~11.2 times at a classification performance loss of less than 2% compared to the original model, but also the inference time reduction of 1.2~2.21 times, and the memory reduction of 1.2~3.8 times in the embedded board.