• Journal of Korea Multimedia Society
• /
• v.18 no.9
• /
• pp.1039-1046
• /
• 2015

We propose a feature generation method for low-resolution face recognition. For this, we first generate new features from the input features (pixels) of a low-resolution face image by adding the higher-order terms. Then, we evaluate the separability of both of the original input features and new features by computing the discriminant distance of each feature. Finally, new data sample used for recognition consists of the features with high separability. The experimental results for the FERET, CMU-PIE and Yale B databases show that the proposed method gives good recognition performance for low-resolution face images compared with other method.

• Journal of Korea Multimedia Society
• /
• v.21 no.12
• /
• pp.1467-1472
• /
• 2018

For the problem that the face image of surveillance video cannot be accurately identified due to the low resolution, this paper proposes a low resolution face recognition solution based on convolutional neural network model. Convolutional Neural Networks (CNN) model for multi-scale input The CNN model for multi-scale input is an improvement over the existing "two-step method" in which low-resolution images are up-sampled using a simple bi-cubic interpolation method. Then, the up sampled image and the high-resolution image are mixed as a model training sample. The CNN model learns the common feature space of the high- and low-resolution images, and then measures the feature similarity through the cosine distance. Finally, the recognition result is given. The experiments on the CMU PIE and Extended Yale B datasets show that the accuracy of the model is better than other comparison methods. Compared with the CMDA_BGE algorithm with the highest recognition rate, the accuracy rate is 2.5%~9.9%.

• Proceedings of the KIPE Conference
• /
• 2001.10a
• /
• pp.708-712
• /
• 2001

A new phase-shifted parallel-input/series-output (PI SO) dual inductor-fed push-pull converter for high-power step­up applications is proposed. This converter is operated at a constant duty cycle and employs an auxiliary circuit to control the output voltage with a phase-shift between the two modules. It features a voltage conversion characteristic which is linear to changes in the control input, and high step-up ratio with a greatly reduced switch turn-off stress resulting in a significant increase in the converter efficiency. It also shows a low ripple content and low root-mean-square (RMS) current in the output capacitor. The operational principle is analyzed and a comparative analysis with the conventional pulse-width-modulated (PWM) PISO dual inductor-fed push-pull converter is presented. A 50kHz, 800W, 350Vdc prototype with an input of 20-32Vdc has also been constructed to validate the proposed converter. The proposed converter compares favorably with the conventional counterpart and is considered well suited to high-power step-up applications.

• Journal of Welding and Joining
• /
• v.8 no.2
• /
• pp.40-52
• /
• 1990

Fatigue behavior of the AH, DH and EH grade TMCP(Thermo-Mechanical Control Process) steels was studied. High cycle and low cycle fatigue tests were carried out for the weldment and base metal of each steel. The results showed that the fatigue limit at 2 * $10^6$ cycles was 33 to 37 kg/$mm^2$ for the base metal and 30 to 34 kg/$mm^2$ for the weldment. The ratio of fatigue limit to tensile strength for TMCP steels was 0.65 to 0.71, which was a value close to the upper limit for the ordinary steels. It was also found that the high cycle fatigue behavior of TMCP steels could be affected by the microstructures of base metal. It will be necessary to have fine structure for TMCP steels to increase the fatigue resistance. In low cycle fatigue test, the fatigue lifetime of AH and DH steels accorded well with the ASME best fit curve, while that of EH steel was considerably lower than the fatigue lifetime of the other steels. Fatigue resistance of the weldment made by high heat input(180kJ/cm) welding was not lower than that made by low heat input(80kJ/cm) welding in case of high cycle fatigue, but the high heat input welding decreased the fatigue resistance in case of low cycle fatigue.

• Journal of Sensor Science and Technology
• /
• v.26 no.5
• /
• pp.342-347
• /
• 2017

We present a precision instrument amplifier (IA) designed for bio-potential acquisition. The proposed IA employs a capacitively coupled instrument amplifier (CCIA) structure to achieve a rail-to-rail input common-mode range and low gain error. A positive feedback loop is applied to boost the input impedance. Also, DC servo loop (DSL) with pseudo resistors is adopted to suppress electrode offset for bio-potential sensing. The proposed amplifier was designed in a $0.18{\mu}m$ CMOS technology with 1.8V supply voltage. Simulation results show the integrated noise of $1.276{\mu}Vrms$ in a frequency range from 0.01 Hz to 1 KHz, 65dB SNR, 118dB CMRR, and $58M{\Omega}$ input impedance respectively. The total current of IA is $38{\mu}A$. It occupies $740{\mu}m$ by $1300{\mu}m$ including the passive on-chip low pass filter.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.7C
• /
• pp.587-593
• /
• 2010

In this paper we design an irregular low-density parity-check (LDPC) code for multiple-input multiple-output (MIMO) system, using a simple extrinsic information transfer (EXIT) chart method. The MIMO systems considered are optimal maximum a posteriori probability (MAP) detector. The MIMO detector and the LDPC decoder exchange soft information and form a turbo iterative receiver. The EXIT charts are used to obtain the edge degree distribution of the irregular LDPC code which is optimized for the MIMO detector. It is shown that the performance of the designed LDPC code is better than that of conventional LDPC code which was optimized for either the Additive White Gaussian Noise (AWGN) channel or the MIMO channel.

• Journal of Advanced Navigation Technology
• /
• v.8 no.2
• /
• pp.176-183
• /
• 2004

In this paper, two-stage low noise amplifier(LNA) for S-band is designed and implemented using ATF54143 HEMT of HP CO. In order to get noise figure and input VSWR to be wanted, it is considered input VSWR and noise figure simultaneously in matching-circuit designing. The fabricated two-stage low noise amplifier has the gain of 27.8dB, input VSWR and output VSWR under 1.5.

• Proceedings of the KIPE Conference
• /
• 2019.07a
• /
• pp.83-85
• /
• 2019

The use of high voltage gain converters is essential for the distributed power generation systems with renewable energy sources such as the fuel cells and solar cells due to their low voltage characteristics. In this paper, a high voltage gain topology combining cascode Inverting Buck-Boost converter and voltage multiplier structure is introduced. In proposed converter, the input voltage is connected in series at the output, the portion of input power is directly delivered to the load which results in continuous input current. In addition, the voltage multiplier stage stacked in proper manner is not only enhance high step-up voltage gain ratio but also significantly reduce the voltage stress across all semiconductor devices and capacitors. As a result, the high current-low voltage switches can be employed for higher efficiency and lower cost. In order to show the feasibility of the proposed topology, the operation principle is presented and the steady-state characteristic is analyzed in detail. A 380W-40/380V prototype converter was built to validate the effectiveness of proposed converter.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.7
• /
• pp.2328-2344
• /
• 2022

2D human pose estimation still faces difficulty in low-resolution images. Most existing top-down approaches scale up the target human bonding box images to the large size and insert the scaled image into the network. Due to up-sampling, artifacts occur in the low-resolution target images, and the degraded images adversely affect the accurate estimation of the joint positions. To address this issue, we propose a multi-resolution input feature fusion network for human pose estimation. Specifically, the bounding box image of the target human is rescaled to multiple input images of various sizes, and the features extracted from the multiple images are fused in the network. Moreover, we introduce a guiding channel which induces the multi-resolution input features to alternatively affect the network according to the resolution of the target image. We conduct experiments on MS COCO dataset which is a representative dataset for 2D human pose estimation, where our method achieves superior performance compared to the strong baseline HRNet and the previous state-of-the-art methods.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.65 no.4
• /
• pp.586-592
• /
• 2016

The current control of Continuous Conduction Mode(CCM) can be implemented by several methods: peak current control; average current control; and hysteresis control. Among these methods, the hysteresis current control is popularly applied in various converter applications because of its simplicity of implementation, fast current control response and inherent peak current limiting capability. However, a current controller with conventional hysteresis band which multiplies the current reference has the disadvantage that the modulation frequency varies in one cycle of the input voltage and, as a result, generates high switching frequency in the low input voltage section. Also it is complicated to design the input filter due to varying switching frequency. This paper proposed an optimum hysteresis-band current control method where the band is generated by using both multiplication method and sum method to maintain the modulation frequency to be nearly constant. This approach can solve the high switching frequency in the low input voltage section, and achieve easy design of input filter. The performance of the proposed converter is verified with the simulation and the experimental works.