• Journal of IKEEE
• /
• v.25 no.2
• /
• pp.273-279
• /
• 2021

In this paper, we propose an algorithm that shortens the learning time by performing individual learning using partitioning the deep learning structure. The proposed algorithm consists of four processes: network classification origin setting process, feature vector extraction process, feature noise removal process, and class classification process. First, in the process of setting the network classification starting point, the division starting point of the network structure for effective feature vector extraction is set. Second, in the feature vector extraction process, feature vectors are extracted without additional learning using the weights previously learned. Third, in the feature noise removal process, the extracted feature vector is received and the output value of each class is learned to remove noise from the data. Fourth, in the class classification process, the noise-removed feature vector is input to the multi-layer perceptron structure, and the result is output and learned. To evaluate the performance of the proposed algorithm, we experimented with the Extended Yale B face database. As a result of the experiment, in the case of the time required for one-time learning, the proposed algorithm reduced 40.7% based on the existing algorithm. In addition, the number of learning up to the target recognition rate was shortened compared with the existing algorithm. Through the experimental results, it was confirmed that the one-time learning time and the total learning time were reduced and improved over the existing algorithm.

• Journal of Broadcast Engineering
• /
• v.27 no.2
• /
• pp.185-197
• /
• 2022

Recently, deep learning networks with high performance for object recognition are emerging. In the case of object recognition using deep learning, it is important to build a training data set to improve performance. To build a data set, we need to collect and label the images. This process requires a lot of time and manpower. For this reason, open data sets are used. However, there are objects that do not have large open data sets. One of them is data required for license plate detection and recognition. Therefore, in this paper, we propose an artificial license plate generator system that can create large data sets by minimizing images. In addition, the detection rate according to the artificial license plate arrangement structure was analyzed. As a result of the analysis, the best layout structure was FVC_III and B, and the most suitable network was D2Det. Although the artificial data set performance was 2-3% lower than that of the actual data set, the time to build the artificial data was about 11 times faster than the time to build the actual data set, proving that it is a time-efficient data set building system.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.17 no.6
• /
• pp.1013-1024
• /
• 2022

In addition to the assessment measure of electric quality levels, load loss are also a factor in hindering the financial profits of electrical sales companies. Therefore, accurate analysis of load losses generated from distributed power networks is very important. The accurate calculation of load losses in the distribution line has been carried out for a long time in many research institutes as well as power utilities around the world. But it is increasingly difficult to calculate the exact amount of loss due to the increase in the congestion of distribution power network due to the linkage of distributed energy resources(DER). In this paper, we develop smart grid big data infrastructure in order to accurately analyze the load loss of the distribution power network due to the connection of DERs. Through the preprocess of data selected from the smart grid big data, we develop a load loss analysis model that eliminated 'veracity' which is one of the characteristics of smart grid big data. Our analysis results can be used for facility investment plans or network operation plans to maintain stable supply reliability and power quality.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.11
• /
• pp.1627-1634
• /
• 2021

This paper introduces a low-power compact ADC circuit for analog Convolutional filter for low-power neural network accelerator SOC. While convolutional neural network accelerators can speed up the learning and inference process, they have drawback of consuming excessive power and occupying large chip area due to large number of multiply-and-accumulate operators when implemented in complex digital circuits. To overcome these drawbacks, we implemented an analog convolutional filter that consists of an analog multiply-and-accumulate arithmetic circuit along with an ADC. This paper is focused on the design optimization of a low-power 8bit SAR ADC for the analog convolutional filter accelerator We demonstrate how to minimize the capacitor-array DAC, an important component of SAR ADC, which is three times smaller than the conventional circuit. The proposed ADC has been fabricated in CMOS 65nm process. It achieves an overall size of 1355.7㎛², power consumption of 2.6㎼ at a frequency of 100MHz, SNDR of 44.19 dB, and ENOB of 7.04bit.

• Korean Journal of Radiology
• /
• v.22 no.11
• /
• pp.1918-1928
• /
• 2021

Objective: With the recent development of various MRI-conditional cardiac implantable electronic devices (CIEDs), the accurate identification and characterization of CIEDs have become critical when performing MRI in patients with CIEDs. We aimed to develop and evaluate a deep learning-based algorithm (DLA) that performs the detection and characterization of parameters, including MRI safety, of CIEDs on chest radiograph (CR) in a single step and compare its performance with other related algorithms that were recently developed. Materials and Methods: We developed a DLA (X-ray CIED identification [XCID]) using 9912 CRs of 958 patients with 968 CIEDs comprising 26 model groups from 4 manufacturers obtained between 2014 and 2019 from one hospital. The performance of XCID was tested with an external dataset consisting of 2122 CRs obtained from a different hospital and compared with the performance of two other related algorithms recently reported, including PacemakerID (PID) and Pacemaker identification with neural networks (PPMnn). Results: The overall accuracies of XCID for the manufacturer classification, model group identification, and MRI safety characterization using the internal test dataset were 99.7% (992/995), 97.2% (967/995), and 98.9% (984/995), respectively. These were 95.8% (2033/2122), 85.4% (1813/2122), and 92.2% (1956/2122), respectively, with the external test dataset. In the comparative study, the accuracy for the manufacturer classification was 95.0% (152/160) for XCID and 91.3% for PPMnn (146/160), which was significantly higher than that for PID (80.0%,128/160; p < 0.001 for both). XCID demonstrated a higher accuracy (88.1%; 141/160) than PPMnn (80.0%; 128/160) in identifying model groups (p < 0.001). Conclusion: The remarkable and consistent performance of XCID suggests its applicability for detection, manufacturer and model identification, as well as MRI safety characterization of CIED on CRs. Further studies are warranted to guarantee the safe use of XCID in clinical practice.

• 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.

• Journal of the Institute of Electronics Engineers of Korea TC
• /
• v.46 no.12
• /
• pp.6-13
• /
• 2009

For X-band phased array systems, a power amplifier, a 6-bit phase shifter, a 6-bit digital attenuator, and a SPDT transmit/receive (T/R) switch are fabricated and measured. All circuits are demonstrated by using CMOS 0.18 um technology. The power amplifier has 2-stage differential and cascade structures. It provides 1-dB gain-compressed output power ($P_{1dB}$) of 20 dBm and power-added-efficiency (PAE) of 19 % at 8-11 GHz frequencies. The 6-bit phase shifter utilizes embedded switched filter structure which consists of nMOS transistors as a switch and meandered microstrip lines for desired inductances. It has $360^{\circ}$ phase-control range and $5.6^{\circ}$ phase resolution. At 8-11 GHz frequencies, it has RMS phase and amplitude errors are below $5^{\circ}$ and 0.8 dB, and insertion loss of $-15.7\;{\pm}\;1,1\;dB$. The 6-bit digital attenuator is comprised of embedded switched Pi-and T-type attenuators resistive networks and nMOS switches and employes compensation circuits for low insertion phase variation. It has max. attenuation of 31.5 dB and 0.5 dB amplitude resolution. Its RMS amplitude and phase errors are below 0.4 dB and $2^{\circ}$ at 8-11 GHz frequencies, and insertion loss is $-10.5\;{\pm}\;0.8\;dB$. The SPDT T/R switch has series and shunt transistor pairs on transmit and receive path, and only one inductance to reduce chip area. It shows insertion loss of -1.5 dB, return loss below -15 dB, and isolation about -30 dB. The fabricated chip areas are $1.28\;mm^2$, $1.9mm^2$, $0.34\;mm^2$, $0.02mm^2$, respectively.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.28 no.6C
• /
• pp.648-658
• /
• 2003

As the usage of image/video contents increase, a security problem for the payed image data or the ones requiring confidentiality is raised. This paper proposed an image encryption methodology to hide the image information. The target data of it is the result from quantization in wavelet domain. This method encrypts only part of the image data rather than the whole data of the original image, in which three types of data selection methodologies were involved. First, by using the fact that the wavelet transform decomposes the original image into frequency sub-bands, only some of the frequency sub-bands were included in encryption to make the resulting image unrecognizable. In the data to represent each pixel, only MSBs were taken for encryption. Finally, pixels to be encrypted in a specific sub-band were selected randomly by using LFSR(Linear Feedback Shift Register). Part of the key for encryption was used for the seed value of LFSR and in selecting the parallel output bits of the LFSR for random selection so that the strength of encryption algorithm increased. The experiments have been performed with the proposed methods implemented in software for about 500 images, from which the result showed that only about 1/1000 amount of data to the original image can obtain the encryption effect not to recognize the original image. Consequently, we are sure that the proposed are efficient image encryption methods to acquire the high encryption effect with small amount of encryption. Also, in this paper, several encryption scheme according to the selection of the sub-bands and the number of bits from LFSR outputs for pixel selection have been proposed, and it has been shown that there exits a relation of trade-off between the execution time and the effect of the encryption. It means that the proposed methods can be selectively used according to the application areas. Also, because the proposed methods are performed in the application layer, they are expected to be a good solution for the end-to-end security problem, which is appearing as one of the important problems in the networks with both wired and wireless sections.

• Journal of IKEEE
• /
• v.27 no.4
• /
• pp.464-471
• /
• 2023

In this paper, we propose a method to reduce age distortion in facial expression image generation using StyleGAN Encoder. The facial expression image generation process first creates a face image using StyleGAN Encoder, and changes the expression by applying the learned boundary to the latent vector using SVM. However, when learning the boundary of a smiling expression, age distortion occurs due to changes in facial expression. The smile boundary created in SVM learning for smiling expressions includes wrinkles caused by changes in facial expressions as learning elements, and it is determined that age characteristics were also learned. To solve this problem, the proposed method calculates the correlation coefficient between the smile boundary and the age boundary and uses this to introduce a method of adjusting the age boundary at the smile boundary in proportion to the correlation coefficient. To confirm the effectiveness of the proposed method, the results of an experiment using the FFHQ dataset, a publicly available standard face dataset, and measuring the FID score are as follows. In the smile image, compared to the existing method, the FID score of the smile image generated by the ground truth and the proposed method was improved by about 0.46. In addition, compared to the existing method in the smile image, the FID score of the image generated by StyleGAN Encoder and the smile image generated by the proposed method improved by about 1.031. In non-smile images, compared to the existing method, the FID score of the non-smile image generated by the ground truth and the method proposed in this paper was improved by about 2.25. In addition, compared to the existing method in non-smile images, it was confirmed that the FID score of the image generated by StyleGAN Encoder and the non-smile image generated by the proposed method improved by about 1.908. Meanwhile, as a result of estimating the age of each generated facial expression image and measuring the estimated age and MSE of the image generated with StyleGAN Encoder, compared to the existing method, the proposed method has an average age of about 1.5 in smile images and about 1.63 in non-smile images. Performance was improved, proving the effectiveness of the proposed method.