http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
High accuracy, Low Power Spread Spectrum Clock Generator to Reduce EMI for Automotive Applications
Lee, Dongsoo,Choi, Jinwook,Oh, Seongjin,Kim, SangYun,Lee, Kang-Yoon The Institute of Electronics and Information Engin 2014 IEIE Transactions on Smart Processing & Computing Vol.3 No.6
This paper presents a Spread Spectrum Clock Generator (SSCG) based on Relaxation oscillator using Up/Down Counter. The current is controlled by a counter and the spread spectrum of the Relaxation Oscillator. A Relaxation Oscillator with temperature compensation using the BGR and ADC is presented. The current to determine the frequency of the Relaxation Oscillator can be controlled. The output frequency of the temperature can be compensated by adjusting the current according to the temperature using the code that is the output from the ADC and BGR. EMI Reduction of SSCG is 11 dB, and Spread down frequency is 150 kHz. The current consumption is $600{\mu}A$ from 5V and the operating frequency is from 2.3 MHz to 5.75 MHz. The rate of change of the output frequency with temperature was approximately ${\pm}1%$. The SSCG is fabricated in a 0.35um CMOS process with active area $250um{\times}440um$.
Extended Siamese Convolutional Neural Networks for Discriminative Feature Learning
Sangyun Lee,홍성준 한국지능시스템학회 2022 INTERNATIONAL JOURNAL of FUZZY LOGIC and INTELLIGE Vol.22 No.4
Siamese convolutional neural networks (SCNNs) has been considered as among the bestdeep learning architectures for visual object verification. However, these models involvethe drawback that each branch extracts features independently without considering the otherbranch, which sometimes lead to unsatisfactory performance. In this study, we propose a newarchitecture called an extended SCNN (ESCNN) that addresses this limitation by learningboth independent and relative features for a pair of images. ESCNNs also have a featureaugmentation architecture that exploits the multi-level features of the underlying SCNN. Theresults of feature visualization showed that the proposed ESCNN can encode relative anddiscriminative information for the two input images at multi-level scales. Finally, we appliedan ESCNN model to a person verification problem, and the experimental results indicate thatthe ESCNN achived an accuracy of 97.7%, which outperformed an SCNN model with 91.4%accuracy. The results of ablation studies also showed that a small version of the ESCNNperformed 5.6% better than an SCNN model.