• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.27 no.10C
• /
• pp.963-972
• /
• 2002

An 8192-point pipelined FFT/IFFT processor core is designed, which can be used in multi-carrier modulation systems such as DUf-based VDSL modem and OFDM-based DVB system. In order to improve the signal-to-quantization-noise ratio (SQNR) of FFT/IFFT results, two-step convergent block floating-point (TS_CBFP) scaling is employed. Since the proposed TS_CBFP scaling does not require additional buffer memory, it reduces memory as much as about 80% when compared with conventional CBFP methods, resulting in area-and power-efficient implementation. The SQNR of about 60-㏈ is achieved with 10-bit input, 14-bit internal data and twiddle factors, and 16-bit output. The core synthesized using 0.25-$\mu\textrm{m}$ CMOS library has about 76,300 gates, 390K bits RAM, and twiddle factor ROM of 39K bits. Simulation results show that it can safely operate up to 50-㎒ clock frequency at 2.5-V supply, resulting that a 8192-point FFT/IFFT can be computed every 164-${\mu}\textrm{s}$. It was verified by Xilinx FPGA implementation.

• Journal of the Institute of Convergence Signal Processing
• /
• v.14 no.2
• /
• pp.110-116
• /
• 2013

The amplification rate of a RF repeater is limited by the feedbacked signals from the same repeater. And an ICS (Interference Cancellation System) repeater has been developed to remove the feedbacked signals. The ICS repeater estimates the amplitudes and the phases of the feedbacked signals and removes the estimated feedback signals from the received input signal of the repeater. However, it requires lots of hardware complexity and this leads to the increase the cost of the repeater. Moreover, the ICS repeater can not solve the pilot pollution problems. To solve these problems, we have studied the implementation and adaptation of smart antenna system for RF repeaters. We have designed a smart antenna system with a switching beam structure in order to reduce the hardware and computational complexity. After analyzing the proposed smart antenna system, we found out that the amplification rate of the proposed repeater increases 23dB compare to the amplification rate of ICS repeater and the output SINR increases 6dB compare to the ICS repeater.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.6
• /
• pp.592-599
• /
• 2020

This paper proposes a Convolutional Recurrent Neural Net (CRNN) structure that can simultaneously reflect both static and dynamic characteristics of seismic waveforms for various earthquake events classification. Addressing various earthquake events, including not only micro-earthquakes and artificial-earthquakes but also macro-earthquakes, requires both effective feature extraction and a classifier that can discriminate seismic waveform under noisy environment. First, we extract the static characteristics of seismic waveform through an attention-based convolution layer. Then, the extracted feature-map is sequentially injected as input to a multi-input single-output Long Short-Term Memory (LSTM) network structure to extract the dynamic characteristic for various seismic event classifications. Subsequently, we perform earthquake events classification through two fully connected layers and softmax function. Representative experimental results using domestic and foreign earthquake database show that the proposed model provides an effective structure for various earthquake events classification.

• Korean Journal of Optics and Photonics
• /
• v.14 no.4
• /
• pp.473-479
• /
• 2003

The purpose of this study is to recover the CIE XYZ tristimulus values of original colors from scanner output signals, and to reproduce true colors on the monitor. The process of this study is composed of three steps; scanner characterization, chromatic adaptation transformation, and color space transformation between and sRGB. Especially, in the process of recovery, scanner stimuli were obtained accurately by dividing the non-linear photometric response curve into two parts. As the result of test to EPSON Expression 1680 scanner, the average color difference between true and recovered XYZ for 228 target colors, 22 test neutrals, and 36 test colors were 1.49, 0.97, and 1.42 $\Delta$ $E_{UV}$ *, respectively. With the transformation from illuminant D50 to illuminant D65, the input signals to sRGB monitor were predicted. Finally, it could be found that displayed colors with predicted input signals were very consistent with true colors. with true colors.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.51 no.4
• /
• pp.49-58
• /
• 2014

A low-swing differential near-ground signaling (NGS) transceiver for low-power high-speed mobile I/O interface is presented. The proposed transmitter adopts an on-chip regulated programmable-swing voltage-mode driver and a pre-driver with asymmetric rising/falling time. The proposed receiver utilizes a new multiple gain-path differential amplifier with feed-forward capacitors that boost high-frequency gain. Also, the receiver incorporates a new adaptive bias generator to compensate the input common-mode variation due to the variable output swing of the transmitter and to minimize the current mismatch of the receiver's input stage amplifier. The use of the new simple and effective impedance matching techniques applied in the transmitter and receiver results in good signal integrity and high power efficiency. The proposed transceiver designed in a 65-nm CMOS technology achieves a data rate of 13 Gbps/channel and 0.3 pJ/bit (= 0.3 mW/Gbps) high power efficiency over a 10 cm FR4 printed circuit board.

• The Journal of the Acoustical Society of Korea
• /
• v.16 no.6
• /
• pp.94-99
• /
• 1997

This paper describes a technique of probabilistic spectral subtraction which uses the knowledge of both noise and speech so as to reduce automatic speech recognition errors in noisy environments. Spectral subtraction method estimates a noise prototype in non-speech intervals and the spectrum of clean speech is obtained from the spectrum of noisy speech by subtracting this noise prototype. Thus noise can not be suppressed effectively using a single noise prototype in case the characteristics of the noise prototype are different from those of the noise contained in input noisy speech. To modify such a drawback, multiple noise prototypes are used in probabilistic subtraction method. In this paper, the probabilistic characteristics of noise and the knowledge of speech which is embedded in hidden Markov models trained in clean environments are used to suppress noise. Futhermore, dynamic feature parameters are considered as well as static feature parameters for effective noise suppression. The proposed method reduced error rates in the recognition of 50 Korean words. The recognition rate was 86.25% with the probabilistic subtraction, 72.75% without any noise suppression method and 80.25% with spectral subtraction at SNR(Signal-to-Noise Ratio) 10 dB.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.43-43
• /
• 2022

유해 조류 대발생은 전국 각지의 인공호소나 하천에서 다발적으로 발생하며, 경관을 해치고 수질을 오염시키는 등 수자원에 부정적인 영향을 미친다. 본 연구에서는 인공호소에서 발생하는 유해 조류 대발생을 예측하기 위해 심층학습 기법을 이용하여 예측 모델을 개발하고자 하였다. 대상 지점은 대청호의 추동 지점으로 선정하였다. 대청호는 금강유역 중류에 위치한 댐으로, 약 150만명에 달하는 급수 인구수를 유지 중이기에 유해 남조 대발생 관리가 매우 중요한 장소이다. 학습용 데이터 구축은 대청호의 2011년 1월부터 2019년 12월까지 측정된 수질, 기상, 수문 자료를 입력 자료를 이용하였다. 수질 예측 모델의 구조는 다중 레이어 퍼셉트론(Multiple Layer Perceptron; MLP)으로, 입력과 한 개 이상의 은닉층, 그리고 출력층으로 구성된 인공신경망이다. 본 연구에서는 인공신경망의 은닉층 개수(1~3개)와 각각의 레이어에 적용되는 은닉 노드 개수(11~30개), 활성함수 5종(Linear, sigmoid, hyperbolic tangent, Rectified Linear Unit, Exponential Linear Unit)을 각각 하이퍼파라미터로 정하고, 모델의 성능을 최대로 발휘할 수 있는 조건을 찾고자 하였다. 하이퍼파라미터 최적화 도구는 Tensorflow에서 배포하는 Keras Tuner를 사용하였다. 모델은 총 3000 학습 epoch 가 진행되는 동안 최적의 가중치를 계산하도록 설계하였고, 이 결과를 매 반복마다 저장장치에 기록하였다. 모델 성능의 타당성은 예측과 실측 데이터 간의 상관관계를 R2, NSE, RMSE를 통해 산출하여 검증하였다. 모델 최적화 결과, 적합한 하이퍼파라미터는 최적화 횟수 총 300회에서 256 번째 반복 결과인 은닉층 개수 3개, 은닉 노드 수 각각 25개, 22개, 14개가 가장 적합하였고, 이에 따른 활성함수는 ELU, ReLU, Hyperbolic tangent, Linear 순서대로 사용되었다. 최적화된 하이퍼파라미터를 이용하여 모델 학습 및 검증을 수행한 결과, R2는 학습 0.68, 검증 0.61이었고 NSE는 학습 0.85, 검증 0.81, RMSE는 학습 0.82, 검증 0.92로 나타났다.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.46 no.2
• /
• pp.72-77
• /
• 2009

In this paper, a 3.2Gb/s clock and data recovery (CDR) circuit for a high-speed serial data communication without the reference clock is described This CDR circuit consists of 5 parts as Phase and frequency detector(PD and FD), multi-phase Voltage Controlled-Oscillator(VCO), Charge-pumps (CP) and external Loop-Filter(KF). It is adapted the PD and FD, which incorporates a half-rate bang-bang type oversampling PD and a half-rate FD that can improve pull-in range. The VCO consists of four fully differential delay cells with rail-to-rail current bias scheme that can increase the tuning range and tuning linearity. Each delay cell has output buffers as a full-swing generator and a duty-cycle mismatch compensation. This materialized CDR can achieve wide pull-in range without an extra reference clock and it can be also reduced chip area and power consumption effectively because there is no additional Phase Locked- Loop(PLL) for generating reference clock. The CDR circuit was designed for fabrication using 0.18um 1P6M CMOS process and total chip area excepted LF is $1{\times}1mm^2$. The pk-pk jitter of recovered clock is 26ps at 3.2Gb/s input data rate and total power consumes 63mW from 1.8V supply voltage according to simulation results. According to test result, the pk-pk jitter of recovered clock is 55ps at the same input data-rate and the reliable range of input data-rate is about from 2.4Gb/s to 3.4Gb/s.

• The KIPS Transactions:PartB
• /
• v.9B no.2
• /
• pp.139-146
• /
• 2002

The robot soccer simulation game is a dynamic multi-agent environment. In this paper we suggest a new reinforcement learning approach to each agent's dynamic positioning in such dynamic environment. Reinforcement learning is the machine learning in which an agent learns from indirect, delayed reward an optimal policy to choose sequences of actions that produce the greatest cumulative reward. Therefore the reinforcement learning is different from supervised learning in the sense that there is no presentation of input-output pairs as training examples. Furthermore, model-free reinforcement learning algorithms like Q-learning do not require defining or learning any models of the surrounding environment. Nevertheless these algorithms can learn the optimal policy if the agent can visit every state-action pair infinitely. However, the biggest problem of monolithic reinforcement learning is that its straightforward applications do not successfully scale up to more complex environments due to the intractable large space of states. In order to address this problem, we suggest Adaptive Mediation-based Modular Q-Learning (AMMQL) as an improvement of the existing Modular Q-Learning (MQL). While simple modular Q-learning combines the results from each learning module in a fixed way, AMMQL combines them in a more flexible way by assigning different weight to each module according to its contribution to rewards. Therefore in addition to resolving the problem of large state space effectively, AMMQL can show higher adaptability to environmental changes than pure MQL. In this paper we use the AMMQL algorithn as a learning method for dynamic positioning of the robot soccer agent, and implement a robot soccer agent system called Cogitoniks.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.17 no.6
• /
• pp.69-78
• /
• 2017

Shannon of the 5G smartphone and Fourier of the signal processing meet in the sampling theorem (2 times the highest frequency 1). In this paper, the initial Shannon Theorem finds the Shannon capacity at the point-to-point, but the 5G shows on the Relay channel that the technology has evolved into Multi Point MIMO. Fourier transforms are signal processing with fixed parameters. We analyzed the performance by proposing a 2N-1 multivariate Fourier-Jacket transform in the multimedia age. In this study, the authors tackle this signal processing complexity issue by proposing a Jacket-based fast method for reducing the precoding/decoding complexity in terms of time computation. Jacket transforms have shown to find applications in signal processing and coding theory. Jacket transforms are defined to be $n{\times}n$ matrices $A=(a_{jk})$ over a field F with the property $AA^{\dot{+}}=nl_n$, where $A^{\dot{+}}$ is the transpose matrix of the element-wise inverse of A, that is, $A^{\dot{+}}=(a^{-1}_{kj})$, which generalise Hadamard transforms and centre weighted Hadamard transforms. In particular, exploiting the Jacket transform properties, the authors propose a new eigenvalue decomposition (EVD) method with application in precoding and decoding of distributive multi-input multi-output channels in relay-based DF cooperative wireless networks in which the transmission is based on using single-symbol decodable space-time block codes. The authors show that the proposed Jacket-based method of EVD has significant reduction in its computational time as compared to the conventional-based EVD method. Performance in terms of computational time reduction is evaluated quantitatively through mathematical analysis and numerical results.