• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.07a
• /
• pp.546-549
• /
• 2020

본 논문에서는 최신 비디오 코딩 기술에서 잔차(Residual)신호 변환을 효율적으로 수행하기 위한 부동기저(Basis)를 사용하는 방법을 제안한다. 기존의 DCT-II 나 DST-VII 과 같은 고정 기저를 사용하는 방법은 대부분의 잔차신호들에 대해 효과적으로 비상관화(decorrelation)를 수행하지만 복잡한 잔차 신호일수록 성능이 떨어지는 문제가 있었다. 이러한 압축 성능하락 문제를 줄이기 위하여 PCA(Principle Component Analysis) 방법 중 하나인 KLT(Karhunen-Loeve Transform)를 이용하여 부동(floating) 변환 기저를 유도하는 방법을 제안한다. 기존의 KLT 를 이용한 변환 커널 유도 방법들의 문제점인 부호화기 및 복호화기 계산 복잡도를 줄이기 위하여 KL 커널을 분해가능한(Separable) 2 개의 1 차원 커널로 유도하는 방법을 제안하고, 원본 잔차신호와 유사한 텍스처를 찾아 커널을 예측하는 과정을 간소화하는 방법을 제안한다. 제안하는 방법은 HEVC 에서 실험되었으며 정지영상 코딩 Main-Profile 에서 평균 1.4%가량의 BD-PSNR(Bjontegaard Delta-Peak Signal to Noise Ratio) 성능 향상을 보였으며 특히 스크린 컨텐츠 영상에서 최대 4.5%의 성능 향상을 보인다.

• The Magazine of the IEIE
• /
• v.41 no.6
• /
• pp.18-26
• /
• 2014

신호의 표현과 압축은 밀접하게 연관되어 있다. 만약 매우 효과적인 신호 표현 방식을 찾을 수 있다면 매우 높은 비율로 신호 압축이 가능하다. 또한 효과적인 신호 표현 방식을 통해 우수한 성능과 낮은 복잡도를 갖는 신호 추정 방식을 유도할 수도 있다. 효과적인 신호 표현 방식은 대상 신호 자체의 성질과 관련되어 있다. 영상 신호등 매우 일반적인 신호가 적절한 변환을 통해 산재된 신호(sparse signal)가 될 수 있음을 많은 연구를 통해 볼 수 있다. 이러한 사실이 compressive sensing(CS)의 기반이다. 즉 신호가 어떠한 변환을 통해 산재된 신호로 표현될 수 있다면 매우 적은 수의 샘플로 이러한 신호를 알아낼 수 있다는 것을 Donoho와 Candes 등이 보였고 이것이 가능한 다양한 조건에 등에 대해 연구되었다. CS는 신호 처리에 근본적인 문제인 효과적인 신호 표현 방식에 직접 연관되어 매우 다양한 분야에 적용될 수 있다. 이 논문에서 CS의 기본적인 개념을 소개한 후 CS가 레이더 신호 처리에 어떻게 도움이 될 수 있는지 살펴본다.

• Proceedings of the Korea Institute of Convergence Signal Processing
• /
• 2000.08a
• /
• pp.21-24
• /
• 2000

Fractional Brownian motion(fBm)은 long-term persistence 특성을 가진 자연 현상, 1/f 잡음, 깊이가 낮은 해저에서의 배경음향잡음 등을 모델링하는데 많이 사용된다. 이 fBm은 nonstationary 유색잡음이다. 이러한 유색잡음 환경 하에서 신호를 검출하기 위한 한 방법은 Fredholm 적분방정식의 해를 구하는 것이다. 이 방정식을 이산화 하면 잡음의 공분산 행렬의 역행렬이 포함되어 계산량이 많다 본 논문에서는 fBm 잡음의 공분산 행렬을 웨이브렛 변환하여 얻어지는 행렬, 즉 fBm의 멀티스케일 성분들의 공분산행렬은 밴드화된 블록들로 근사화할 수 있다는 성질을 이용하여 적은 계산량으로 신호를 검출하는 알고리즘을 제안한다.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.40 no.1
• /
• pp.19-29
• /
• 2003

This paper proposes a CMOS readout circuit for uncooled micro-bolometer arrays adopting a four-point step calibration technique. The proposed readout circuit employing an 11b analog-to-digital converter (ADC), a 7b digital-to-analog converter (DAC), and an automatic gain control circuit (AGC) extracts minute infrared (IR) signals from the large output signals of uncooled micro-bolometer arrays including DC bias currents, inter-pixel process variations, and self-heating effects. Die area and Power consumption of the ADC are minimized with merged-capacitor switching (MCS) technique adopted. The current mirror with high linearity is proposed at the output stage of the DAC to calibrate inter-pixel process variations and self-heating effects. The prototype is fabricated on a double-poly double-metal 1.2 um CMOS process and the measured power consumption is 110 ㎽ from a 4.5 V supply. The measured differential nonlinearity (DNL) and integrat nonlinearity (INL) of the 11b ADC show $\pm$0.9 LSB and $\pm$1.8 LSB, while the DNL and INL of the 7b DAC show $\pm$0.1 LSB and $\pm$0.1 LSB.

• Journal of Broadcast Engineering
• /
• v.21 no.5
• /
• pp.803-815
• /
• 2016

In this paper, we design a direct radio frequency (RF) sampling receiver for multiband GNSS signals and demonstrate its performance. The direct RF sampling is a technique that does not use an analog mixer, but samples the passband signal directly, and all receiver processes are done in digital domain, whereas the conventional intermediate frequency (IF) receiver samples the IF band signals. In contrast to the IF sampling receiver, the RF sampling receiver is less complex in hardware, reconfigurable, and simultaneously converts multiband signals to digital signals with an analog-to-digital (AD) converter. The reconfigurability and simultaneous reception are very important in military applications where rapid change to other system is needed when a system is jammed by an enemy. For simultaneous reception of multiband signals, the sampling frequency should be selected with caution by considering the carrier frequencies, bandwidths, desired intermediate frequencies, and guard bands. In this paper, we select a sampling frequency and design a direct RF sampling receiver to receive multiband global navigation satellite system (GNSS) signals such as GPS L1, GLONASS G1 and G2 signals. The receiver is implemented with a commercial AD converter and software. The receiver performance is demonstrated by receiving the real signals.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.5
• /
• pp.462-469
• /
• 2014

As a basic study on both engineering applications and representation methods of synesthesia, this paper aims at building basic system which converts a muscular sense into both visual and auditory elements. As for the building method, data of the muscular sense can be acquired through roll and pitch signals which are calculated from both three-axis acceleration sensor and the two-axis gyro sensor. The roll and pitch signals are then converted into both visual and auditory information as outputs. The roll signals are converted into both intensity elements of the HSI color model and octaves as one of auditory elements. In addition, the pitch signals are converted into both hue elements of the HSI color model and scales as another one of auditory elements. Each of the extracted elements of the HSI color model is converted into each of the three elements of the RGB color model respectively, so that the real-time output color signals can be obtained. Octaves and scales are also converted and synthesized into MIDI signals, so that the real-time sound signals can be obtained as anther one of output signals. In experiments, the results revealed that normal color and sound output signals were successfully obtained from roll and pitch values that represent muscular senses or physical movements, depending on the conversion relationship based on the similarity between color and sound.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.8
• /
• pp.1222-1230
• /
• 2022

As the number of signal sources with low detectability by using various modulation techniques increases, research to classify signal modulation methods is steadily progressing. Recently, a Convolutional Neural Network (CNN) deep learning technique using FFT as a preprocessing process has been proposed to improve the performance of received signal classification in signal interference or noise environments. However, due to the characteristics of the FFT in which the window is fixed, it is not possible to accurately classify the change over time of the detection signal. Therefore, in this paper, we propose a CNN model that has high resolution in the time domain and frequency domain and uses wavelet transform as a preprocessing process that can express various types of signals simultaneously in time and frequency domains. It has been demonstrated that the proposed wavelet transform method through simulation shows superior performance regardless of the SNR change in terms of accuracy and learning speed compared to the FFT transform method, and shows a greater difference, especially when the SNR is low.

• Communications for Statistical Applications and Methods
• /
• v.3 no.2
• /
• pp.85-91
• /
• 1996

일본의 품질 전문가 다구찌가 제안한 파라미터 설계의 장단점에 대해 많은 논의가 있어왔다. 본 연구에서는 신호 대 잡음 비율에 초점을 두어, 이론적 쟁점이 무엇인지를 요약하고, 그 문제점을 모의실험을 통해 구체적으로 지적하였다. 한편 신호 대 잡음 비율에 대한 대안 중에서 Box의 변수변환 방법을 소개하였다.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.47 no.2
• /
• pp.35-42
• /
• 2010

In this study, crosstalk between dominant interconnect pairs in an A/D converter circuit is analyzed in frequency domain and effects of termination conditions on crosstalk are described, based on the practical circuit conditions. An A/D converter circuit is a mixed circuit where both clean and noisy signals coexist such that the circuit probably suffers from distortion by crosstalk. An analog input signal and the reference voltage signal, which dominate the overall conversion performance of the A/D converter circuit, are ready to be distorted by crosstalk and include specific termination conditions, such as non-matching and capacitive termination, respectively. Thus, this study presents the model of crosstalk considering impedance mismatch at both ends and analyzes effects of the practical termination conditions in the analog input and the reference voltage interconnects on crosstalk. A typical circuit configuration of the two interconnects is described and crosstalk including near-end and far-end termination impedances is modeled. Effects of the near-end impedance mismatch in the analog input interconnect and the far-end capacitive termination in the reference voltage interconnect are estimated in the frequency domain by using the model of crosstalk and experiments are performed to confirm the estimated results. Microstrip lines are used as interconnects, involving the increase of loss in high frequencies.

• Phonetics and Speech Sciences
• /
• v.12 no.3
• /
• pp.55-63
• /
• 2020

Voice conversion can be applied to various voice processing applications. It can also play an important role in data augmentation for speech recognition. The conventional method uses the architecture of voice conversion with speech synthesis, with Mel filter bank as the main parameter. Mel filter bank is well-suited for quick computation of neural networks but cannot be converted into a high-quality waveform without the aid of a vocoder. Further, it is not effective in terms of obtaining data for speech recognition. In this paper, we focus on performing voice-to-voice conversion using only the raw spectrum. We propose a deep learning model based on the transformer network, which quickly learns the voice conversion properties using an attention mechanism between source and target spectral components. The experiments were performed on TIDIGITS data, a series of numbers spoken by an English speaker. The conversion voices were evaluated for naturalness and similarity using mean opinion score (MOS) obtained from 30 participants. Our final results yielded 3.52±0.22 for naturalness and 3.89±0.19 for similarity.