컨볼루션 혼합신호의 암묵 잡음분리방법

이행우 한국전자통신학회 2022 한국전자통신학회 논문지 Vol.17 No.3

본 논문은 시간지연 컨볼루션 혼합신호의 암묵잡음분리방법에 관한 것이다. 폐쇄된 공간에서 음향신호의 혼합모델은 다채널이기 때문에 convolutive 암묵신호분리방법을 적용하며 두 마이크 입력신호의 시간지연된 데이터 메모리를 사용한다. 신호분리를 위해 분리계수를 직접 계산하지 않고 역방향 모델을 이용하여 혼합계수를 산출하며 계수갱신은 2차 통계적 성질을 기반으로 반복적인 계산에 의해 수행됨으로써 음성신호를 추정해낸다. 제안한 암묵신호분리의 성능을 검증하기 위해 많은 시뮬레이션을 수행하였다. 모의실험 결과, 이 방법을 사용한 잡음분리는 컨볼루션혼합에 상관없이 안전하게 동작하고, 일반적인 적응 FIR(Finite Impulse Response) 필터구조에 비해 PESQ(Perceptual Evaluation of Speech Quality)가 0.3점 개선되는 것으로 나타났다.

음향반향제거기의 구현을 위한 개선된 직교투사법

이행우 대한전자공학회 2006 電子工學會論文誌-SP (Signal processing) Vol.43 No.2

This paper proposes the improved orthogonal projection method as a new technique advancing the performance of the acoustic echo canceller. Comparing with the widely used NLMS adaptive algorithm which is simple and stable, it shows that this method has the improvement of the convergence speed for signals with the large auto-correlation, and has small computational quantities. In order to testify performances of the orthogonal projection method whom this paper proposes, we have coded a simulation program and executed computer simulations. We observed convergence curves by using two adaptive algorithm for noises and speeches. From simulation results for two input signals, the proposed method shows the high ERLE and the fast convergence and the stable operation in case of using speeches as well as noises. 본 논문은 음향반향제거기의 성능을 향상시키기 위한 새로운 기술로서 개선된 직교투사법을 제안한다. 간단하고 안정되어 널리 사용되고 있는 기존의 NLMS 알고리즘과 비교하여, 이 기술은 음성신호와 같이 상호상관도가 높은 신호에 대해 수렴속도가 증가하고 연산량이 감소한다. 제안하는 직교투사법의 성능을 시험하기 위하여 프로그램을 코딩하고 시뮬레이션을 수행하였다. 두 알고리즘을 사용하여 음성과 잡음에 대한 수렴특성을 관찰하였다. 실험 결과, 제안한 방법은 잡음뿐만 아니라 음성에 대해서도 높은 ERLE와 수렴속도 및 수렴의 안정성 등을 보여 주었다.

음향반향제거기에서 암묵신호분리를 이용한 동시통화처리

이행우,Lee, Haengwoo 디지털산업정보학회 2016 디지털산업정보학회논문지 Vol.12 No.1

This paper is on an acoustic echo canceller solving the double-talk problem by using the blind signal separation technology. The acoustic echo canceller may be deteriorated or diverged during the double-talk period. So we use the blind signal separation to detect the double talking by separating the near-end speech signal from the mixed microphone signal. The blind signal separation extracts the near-end signal from dual microphones by the iterative computations using the 2nd order statistical character in the closed reverberation environment. By this method, the acoustic echo canceller operates irrespective of the double-talking. We verified performances of the proposed acoustic echo canceller in the computer simulations. The results show that the acoustic echo canceller with this algorithm detects the double-talk periods well, and then operates stably without diverging of the coefficients after ending the double-talking. The merits are in the simplicity and stability.

성토층 하부의 자갈층 유·무에 따른 침투특성 변화

이행우,장병욱,장웅희,김시운,Lee Haeng Woo,Chang Pyoung Wuck,Chang Woong Hee,Kim See-Woon 한국농공학회 2005 한국농공학회논문집 Vol.47 No.4

A series of laboratory tests was carried out fur analyzing of seepage characteristics of two-layers embankment model which consists of gravel and earth fill layers. Gravel layers were built under the earth fill for a half and one-third width of earth fill of the model. Permeability of earth fill was ranged between $5.00\times10^{-5}\~3.00\times10^{-4}\;m/s$.. The tests were performed with hydraulic gradients(i), $0.10\~0.55$. From the test results, hydraulic head of earth fill with gravel layer was 1.6 times higher than that of earth fill without gravel layer. Seepage rate was increased up to $4\~22$ times and safety factor for piping was decreased to $13\~43\;\%$ comparing the earth fill with gravel layer to that without gravel layer. The gravel layer under the earth fill could, in general, give some serious seepage problems to seadike embankment.

연직배수재(VDs)에 의한 오염지반정화 메커니즘 연구

이행우,장병욱,강병윤,김현태,Lee Haeng Woo,Chang Pyoung Wuck,Kang Byung Yoon,Kim Hyun Tae 한국농공학회 2005 한국농공학회논문집 Vol.47 No.5

In-situ soil remediation mechanism through the vertical drains (VDs) is analyzed with numerical model as the error and complementary error function. Results from in-situ test and analysis indicate that the contaminant concentration ratio as initial one ( C/$C_0$) increases as the radius ratio ( r/R) increases from the injection well, and also increases as the depth ratio ( z/ H) increases from the top of contaminated area. The elapse time needed to attain $50\%$ and $90\%$ clean up level ($ t_{50},\;t_{90}$) increases as the radius ratio ( r/R) and the depth ratio ( z/ H) increase. As above results, the procedure of soil flushing in contaminated area using vertical drains makes progress from the top of injection well to the bottom of extraction well.

비상관기를 이용한 디지털 보청기용 음향궤환제거기

이행우,Lee, Haeng-Woo 한국정보통신학회 2008 한국정보통신학회논문지 Vol.12 No.5

최근 고령화 사회로 접어들면서 보청기의 수요가 점차 증대되고 있으며, 이에 따라 우수한 성능의 음향궤환제거기가 요구되고 있다. 본 논문에서는 음향궤환제거기의 수렴 특성을 향상시키기 위한 새로운 적응 알고리즘을 제안하였다. 이 알고리즘은 음성신호와 같이 상관도가 높은 신호를 사용하여 적응필터의 계수를 갱신할 때 신호의 상관도를 감소시켜주기 위해 비상관기를 사용함으로서 수렴성능을 향상시킨다. 본 연구에서 제한 음향궤환제거기의 성능을 검증하기 위하여 시뮬레이션 프로그램을 작성하고 시뮬레이션을 수행하였다. 입력신호의 종류와 적응 알고리즘에 따라서 궤환제거기의 수렴특성을 관찰하였다. 시뮬레이션 결과, 제안한 궤환제거기는 유색신호를 입력으로 사용하는 경우, 약 5-10 dB의 SNR 성능이 개선되는 것을 보여 주었다. This paper is on a new adaptive algorithm which can cancel the acoustic feedback signals in the digital hearing aids. The proposed algorithm uses the normalized LMS algorithm with decorrelators. By doing so, it can be reduced the autocorrelation for the voice signals. To analyze the convergence characteristics of the proposed algorithm, the simulations were carried out about various input signals. And we had compared the performances of convergence for this algorithm with the ones for the NLMS algorithm. As the results of simulations, it is proved that the feedback canceller adopting this algorithm shows about 5-10 dB more high SNR than the NLMS algorithm for the colored inputs.

Hough 변환을 이용한 암묵신호분리방법

이행우 (사)디지털산업정보학회 2014 디지털산업정보학회논문지 Vol.10 No.3

This paper is on the blind signal separation(BSS) method by the geometric method. To separate the signal sources, we use Hough transform and BSS. Hough transform is a geometric method which let us know the local informations of the signal. We find the orientations of signals by Hough transform and know the number of signal sources. When the number of sensors is more than the number of sources. the BSS algorithm can separate the mixtures well in the time domain. This algorithm has a good performance in converging fast. We had checked up the quality of the algorithm after separating the mixed signals. The results of simulations show that this BSS method has the abnormal waveforms due to unconverging coefficients in the beginning, and stably has the separated waveforms which almost equal to the sources in the most period.

S-LMS 알고리즘을 이용한 음향반향제거기의 FPGA구현

이행우 대한전자공학회 2004 電子工學會論文誌-SD (Semiconductor and devices) Vol.41 No.9

본 논문은 휴대폰의 자동차 핸즈프리를 위한 음향반향제거기의 설계 및 구현에 관한 것이다. 이를 구현하기 위해 효율적인 적응 알고리즘의 개발과 VHDL 회로설계, 그리고 테스트 보드 제작 등을 수행하였다. 적응 알고리즘은 간단하고 안정적이며 수렴속도가 빠른 새로운 S-LMS 알고리즘을 사용하였다 음향반향제거기는 15-bit, 128-tap 회로구조로서 50,000 게이트급 FPGA에 구현하였다. 그리고 회로동작을 검증하기 위하여 보드를 제작하였는데, 2개의 CODEC과 제어입력용 DIP 스위치 및 출력표시용 LED 등이 포함되어 있다. 백색 가우스잡음을 사용하여 시험한 결과, 구현된 회로의 ERLE는 최대 30dB의 반향제거 성능을 나타내는 것으로 확인되었다. This paper describes a new adaptive algorithm which can reduce the required computation quantities in the adaptive filter. The proposed S-LMS algorithm uses only the signs of the normalized input signal rather than the input signals when coefficients of the filter are adapted. By doing so, there is no need for the multiplications and divisions which are mostly responsible for the computation quantities. To analyze the convergence characteristics of the proposed algorithm, the condition and speed of the convergence are derived mathematically. Also, we simulate an echo canceller adopting this algorithm and compare the performances of convergence for this algorithm with the ones for the other algorithm. As the results of simulations, it is proved that the echo canceller adopting this algorithm shows almost the same performances of convergence as the echo canceller adopting the SIA algorithm.

이산 Wavelet 변환을 이용한 딥러닝 기반 잡음제거기

이행우 한국전자통신학회 2023 한국전자통신학회 논문지 Vol.18 No.6

본 논문에서는 음향신호의 배경잡음을 감쇠하기 위한 새로운 알고리즘을 제안한다. 이 알고리즘은 이산 웨이블릿 변환(DWT: Discrete Wavelet Transform) 후 기존의 적응필터를 대신 FNN(Full-connected Neural Network) 심층학습 알고리즘을 이용하여 잡음감쇠 성능을 개선하였다. 입력신호를 단시간 구간별로 웨이블릿 변환한 다음 1024-1024-512-neuron FNN 딥러닝 모델을 이용하여 잡음이 포함된 단일입력 음성신호로부터 잡음을 제거한다. 이는 시간영역 음성신호를 잡음특성이 잘 표현되도록 시간-주파수영역으로 변환하고 변환 파라미터에 대해 순수 음성신호의 변환 파라미터를 이용한 지도학습을 통하여 잡음환경에서 효과적으로 음성을 예측한다. 본 연구에서 제안한 잡음감쇠시스템의 성능을 검증하기 위하여 Tensorflow와 Keras 라이브러리를 사용한 시뮬레이션 프로그램을 작성하고 모의실험을 수행하였다. 실험 결과, 제안한 심층학습 알고리즘을 사용하면 기존의 적응필터를 사용하는 경우보다 30%, STFT(Short-Time Fourier Transform) 변환을 사용하는 경우보다는 20%의 평균자승오차(MSE: Mean Square Error) 개선효과를 얻을 수 있었다. In this paper, we propose a new algorithm for attenuating the background noises in acoustic signal. This algorithm improves the noise attenuation performance by using the FNN (Full-connected Neural Network) deep learning algorithm instead of the existing adaptive filter after wavelet transform. After wavelet transforming the input signal for each short-time period, noise is removed from a single input audio signal containing noise by using a 1024-1024-512-neuron FNN deep learning model. This transforms the time-domain voice signal into the time-frequency domain so that the noise characteristics are well expressed, and effectively predicts voice in a noisy environment through supervised learning using the conversion parameter of the pure voice signal for the conversion parameter. In order to verify the performance of the noise reduction system proposed in this study, a simulation program using Tensorflow and Keras libraries was written and a simulation was performed. As a result of the experiment, the proposed deep learning algorithm improved Mean Square Error (MSE) by 30% compared to the case of using the existing adaptive filter and by 20% compared to the case of using the STFT (Short-Time Fourier Transform) transform. effect was obtained.

성토제 하부에 매설된 사석층의 침투특성

이행우,장병욱 한국지반공학회 2005 한국지반공학회논문집 Vol.21 No.8

Rocks are dumped to soft marine ground in order to improve trafficability and construction conditions in the tideland reclamation construction sites. Though this rock layer under earth fill has caused in a serious seepage problems after construction, seepage behaviors of this embankment structure is not correctly investigated. Water flow through rock layers is, in general, known as Non-Darcy’s flow. However, the embedded rock layer under earth fill is not known whether its flow is governed by Darcy’s or Non-Darcy’s law. Therefore, a numerical analysis, laboratory model test and filed investigations were performed for analyzing the those seepage characteristics in this research. Results show that there is significance of 95% of confidence between observed heads and seepage rates, and the calculated ones by SAMTLE which is developed under the assumption that the water flows through the two-layer system obey the Darcy’s flow. And after operating the hydraulic gradient(i) of 0.10~0.55 upon laboratory model, these seepage characteristics of the embedded rock layer show that Reynolds Numbers are less than 10 and the relationship between these velocities of rock layer(v) and hydraulic gradients(i) is linearly proportional with more than 0.79 of the coefficient of correlation (R2). And the Reynolds Number of the velocity calculated by the relation of in the embedded rock layer of ○○ sea dike is 1~6. It shows also laminar flow. Based on these results, it is concluded that the seepage characteristics of embedded rock layer under earth fill can be laminar and Darcy’s flow.