• 대한전기학회논문지:시스템및제어부문D
• /
• 제51권2호
• /
• pp.69-77
• /
• 2002

Speaker verification can be classified in two categories, text-dependent speaker verification and text-independent speaker verification. In this paper, we discuss text-dependent speaker verification. Text-dependent speaker verification system determines whether the sound characteristics of the speaker are equal to those of the specific person or not. In this paper we obtain the speaker data using a sound card in various noisy conditions, apply a new Hidden LMS (Least Mean Square) adaptive algorithm to it, and extract LPC (Linear Predictive Coding)-cepstrum coefficients as feature vectors. Finally, we use a competitive learning neural network for speaker verification. The proposed hidden LMS adaptive filter using a neural network reduces noise and enhances features in various noisy conditions. We construct a separate neural network for each speaker, which makes it unnecessary to train the whole network for a new added speaker and makes the system expansion easy. We experimentally prove that the proposed method improves the speaker verification performance.

• Journal of electromagnetic engineering and science
• /
• 제4권1호
• /
• pp.1-7
• /
• 2004

In this paper the Individual tap - Least Mean Square(IT-LMS) algorithm is applied to the adaptive multipath channel equalization using hybrid-type Genetic Algorithm(GA) for achieving lower minimum Mean Squared Error(MSE). Owing to the global search performance of GA, LMS-type equalizers combined with it have shown preferable performance in both global and local search but those still have unsatisfying minimum MSE performance. In order to lower the minimum MSE we investigated excess MSE of IT-LMS algorithm and applied it to the hybrid GA equalizer. The high convergence rate and lower minimum MSE of the proposed system give us reason to expect that it will perform well in practical multi-path channel equalization systems.

• 한국소음진동공학회:학술대회논문집
• /
• 한국소음진동공학회 2009년도 추계학술대회 논문집
• /
• pp.231-239
• /
• 2009

The filtered-x LMS (FX-LMS) algorithm has been applied to the active noise control (ANC) system in an acoustic duct. This algorithm is designed based on the FIR (finite impulse response) filter, but it has a slow convergence problem because of a large number of zero coefficients. In order to improve the convergence performance, the step size of the LMS algorithm was modified from fixed to variable. However, this algorithm is still not suitable for the ANC system of a short acoustic duct since the reference signal is affected by the backward acoustic wave propagated from a secondary source. Therefore, the recursive filteredu LMS algorithm (FU-LMS) based on infinite impulse response (IIR) is developed by considering the backward acoustic propagation. This algorithm, unfortunately, generally has a stability problem. The stability problem was improved by using an error smoothing filter. In this paper, the recursive LMS algorithm with variable step size and smoothing error filter is designed. This recursive LMS algorithm, called FU-VSSLMS algorithm, uses an IIR filter. With fast convergence and good stability, this algorithm is suitable for the ANC system in a short acoustic duct such as the intake system of an automotive. This algorithm is applied to the ANC system of a short acoustic duct. The disturbance signals used as primary noise source are a sinusoidal signal embedded in white noise and the chirp signal of which the instantaneous frequency is variable. Test results demonstrate that the FU-VSSLMS algorithm has superior convergence performance to the FX-LMS algorithm and FX-LMS algorithm. It is successfully applied to the ANC system in a short duct.

• 전자공학회논문지 IE
• /
• 제45권1호
• /
• pp.50-57
• /
• 2008

E-Learning 시스템에서 학습자의 학습 욕구와 효과를 높이기 위하여 다양한 학습 콘텐츠를 적용하고 있다. 이러한 학습 콘텐츠로는 텍스트, 동영상, 소리, 그림 등을 들 수 있다. 그러나 파일크기가 큰 멀티미디어 학습 콘텐츠는 많은 전송 서비스 시간을 필요로 한다. 본 논문에서는 LCMS에서 관리 및 처리되는 멀티미디어 학습 콘텐츠를 보다 빠르고 효율적으로 서비스하기 위한 LMS의 스케줄링 기법을 제안하고자 한다. 이를 위하여 LMS에 스케줄러와 메시지 큐를 두었으며, 학습이 진행되는 동안 학습 콘텐츠 요구에 대한 결과정보를 LMS에 저장하였다. 학습자의 학습 콘텐츠 요구가 있을 경우 LCMS에 접속하지 않고 LMS에 저장된 학습 콘텐츠 정보를 활용함으로서 보다 빠르고 효율적인 학습 콘텐츠 지원이 가능하도록 하였다. 본 기법의 적용결과로서 학습 초기에는 기존의 기법에 비하여 학습 콘텐츠 서비스가 늦게 나타났으나 학습이 진행될수록 보다 빠른 서비스가 가능하였다.

• 디지털콘텐츠학회 논문지
• /
• 제5권4호
• /
• pp.295-299
• /
• 2004

본 논문에서는 IEEE802.15.3(HDR-WPAN) 시스템에서 LMS 알고리즘과 RLS 알고리즘의 성능을 비교 분석하였다. LMS 알고리즘은 구현하기가 쉽고 계산량이 적은 장점이 있는 반면, 수렴 속도가 느리며, RLS 알고리즘은 계산량이 많으나, 수렴속도가 빠른 장점을 가지고 있다. HDR-WPAM 시스템을 기반으로 같은 환경 하에서 LMS 알고리즘을 사용했을 경우 250 샘플 이후에 채널에 적응된 등화가 이루어졌고 RLS 알고리즘을 사용했을 경우 50 샘플 이후에 등화가 이루어 졌다. 이를 통해, HDR-WPAN 시스템에서 보다 안정적이며, 빠른 등화 처리를 위해서는 LMS 알고리즘보다 RLS 알고리즘을 통한 적응 등화 구현이 효과적임을 시뮬레이션을 통해 확인하였다.

• 한국통신학회논문지
• /
• 제18권2호
• /
• pp.178-183
• /
• 1993

본 논문에서는 LMS 적응 알고리즘이 수렴하기 위한 스텝크기의 적정 범위를 등화기 계수의 양자화 오차와 초과 MSE를 고려하여 새로 제시하였으며 이의 타당상을 트랜스버설 등화기의 시뮬레이션을 통해 보았다.

• 소음진동
• /
• 제4권1호
• /
• pp.59-69
• /
• 1994

A new LMS algorithm titled constrained LMS' is proposed for problems with constrained structure. The conventional LMS algorithm can not be used because it destroys the constrained structures of the weights or parameters. Proposed method uses error-back propagation, which is popular in training neural networks, for error minimization. The illustrative examplesare shown to demonstrate the applicability of the proposed algorithm.

• 소음진동
• /
• 제8권3호
• /
• pp.485-493
• /
• 1998

Many of the adaptive noise control systems utilize a form of the least mean square (LMS) algorithms. In the active control of noise, it is common practice to locate an error microphone far from the control source to avoid the near-field effects by evanescent waves. Such a distance between the control source and the error microphone makes a certain level of time-delay inevitable and, hence, may yield undesirable effects on the convergence properties of control algorithms such as filtered-x LMS. This paper discusses the dependence of the convergence rate on the acoustic error path in these popularalgorithms and introduces new algorithms which increase the convergence region regardless of the time-delay in the acoustic error path. Performances of the new LMS algorithms are presented in comparison with those by the conventional algorithms based on computer simulations and experiments.

• 대한전자공학회논문지SP
• /
• 제45권2호
• /
• pp.90-94
• /
• 2008

적응 필터의 입력 신호의 상관도 (correlation)가 클 경우 LMS 알고리즘의 수렴 속도는 상당히 느려지게 된다. 본 논문에서는 입력 신호의 상관도가 높은 상황에서 수렴 속도를 향상시킬 수 있는 적응 필터링 알고리즘을 제안한다. 입력 신호에 대하여 직교성을 가지도록 변환을 인위적으로 가하여 LMS 알고리즘의 한계를 극복한다. 제안한 알고리즘의 성능 향상은 시스템식별 모델을 통하여 그 수렴 속도의 개선을 확인하며 또한 시변 환경 하에서 적응 필터의 시변 추적 능력을 통해 보여 진다.

• 전기학회논문지
• /
• 제60권1호
• /
• pp.150-155
• /
• 2011

In an active noise control systems, an IIR filter may cause a problem in stability beacause of its poles. For IIR filter, its poles goes sometimes out of a unit circle in a z-plane in the transition state, where the adaptive algorithm converges to the optimum value, which causes the system to diverge. Fuzzy LMS algorithm has a better convergence property than conventional LMS algorithms, but is not applicable to IIR filter because of the reasons. Stabilized adaptive algorithm could be improves stability by moving the pole of IIR filer toward the origin forcibly in the transient state, and by introducing forgetting factor to maintain the optimum convergence when it reaches to the steady state. In this paper, We proposed stabilized adaptive fuzzy LMS algorithms with IIR filter structures, for single channel active noise control with ill conditioned signal case. Computer simulations were performed to show the effectiveness of a proposed algorithm.