Earticle

본 논문은 반향 제거 평균 예측 LMS 알고리즘을 이용하여 반향 잡음에 강인한 연속 음성 인식 모델인 CHMM 모델을 구성하는 방법을 제안하였다. 변화하는 반향 잡음에 적응하고 연속 음성 인식 성능 향상을 위한 반향 잡음 제거 평균 예측 LMS 알고리즘을 이용하여 CHMM 모델을 구성하였다. 제안한 알고리즘에 의해 구성된 CHMM 모델에 대하여 연속 인식 성능을 평가하였다. 실험 결과 변화하는 환경 잡음을 제거하여 얻은 음성의 SNR은 평균 1.93dB이 향상되었고 연속 음성의 인식률은 2.1% 향상되었다.

In this paper, the echo noise robust CHMM learning model using echo cancellation average estimator LMS algorithm is proposed. To be able to adapt to the changing echo noise. For improving the performance of a continuous speech recognition, CHMM models were constructed using echo noise cancellation average estimator LMS algorithm. As a results, SNR of speech obtained by removing Changing environment noise is improved as average 1.93dB, recognition rate improved as 2.1%.

Firstly the Decomposition Least Mean (D‐LMS) algorithm is analyzed deeply in theoretical, then a variable step size D‐LMS algorithm is proposed. By using the MATLAB based simulation model, the system identification results of the stable step size D‐LMS algorithm and the ones of the variable step‐size D‐LMS algorithm are both obtained. The simulation results show that compared with the stable step size D‐LMS algorithm, this variable step‐size D‐LMS algorithm based on wavelet transform has the better convergence ability and tracing ability, but the less state errors.

음성 인식 시스템은 다양하게 변화하는 환경 잡음에 빠르게 적응할 수 없어서 인식 성능을 저하시키는 요인이 된다. 본 논문에서는 평균 예측 LMS 알고리즘을 이용하여 반향 잡음에 강인하게 하는 방법으로 HMM 학습 모델을 구성하는 방법을 제안하였으며, 변화하는 반향 잡음에 적응하도록 HMM 학습 모델을 구성하여 인식 성능을 평가하였다. 실험 결과 변화하는 환경 잡음을 제거하여 얻은 음성의 SNR은 평균 3.1dB이 향상되었고 인식률은 3.9% 향상되었다.

The speech recognition system can not quickly adapt to varied environmental noise factors that degrade the performance of recognition. In this paper, the echo noise robust HMM learning model using average estimator LMS algorithm is proposed. To be able to adapt to the changing echo noise HMM learning model consists of the recognition performance is evaluated. As a results, SNR of speech obtained by removing Changing environment noise is improved as average 3.1dB, recognition rate improved as 3.9%.

In this study, a response model of a beam structure was established through finite element analysis by analyzing the vibration response to external excitation. The vibration control performance of the beam was then evaluated by applying the narrow-band Fx-LMS algorithm for active structural control. The transfer function was obtained at the error sensor location when the structure was excited and the three-axis actuator was operated. The performance of the active control was investigated with 18 channels for error input and actuator output. When the equipment is exciting, the response of the error sensor is the primary path, and when the inertial 3-axis actuator operates, the response of the error sensor position is the secondary path, and the Fx-LMS algorithm is applied. The simulation was performed by changing the control parameters so that the response of the error sensor can satisfy the target performance. From the results of this study, the acceleration results over time showed about 70% vibration reduction after active control, and the average error value of the error sensor also decreased by about 68%. In addition, it was confirmed that real-time control of a system with 18 sensors and 18 actuators is possible even if the secondary path is configured in two orders.

Beam forming is a technique in signal processing that has found widespread applications in fields such as radar, wireless communication, bio-medical ultrasounds and so on. Beam forming consist of an array of antennas whose radiation pattern are adjusted to a particular direction. This paper aims in implementing a two stage design for acoustic beam forming using MATLAB and Verilog HDL. First stage is a delay and sum beam former used to obtain beams from a particular direction while the second stage is an LMS model based on Least Mean Square (LMS) algorithm. This stage improves the efficiency of delay sum by removing the unwanted signals. Different hardware parameters like logic utilization, memory usage, CPU time and delay of this design are analyzed to determine its characteristics as compared to delay sum and LMS. This study found that the two stage model gives better beam forming and noise removal than when the stages are implemented independently.

Removal of noises from real-time speech signal is a typical problem. The signal interference initiated by background noise is a major problem in voice communication systems. Adaptive Filtering methods have emerged as an important technology for communication systems. This technique has been employed to improve the quality of the speech signal by cancelling the undesirable phenomenon such as acoustic noise. In this paper, for the removal of additional noise from speech signal an adaptive filter has been designed using LMS, NLMS, SLMS and VSS-LMS algorithms. This paper presents the instigation of Least Mean Square algorithm (LMS), Normalized Least Mean Square algorithm (NLMS), Sign Least mean square algorithm (SLMS) and Variable step size (VSS) algorithm on an infinite impulse response (IIR) filter using adaptive filter toolkit of LabVIEW software. User interface is designed using LabVIEW to obtain the learning curves for these adaptive algorithms. The final results show the comparison of the performance of the entire proposed algorithms with each other. The complete performance of the designed system in terms of stability and convergence rate has been observed.

최근 정보화 및 컴퓨터 비전 기술의 발전과 함께 객체의 인식 및 추적 기능을 가진 CCTV시스템이 다양한 분야에서 연구되고 있다. 하지만 실외환경에서 발생할 수 있는 그림자의 변화, 조명의 변화, 움직이는 요소들과 같은 배경의 변화는 객체 인지성능에 영향을 주게 된다. 따라서 실외환경에서 배경의 변화를 실시간으로 갱신하기 위해 본 논문에서는 다양한 배경 모델링 기법들을 분석하고, 가중치를 기반으로 한 배경 갱신 알고리즘을 제안한다. 실험을 통해 제안한 알고리즘의 객체 검출 성능은 이전 연구의 객체 검출 성능을 유지하며, 오인식 된 객체 수가 이전 연구에 비해 감소됨을 확인하였다.

Recently, with the rapid advancement in information and computer vision technology, a CCTV system using object recognition and tracking has been studied in a variety of fields. However, it is difficult to recognize a precise object outdoors due to varying pixel values by moving background elements such as shadows, lighting change, and moving elements of the scene. In order to adapt the background outdoors, this paper presents to analyze a variety of background models and proposed background update algorithms based on the weight factor. The experimental results show that the accuracy of object detection is maintained, and the number of misrecognized objects are reduced compared to previous study by using the proposed algorithm.

IN ADAPTIVE LINEAR PREDICTION, AN ADAPTIVE CAPABILITY IS BUILT INTO THE PROCESSOR SUCH THAT AS THE IMAGE STATISTICS CHANGE, THE PREDICTION FILTER COEFFICIENTS THEMSELVES CHANGE, PRODUCING A NEW FILTER MORE CLOSELY OPTIMIZED TO THE NEW SET OF IMAGES STATISTICS. THE LMS ALGORITHM MAY BE USED TO ADAPT THE COEFFICIENT OF AN ADAPTIVE PREDICTION FILTER FOR IMAGE SOURCE ENCODING. IN THIS PAPER, TWO CODING SYSTEMS USING DPCM AND LMS ALGORITHMS RESPECTIVELY FOR OBTAINING THE FIRST TRANSFORMED COEFFICIENT IN HYBRID CODING ARE COMPARED.

Filtered-x LMS algorithm maybe the most popular control algorithm used in DSP implementations of active noise and vibration control system. The algorithm converges on a timescale comparable to the response time of the system to be controlled, and is found to be very robust. If the pure tone reference signal is synchronously sampled, it is found that the behavior of the adaptive system can be completely described by a matrix of linear, time invariant, transfer functions. This is used to explain the behavior observed in simulations of a simplified single input, single output adaptive system, which retains many of the properties of the multichannel algorithm.

고정 소수점 처리기만큼 낮은 복잡도와 비용으로 넓은 동작 영역의 데이터 처리가 가능한 블록 부동 소수점 체제에서 블록 LMS 알고리즘의 구현을 위한 기법을 제시하였다. 제안 기법은 필터 계수 및 데이터의 표현을 위한 적절한 포맷을 적용하였다. 또한, 시변 mantissa와 시변 exponent를 갖는 스텝 크기에 대해 scaled 표현을 적용하였다. Scaled 표현과 새로운 상한을 이용하여, 필터 계수의 무게 mantissa와 exponent에 대한 업데이트 관계를 개발하였으며, 오버 플로우가 발생하지 않도록 할 뿐만 아니라 이미 직접 곱해진 미소량도 고려하였다. 또한 필터 계수의 mantissa와 필터 출력 역시 고속 블록 LMS 알고리즘 기법의 적절한 수정에 의해 더욱 빠르게 평가할 수 있는 방법을 보였다.

A scheme is proposed for implementing the block LMS algorithm in a block floating point framework that permits processing of data over a wide dynamic range at a processor complexity and coat as low as that of a fixed point processor. The proposed scheme adopts appropriate formats for representing the filter coefficients and the data. Using these and a new upper bound on the step size, update relations for the filter weight mantissas and exponent are developed, taking care so that neither overflow occurs, nor are quantifies which are already very small multiplied directly. It is further shown how the mantissas of the filter coefficients and also the filter output can be evaluated faster by suitably modifying the approach of the fast block LMS algorithm

It is well-known that the convergence rate gets worse when an input signal to an adaptive filter is correlated. In this paper we propose a new adaptive filtering algorithm that makes the convergence rate highly improved even for highly correlated input signals. By introducing an orthogonal constraint between successive input signal vectors, we overcome the slow convergence problem caused by the correlated input signal. Simulation results show that the proposed algorithm yields highly improved convergence speed and excellent tracking capability under both time-invariant and time varying environments, while keeping both computation and implementation simple.

In this paper, convergence properties of the LMS, LMF, and LVCMS algorithms are investigated under the assumption of the uncorrelated Gaussian input data. By treating these algorithms as special cases of more general algorithm family, unified results on these algorithms are obtained. First the upper bound on the step size parameter is obtained. Second, an expression for misadjustment is obtained. These theoretical results confirm earlier LMS works. Further, the results explain why the LMS and LVCMS algorithms are experiencing difficulties with plant noise having heavier tailed densities. Simulation results agree with theoretical expectation closely for various plant noise statistics.

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.

It is well-blown that the convergence rate gets worse when an input signal to an adaptive filter is correlated. In this paper we propose a new adaptive filtering algorithm that makes the convergence rate much improved even for highly correlated input signals. By introducing an orthogonal constraint between successive input signal vectors we overcome the slow convergence problem of the LMS algorithm with the correlated input signal. Simulation results show that the proposed algerian yields fast convergence speed and excellent tracking capability under both time-invariant and time-varying environments, while keeping both computation and implementation simple.

The conventional normalized least mean square (NLMS) algorithm is the most widely used for adaptive identification within a non-stationary input context. The convergence of the NLMS algorithm is independent of environmental changes. However, its steady state performance is impaired during input sequences with low dynamics. In this paper, we propose a new NLMS algorithm which is, in the steady state, insensitive to the time variations of the input dynamics. The square soot (SR)-NLMS algorithm is based on a normalization of the LMS adaptive filter input by the Euclidean norm of the tap-input. The tap-input power of the SR-NLMS adaptive filter is then equal to one even during sequences with low dynamics. Therefore, the amplification of the observation noise power by the tap-input power is cancelled in the misadjustment time evolution. The harmful effect of the low dynamics input sequences, on the steady state performance of the LMS adaptive filter are then reduced. In addition, the square root normalized input is more stationary than the base input. Therefore, the robustness of LMS adaptive filter with respect to the input non stationarity is enhanced. A performance analysis of the first- and the second-order statistic behavior of the proposed SR-NLMS adaptive filter is carried out. In particular, an analytical expression of the step size ensuring stability and mean convergence is derived. In addition, the results of an experimental study demonstrating the good performance of the SR-NLMS algorithm are given. A comparison of these results with those obtained from a standard NLMS algorithm, is performed. It is shown that, within a non-stationary input context, the SR-NLMS algorithm exhibits better performance than the NLMS algorithm.

Variable step size LMS(VS-LMS) algorithms improve performance of LMS algorithm by means of varying the step size. This paper presents a new VS-LMS algorithm using normalized absolute estimation error. Normalizing the estimation error to the expected valus of the desired signal, we determined the step size using the relative size of estimation error, Because parameters and computational load are less, our algorithm is easy to implement in hardware. The performance of the proposed algorithm is analyzed theoretically and estimated through simulations. Based on the theoretical analysis and computer simulations, the proposed algorithm is shown to be effective compared to conventional VS-LMS algorithms.

This paper deals with the application of the LMS algorithm as a decision-directed adaptive equalizer in a communication receiver which also employs a sophisticated decoding scheme such as the Viterbi algorithm, in which the desired signal, hence the error, is not available until several symbol intervals later because of decoding delay. In such applications the implemented weight updating algorithm becomes DLMS and major penalty is reduced convergence speed. Therefore, every effort should by made to keep the delay as small as possible if it is not avoidable. In this paper we present a modified implementation in which the effects of the decoding delay can be avioded and perform some computer simulations to check the validity and the performance of the new implementation.

Application of the Filtered-X LMS adaptive filter to active noise cancellation requires to estimate the transfer characteristics between the output and the error signal of the adaptive canceler. In this paper, we derive an adaptive cancellation algorithm and analyze is convergence behavior when the acoustic noise is assumed to consist of multiple sinusoids. The results of the convergence analysis of the Filtered-X LMS algorithm indicate that the effects of parameter estimation inaccuracy on the convergence behavior of the algorithm are characterize by two distinct components : Phase estimation error and estimated magnitude. In particular, the convergence of the Filtered-X LMS algorithm is show to be strongly affected by the accuracy of the phase response estimate. Simulation results of the algorithm are presented which support the theoretical convergence analysis.

The adaptive array antenna with the LMS algorithm has the advantage in that it can perform train't because of its slower convergencerate. In this paper, the \ulcornerLMS algorithm is applied to the adaptive array so that the convergence rate can be improved, and the performance of he adaptive array by the \ulcornerLMS algorithm is compared to, that of the LMS adaptive array. It is shown that the adaptive array by the \ulcornerLMS algorithm is superior to the LMS adaptive array in the narrow frequency band.

In this paper, enhancement of speech corrupted by additive white or colored noise is stuided. The nuconstrained frequency-domain block least-mean-square (UFBLMS) adaptation algorithm and its frequency-weighted version are newly applied to speech enhancement. For enhancement of speech degraded by white noise, the performance of the UFBLMS algorithm is superior to the spectral subtraction method or Wiener filtering technique by more than 3 dB in segmented frequency-weighted signal-to-noise ratio(FWSNERSEG) when SNR of speech is in the range of 0 to 10 dB. As for enhancement of noisy speech corrupted by colored noise, the UFBLMS algorithm is superior to that of the spectral subtraction method by about 3 to 5 dB in FWSNRSEG. Also, it yields better performance by about 2 dB in FWSNR and FWSNRSEG than that of time-domain least-mean-square (TLMS) adaptive prediction filter(APF). In view of the computational complexity and performance improvement in speech quality and intelligibility, the frequency-weighted UFBLMS algorithm appears to yield the best performance among various algorithms in enhancing noisy speech corrupted by white or colored noise.