【摘要】：為了應(yīng)對日益嚴重的噪聲污染問題,有源噪聲控制技術(shù)得到日益廣泛的應(yīng)用。對常用的前饋有源噪聲控制系統(tǒng)而言,自適應(yīng)算法的性能好壞直接決定了控制器的降噪性能。與時域自適應(yīng)算法相比,頻域自適應(yīng)算法有較快的收斂速度和較低的運算量,這使得其很適合在實時有源噪聲控制系統(tǒng)中使用。但在有源噪聲控制系統(tǒng)中,由于參考通道和次級通道的存在,系統(tǒng)往往會面臨濾波器階數(shù)不足和非因果的問題。本文重點探索頻域自適應(yīng)算法在有源噪聲控制系統(tǒng)中的性能。首先,本文介紹了塊LMS(BLMS)算法、頻域LMS(FDLMS)算法和LMS算法誤差與維納誤差的關(guān)系。FDLMS算法是BLMS算法的快速實現(xiàn)方式,其收斂速度可以通過使用歸一化各頻點功率得到的收斂因子來獲得改善。LMS算法誤差為維納誤差作為輸入信號的線性時不變系統(tǒng)的輸出,系統(tǒng)函數(shù)主要依賴于參考信號功率譜估計。對于某些信號,可能存在LMS算法誤差小于維納誤差的情況。其次,針對有源噪聲控制系統(tǒng)中遇到的非因果問題,分析了歸一化頻域算法(NFDLMS)在非因果條件下穩(wěn)態(tài)解的特性,發(fā)現(xiàn)NFDLMS算法的收斂結(jié)果與參考信號功率譜估計的平坦程度有關(guān)。參考信號功率譜平坦度越高,穩(wěn)態(tài)解越接近維納解。與之相對應(yīng)的一個值得關(guān)注的結(jié)論是：在非因果條件下,濾波器階數(shù)越高,NFDLMS算法的收斂結(jié)果可能越差。仿真和實驗驗證了理論分析的有效性。最后,針對非因果條件和濾波器階數(shù)不足時NFDLMS算法收斂不到維納解的問題,提出一種改進歸一化頻域算法(MNFDLMS).理論證明MNFDLMS算法在非因果條件和濾波器階數(shù)不足時也能收斂到維納解。仿真和實驗同樣驗證了理論分析的有效性。
[Abstract]:In order to deal with the increasingly serious problem of noise pollution, active noise control technology has been increasingly widely used. For the commonly used feedforward active noise control systems, the performance of the adaptive algorithm directly determines the noise reduction performance of the controller. Compared with the time-domain adaptive algorithm, the frequency-domain adaptive algorithm has higher convergence speed and lower computational complexity, which makes it suitable for use in real-time active noise control systems. However, in active noise control systems, due to the existence of reference channels and secondary channels, the system often faces the problems of insufficient filter order and non-causality. This paper focuses on the performance of frequency domain adaptive algorithm in active noise control system. First of all, this paper introduces the relationship between block LMS (BLMS) algorithm, frequency-domain LMS (FDLMS) algorithm and LMS algorithm error and Wiener error. FDLMS algorithm is the fast implementation of BLMS algorithm. The convergence rate can be improved by using the convergence factor of normalized power at each frequency point. The error of LMS algorithm is the output of linear time-invariant system with Wiener error as input signal. The system function mainly depends on the power spectrum estimation of the reference signal. For some signals, the LMS algorithm error may be less than Wiener error. Secondly, aiming at the non-causality problem in active noise control system, the characteristics of steady-state solution of normalized frequency-domain algorithm (NFDLMS) under non-causality condition are analyzed. It is found that the convergence of NFDLMS algorithm is related to the flatness of power spectrum estimation of reference signal. The higher the power spectrum flatness of the reference signal, the closer the steady-state solution is to the Wiener solution. The corresponding conclusion is that the higher the filter order, the worse the convergence result of NFDLMS algorithm under the condition of non-causality. The effectiveness of the theoretical analysis is verified by simulation and experiments. Finally, an improved normalized frequency domain algorithm (MNFDLMS).) is proposed to solve the problem that the NFDLMS algorithm can not converge to the Wiener solution when the non-causality condition and the filter order are insufficient. It is proved theoretically that the MNFDLMS algorithm can converge to Wiener solution when the non-causality condition and the filter order are not sufficient. Simulation and experiment also verify the effectiveness of the theoretical analysis.
【學(xué)位授予單位】：南京大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TB535

【共引文獻】

相關(guān)碩士學(xué)位論文 前1條

1 王月琳;封閉空間主動噪聲控制試驗平臺[D];清華大學(xué);2014年

，

本文編號：2292258