結合小波函數與FXLMS演算法於主動式噪音控制

Active Noise Control by Using Wavelet Function and FXLMS Algorithm

巫國煒(Guo-Wei Wu) ； 蕭俊祥(Jin-Siang Shaw)

《臺北科技大學學報》 45卷1期 (2012/09) Pp. 1-12

https://doi.org/10.29768/JNTUT.201209.0001

摘要

本研究以回授式主動噪音控制於管路減噪系統，利用聲源送出主噪音，經噪音計擷取訊號由A/D轉換給數位訊號處理作為第二路徑控制器使用，運算以X濾波最小均方法FXLMS演算法即時更新控制器權重，計算後輸出D/A轉換給放大器產生電壓訊號，推動喇叭產生反噪音與主噪音振幅相同，相位差180度的破壞性干涉，使殘餘噪音變小。實驗方法首先利用離線鑑別方式求得系統第二路徑之IIR濾波器權重，再由回授控制器FIR濾波器權重以FXLMS演算法不斷更新和學習。最後，由TI TMS320C6713 DSK卡每秒取樣頻率8000Hz，實現管路減噪控制。另外模擬實際環境所造成量測雜訊，在誤差訊號內加入白雜訊、雙頻及四頻週期訊號干擾，比較FXLMS演算法、FXLMS結合小波函數及FXLMS結合低通濾波器，對噪音源有效收斂與減噪評估。實驗與MATLAB模擬結果，主聲源頻率400～800Hz時FXLMS結合小波函數減噪效果最佳，受到白雜訊、雙頻及四頻週期訊號干擾，分別有21.9dBA、18.7dB及8.4dB減噪效果，比FXLMS演算法及結合低通濾波器更佳，證明ANC結合小波函數於管路減噪有效性。

關鍵字

主動式噪音控制 ； Haar小波 ； 數位訊號處理器

並列摘要

This study focuses on the feedback active noise control (ANC) for noise reduction in an acoustic duct. Noise signal from the sound source (i.e., the primary noise) is picked up by a sound-level meter and is A/D converted for the digital signal processor (DSP), which is used as the controller in the duct system. The controller parameters (weights) are updated based on the filtered-X least mean squares (FXLMS) algorithm in real-time. The processed signals from the controller are D/A converted and then amplified to generate anti-noise to suppress the duct noise. Anti-noise, which is of identical amplitude but a 180° phase difference to the primary noise, can interfere destructively with the primary noise so that the residue noise can be minimized. In the experiments, the weights of an IIR filter are first obtained by offline identification technique for the secondary path dynamics. The controller FIR filter weights are constantly updated and trained in the feedback ANC by using the FXLMS algorithm. To implement the noise reduction system, a TI TMS320C6713 DSK card is used at a sampling rate of 8000 Hz. In addition, to simulate the effect of the sensor noise in an actual environment, signal contamination, including white noise, and a dual-tone and a 4-tone periodic noises, is introduced to the error signal. Three control schemes, the FXLMS algorithm, the combined FXLMS and Haar wavelet function, the combined FXLMS and a low-pass filter, are compared for the effectiveness of noise reduction. From the experiments and the results of MATLAB simulation, the method of combining FXLMS with the Haar wavelet function proves to be the most effective in noise reduction. In a single frequency sound source with frequency of 400 Hz to 800 Hz, the noise reductions when the error signal is interfered with white noise, dual-tone and 4-tone periodic noises are respectively 21.9dB, 18.7dB, and 8.4dB. The results indicate the effectiveness of the combined FXLMS and Haar wavelet function for reducing duct noises.

並列關鍵字

Filtered-X Least Mean Squares(FXLMS) ； Active Noise Control ； Haar Wavelet ； Digital Signal Processor