透過您的圖書館登入
IP:18.191.146.8
  • 學位論文

最小均方演算法以及功率頻譜密度差異值用於雜訊消除的分析

The Research on LMS and ΔPSD-Based Noise Reductio

指導教授 : 黃紹華

摘要


本論文針對環境中規律性雜訊,使用LMS與ΔPSD等方法做雜訊消除。本論文分為3個部分: 1. 分析功率頻譜函數的差値(Δ Power Spectrum Density)用於規律性雜訊與人聲的判別, 2. 使用LMS(least-mean-square)用於適應性雜訊學習與消除, 3. 去雜訊後語音使用增益,以加強人聲部分的訊號。 本論文所提出的方法,對於使用筆記型電腦內建麥克風或是在有規律雜訊干擾環境中使用網路語音通訊者,都可以將背景規律的雜訊消除,提升網路語音通話品質。 實驗結果證實,使用ΔPSD與LMS可以有效降低規律性雜訊的干擾,在CPU運算時間短,用於即時通訊上有著顯著的優點。

並列摘要


The LMS and ΔPSD-based method are employed to reduce the noise for the communication system. Three topics are employed to implement the noise reduction. First, the delta function of Power Spectrum Density is employed to distinguish the noise and un-noise signal. Second, the LMS(least-mean-square) method is employed to learn and reduce the noise adaptively. Third, the filter’s gain are employed to recover the gain of un-noise signal. The LMS and ΔPSD method are implemented in the VoIP to inspect the performance. The experimental results confirm that the LSM and ΔPSD-based method reduce the noise efficiently.

並列關鍵字

LMS ΔPSD noise reduction

參考文獻


[16] 黃信德,迴音消除於網路電話系統之研究與分析,台北科技大學,碩士論文,2005,pp.32
[3] S. Furui, “Cepstral Analysis Techniques for Automatic Speaker Verification,” IEEE Trans. On ASSP, 1981.
[4] A. Viikki and K. Laurila, “Cepstral Domain Segmental Feature Vector Normalization for NoiseRobust Speech Recognition,” Speech Communication, Vol. 25, 1998.
[5] X. Huang, A. Acero and H. Hon, “Spoken Language Processing: A Guide to Theory, Algorithm and System Development,” Prentice Hall PTR Upper Saddle River, NJ, USA, 2001.
[8] “Non-Linear Transformations of the Feature Space for Robust Speech Recognition”, A´ngel de la Torre, Jose´ C. Segura, Carmen Ben´ıtez, Antonio M. Peinado, Antonio J. Rubio, Dpto. Electr´onica y Tecn. Comp., Universidad de Granada, 18071 GRANADA (Spain), 2002

延伸閱讀