Please use this identifier to cite or link to this item:
On-line Spectrum Estimation for Speech Enhancement Using ρ-LMS
|關鍵字:||Speech Enhancement;語音增強;Noise Cancellation;噪音消除||出版社:||電機工程學系所||引用:|| Thomas F. Quatieri, Discrete-Time Speech Signal Processing: Principles and Practice. Pearson education Taiwan Ltd, Prentice Hall PTR, Vol. 13, pp. 665-708, Fed. 2005.  Y. Ephraim and D. Malah, “Speech Enhancement Using a Minimum Mean-Square Error Short-Time Spectral Amplitude Estimator,” IEEE Trans. Acoustics, Speech, Signal Processing, Vol. 32 (6), pp. 1109-1211, Dec. 1984.  Y. Ephraim and D. Malah, “Speech Enhancement Using a Minimum Mean-Square Error Log-Spectral Amplitude Estimator,” IEEE Trans. Acoustics, Speech, Signal Processing, Vol. 33 (2), pp. 443-445, Dec. 1985.  C.H. You, S.N. Koh, and S. Rahardja, “β-order MMSE spectral amplitude estimation for speech enhancement,” IEEE Trans. Speech Audio Processing. Vol. 13 (4), pp. 475-486, July 2005.  C.H. You, S.N. Koh, and S. Rahardja, “Adaptiveβ-Order MMSE Estimation for Speech Enhancement,” Proc. IEEE Int. Conf. Acoust., Speech and Signal Processing, ICASSP-3, Vol. 1, pp. 852-855, 2003.  C.H. You, S.N. Koh, and S. Rahardja, “Masking-basedβ-order MMSE speech enhancement,” Speech Communication 48, pp.57-80, 2006.  Wu, W. R. and P. C. Chen, “Adaptive AR Modeling in White Gaussian Noise, ” IEEE Trans. Signal Processing, Vol. 45, pp. 1184-1192, May 1997  John R. Treichler, C. Richard Johnson, Jr.,and Michael G. Larimore, Theory and Design of Adaptive Filter. Tom Robbins, Prentice Hall, Inc., Vol. 4, pp. 92-115, 2001.  J. S. Lim, “Evaluation of a correlation subtraction method enhancing speech degraded by additive white noise,” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-26, no. 5, pp. 471-472, Oct. 1978.  R. Martin, “Spectral subtraction based on minimum statistics,” in Proc. Eur. Signal Processing Conf., 1994, pp. 1182-1185.||摘要:||
語音增強技術的發展，是為了消除干擾雜訊的影響，來改善音質、增加語音清晰度，或減少聽者的疲勞。常見的語音增強方法有頻譜刪減法(Spectral Subtraction)，最小均方誤差法(Minimum Mean-Square Error, MMSE)。此二類方法，都必須先取得訊號的頻譜值，(頻譜消去法中需要知道雜訊頻譜值，而最小均方誤差法中需要已知語音頻譜值)，才能進行消除雜訊的處理。然而，當輸入訊號只有包含雜訊干擾的語音時，如何取得上述之參數，便成為一個需要解決的問題。
傳統上，取得雜訊頻譜的方法是藉由所謂的VAD (Voice Activity Detector)，根據個別處理的音框的特性，判斷其為雜訊音框或是含語音之音框，進而產生新的雜訊音框值。另一方面，傳統上取得語音頻譜的方法是以前一個音框在消除完雜訊後的輸出訊號頻譜，當作目前音框中的語音訊號頻譜值。
The objective of developing speech enhancement (SE) technology for eliminating noise disturbance is to improve the overall quality, to increase intelligibility, or to reduce listener fatigue by suppressing noise and eliminating the distortion. There are two methods for speech enhancements, one is spectral subtraction (SS), and the other is minimum mean-square error (MMSE). First, we have to calculate the signal spectrum for denosing process in these two methods. The former requires the noise spectrum, and the latter requires the speech spectrum. Nevertheless, how to calculate the above parameters become a problem we have to solve only given a noisy speech signal.
The conventional approach to noise spectrum estimation is called “VAD” (Voice Activity Detector), which can detect where the processing frame is a speech presence one or not, and estimate a new noise based on the characteristics of the processing frame. On the other hand, the conventional method for obtain the speech spectrum is to utilize the denoised signal spectrum of last frame instead of current speech spectrum.
In this thesis, we propose an approach to the speech spectrum and noise spectrum estimation based on the analysis of AR model of the speech signal. In contrast, our proposed approach outperforms the conventional methods on estimating and denoising.
|Appears in Collections:||電機工程學系所|
Show full item record
TAIR Related Article
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.