Please use this identifier to cite or link to this item:
標題: 結合即時訊號頻譜與即時雜訊頻譜估計之語音增強演算法
Improved speech enhancement with on-line signal and noise spectrum estimation
作者: 簡才淦
Chien, Tsai-Kan
關鍵字: speech enhancement;語音增強;spectral distortion;Tho-LMS;noise spectrum estimation;spectral subtraction;Beta-order minimum mean square;overlap and add;頻譜失真;Tho-LMS;雜訊頻譜能量估計;頻譜相減法;Beta -最小均方誤差法;重疊相加
出版社: 電機工程學系所
引用: [1] C.H. You,S.N. Koh,and S. Rahardja‚“Beta-order MMSE spectral amplitude estimation for speech enhancement,”IEEE Trans. Speech and Audio Processing, vol. 13,no. 4, pp. 475-486,July 2005. [2] T. Shimamura and J. Yamauchi‚“Non-stationary noise estimation utilizing harmonic structure for spectral subtraction”Signals, Systems and Computers‚vol. 2‚pp. 2305- 2309‚Nov. 2004. [3] R. Martin‚“Spectral subtraction based on minimum statistics”‚Proc.EUSIPCO 94‚pp. 1182-1185 , Sept.1994. [4] 侯政宇,“使用Tho-LMS 於即時頻譜估計之語音增強研究”,國立中興大學電機工程學系碩士論文,2007 [5] W.R. Wu and P.C. Chen‚“Adaptive AR Modeling in White Gaussian Noise” IEEE Trans. Signal Processing‚vol. 45‚no. 5‚pp. 1184-1192‚May 1997. [6] John R. Treichler‚C.Richard Johnson‚Jr.‚and Michael G. Larimore‚“Theory and design of adaptive filter”‚prentice Hall ,pp.92-115,2001. [7] A. de Cheveigne,“Separation of concurrent harmonic sounds:Fundamental frequency estimation and a time-domain cancellation model of auditory processing” J. Acoust. Soc. Am., vol.93, no. 6, pp. 3271–3290, June 1993. [8] Z. Goh, K. Tan and B. T. G. Tan, “Kalman-filtering speech enhancement method based on a voiced-unvoiced speech model,”IEEE Trans. Speech Audio Processing, vol. 7, no. 5, pp. 510–524, September 1999. [9] J. H. L. Hansen and M. A. Clements, “Constrained iterative speech enhancement with application to speech recognition,”IEEE Trans. Signal Processing, vol. 39, no. 4, April 1991. [10] N. Virag, “Single channel speech enhancement based on masking properties of the human auditory system,” IEEE Trans. Speech and Audio Processing, vol. 7, pp. 126–137, March 1999. [11] Y. Ephraim and D. Malah, “Speech enhancement using a minimum mean-square error short-time spectral amplitude estimatior” IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-32,no.6, pp. 1109-1121,Dec.1984. [12] Y. Ephraim and D. Malah‚“Speech enhancement a minimum mean-square error log-spectral amplitude estimatior”IEEE Trans. Acoust., Speech, Signal Process., vol. ASSP-33,no. 2, pp. 443-445,Apr.1985. [13] Y.Ephraim and D.Malamh,“Speech enhancement using a minimum mean-square error log-spectral amplitude estimation,”IEEE Trans. Acoustics,signal processing,vol.33 (2), pp. 443-445,Dec.1985. [14] K.G. Wu and P.C. Chen‚“Efficient speech enhancement using spectral subtraction for car hands-free application” IEEE Consumer Electronics, 2001. ICCE. International Conference on ‚ pp. 220-221‚June 2001. [15] I. Cohen,“Noise spectrum estimation in adverse environments:Improved minima controlled recursive averaging‚”IEEE Trans. Speech and Audio Processing, vol. 11,no. 5, pp. 466-475,Sept. 2003.
本論文旨在研究如何把這些受到雜音干擾的訊號(noisy signal)把雜訊消除,讓接收端的訊號能盡量和原始傳送端語音訊號能夠相同,在增強語音抑制雜訊的同時,還要考慮降低頻譜失真度(spectral distortion)的現象,使收話者可以清楚得聽到傳話者要表達的辭句。在此提出 Tho-LMS方法先將受雜訊干擾的訊號作初步消除雜訊,之後經過傅利葉轉換轉成頻域,得到即時訊號頻譜,並利用雜訊頻譜能量估計的雜訊估測方法-相減法、非語音中斷法、最小統計法,再經過語音增強(speech enhancement)的方法把雜訊消除,傳統語音增強法如頻譜相減法(spectral subtraction,SS)、韋納濾波器(wiener filter)。在論文中則是使用另一種語音增強,稱為Beta-最小均方誤差法(Beta-MMSE),係將含雜訊訊號直接乘上一個含Beta變數的增益函數(gain function),此最佳Beta值是利用雜訊頻譜估計出的雜訊能量值計算出的音框訊雜比(frame SNR);最後,將其增強後的訊號作傅利葉反轉換回時域,並使用重疊相加(overlap and add),重建原始乾淨的語音訊號。


The thesis means how to reduce noise for noisy speech signal‚then received signal can approach transmitted clean signal. When reducing noise for speech enhancement‚it will still consider to reduce spectral distortion phenomenon. So the receiver clearly listens sentences that the transmitter wants to speak. This paper uses Tho-LMS method that reduces noise for noisy speech signal‚and gets on-line signal spectrum that transfers to frequency domain with Fourier transform. We use noise spectrum estimation methods to get noise energy spectrum, such as subtraction method‚non-speech pause method‚and minimum statistics method, that supply speech enhancement to reduce noise.
Traditional speech enhancement methods like spectral subtraction and Wiener filter. The paper uses the other speech enhancement that is called Beta-order minimum mean square. After noise spectrum estimation we get frame signal-to-noise ratio,it can be calculated optimum Beta variable. And the noisy speech signal is directly multiplied the gain function that includes Beta variable. Finally, the enhanced speech signal transfers to time domain with inverse Fourier transform, and we use overlap and add to get reconstructed clean speech signal.
其他識別: U0005-2708200822363400
Appears in Collections:電機工程學系所

Show full item record

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.