整合全卷積神經網路、時序低通濾波與時頻遮罩法之語音強化

Speech Enhancement Based on the Integration of Fully Convolutional Network, Temporal Lowpass Filtering and Spectrogram Masking

劉冠毅(Kuan-Yi Liu)

指導教授 ： 洪志偉

國立暨南國際大學/科技學院/電機工程學系/碩士(2019年)

https://doi.org/10.6837/ncnu201900216

摘要

在本研究中，我們注重於消除語音信號中噪聲所帶來的失真問題，並開發了兩種新穎的無監督語音強化演算法，包括時序低通濾波器（TLF）與時頻遮罩法（RMM）。這兩種方法都是在語音信號的頻域振幅上進行處理。 TLF使用簡單移動平均濾波器來強調語音信號的低調變頻率區域，這可使語音信號擁有更豐富的語言信息並表現出更高的信噪比（SNR）。相比之下，在RMM中，我們使用遮罩值以逐點方式直接與語音頻譜相乘，而這些所使用的遮罩值與它們所對應的幅度成正比。此研究是在TIMIT資料庫上所進行的，初步實驗顯示這兩種新方法可以有效的提高被雜訊干擾的語音品質，並且它們都能與近期眾所皆知的監督語音增強方法全卷積網路（FCN）形成良好的加成性，使降噪的能力更好並得到更清晰的語音品值。

關鍵字

語音強化 ； 頻譜 ； 全卷積網路 ； 時序低通濾波器 ； 時頻遮罩法

並列摘要

In this study, we focus on the issue of noise distortion in speech signals, and develop two novel unsupervised speech enhancement algorithms including temporal lowpass filtering (TLF) and relative-to-maximum masking (RMM). Both of these two algorithms are conducted on the magnitude spectrogram of speech signals. TLF uses a simple moving-average filter to emphasize the low modulation frequencies of speech signals, which are believed to contain richer linguistic information and exhibit higher signal-to-noise ratios (SNR). Comparatively, in RMM we apply a mask that is directly multiplied with the speech spectrogram in a point-wise manner, and the used masking value is directly proportional to the magnitude of each temporal-frequency (T-F) point in the spectrogram. The preliminary experiments conducted on a subset of TIMIT database show that the two novel methods can promote the quality of noise-corrupted speech signals significantly, and both of them can be integrated with a well-known supervised speech enhancement scenario, namely fully convolutional network, to achieve even better perceptual speech quality values.

並列關鍵字

speech enhancement ； spectrogram ； fully convolutional network ； temporal lowpass filtering ； relative-to-maximum masking

參考文獻

[1]D. O' Shaughnessy, "Speech communications: human and machine," 2nd ed., Hyderabad, India: University Press (India) Pvt. Ltd., 2007.

[2]Y. Ephraim, H. L. Ari and W. Roberts, "A brief survey of speech enhancement," Electrical Engineering Handbook, 3rd ed. Boca Raton, FL: CRC, 2006.

[3]P. C. Loizou, "Speech enhancement: theory and practice," Taylor and F. Group, Eds. Boca Raton, FL, USA: CRC Press, 2013.

[4]R. Martin, "Spectral subtraction based on minimum statistics," In Proc. European Conference Signal Processing, pp. 1182–1185, 1994.

[5]P. Krishnamurthy, and S. R. M. Prasanna, "Modified spectral subtraction method for enhancement of noisy speech," in Proc. International Conference Signal, Image Processing, pp.146-150, Dec. 2005.