Page 71 - 201901

P. 71

第 38 卷第 1 期郭颖等：偏度最大化多通道逆滤波语声去混响研究 67

尤其在混响较强的情况下优势更为明显，且算法复 [10] Papoulis A, Hoﬀman J G. Probability, random variables,
杂度更低，对高斯噪声的鲁棒性更强。应该指出的 and stochastic processes[M]. New York: McGraw-Hill,
2013.
是，本文所提的方法主要用于抑制早期混响所引起
[11] Gillespie B W, Malvar H S, Florêncio D A F. Speech
的谱染色现象，而对较长混响时间所引起的拖尾现 dereverberation via maximum-kurtosis subband adaptive
象抑制不明显，结合谱减法等后处理方法可以对残 ﬁltering[C]//Acoustics, Speech, and Signal Processing,
余晚期混响进行抑制，进而进一步提升可懂度。其 2001. Proceedings.(ICASSP’01). 2001 IEEE Interna-
tional Conference on. IEEE, 2001, 6: 3701–3704.
次，在研究中发现，在混响较强情况下，相比于多通 [12] Wu M, Wang D L. A two-stage algorithm for one-
道方法，单通道算法表现出了明显的局限性。另外， microphone reverberant speech enhancement[J]. IEEE
在实际应用中，本文所提方法的实时处理问题也是 Transactions on Audio Speech & Language Processing,
2006, 14(3): 774–784.
值得进一步深入研究的。
[13] Pääjärvi P, Leblanc J. Skewness maximization for impul-
sive sources in blind deconvolution[C]//Nordic Signal Pro-
参考文献 cessing Symposium: 09/06/2004-11/06/2004. Helsinki
University of Technology, 2004: 304–307.
[14] Mosayyebpour S, Sheikhzadeh H, Gulliver T A, et al.
[1] Flanagan J L, Johnston J D, Zahn R, et al. Computer-
Single-microphone LP residual skewness-based inverse ﬁl-
steered microphone arrays for sound transduction in large
tering of the room impulse response[J]. IEEE Transactions
rooms[J]. The Journal of the Acoustical Society of Amer-
on Audio, Speech & Language Processing, 2012, 20(5):
ica, 1985, 78(5): 1508–1518.
1617–1632.
[2] 肖栋, 向阳, 卓瑞岩, 等. 基于波束形成的多类型多声源定位
[15] Yoshioka T, Hikichi T, Miyoshi M, et al. Robust decom-
研究 [J]. 应用声学, 2017, 36(3): 220–227.
position of inverse ﬁlter of channel and prediction error
Xiao Dong, Xiang Yang, Zhuo Ruiyan, et al. Location of
ﬁlter of speech signal for dereverberation[C]//Signal Pro-
multiple sound source with multi-type based on beam-
cessing Conference, 2006 14th European. IEEE, 2006:
forming[J]. Journal of Applied Acoustics, 2017, 36(3):
1–5.
220–227.
[16] Rabiner L R, Schafer R W. Digital processing of speech
[3] Boll S F. Suppression of acoustic noise in speech using
signals[M]. Englewood Cliﬀs, NJ: Prentice-Hall, 1978.
spectral subtraction[C]//Acoustics, Speech, and Signal
[17] Allen J B, Berkley D A. Image method for eﬃciently sim-
Processing, IEEE International Conference on ICASSP.
ulating small-room acoustics[J]. The Journal of the Acous-
IEEE, 1979: 200–203.
tical Society of America, 1979, 65(4): 943–950.
[4] Li R, Bao C, Xia B, et al. Speech enhancement us-
[18] 饶宇安. 关于声学比 -混响时间 -语言清晰度关系的实验与理
ing the combination of adaptive wavelet threshold and
论计算 [J]. 声学学报, 1981, 17(1): 20–33.
spectral subtraction based on wavelet packet decomposi-
Rao Yu’an. Experimental and theoretical calculation of
tion[C]//Signal Processing (ICSP), 2012 IEEE 11th Inter-
the relationship among acoustic ratio, reverberation time
national Conference on. IEEE, 2012, 1: 481–484.
and speech intelligibility[J]. Acta Acustica, 1981, 17(1):
[5] Lebart K, Boucher J M, Denbigh P N. A new method
20–33.
based on spectral subtraction for speech dereverbera-
[19] Habets E A P. Single- and multi-microphone speech dere-
tion[J]. Acta Acustica united with Acustica, 2001, 87(3):
verberation using spectral enhancement[J]. Dissertation
359–366.
Abstracts International, 2007, 68(4): 10.6100/IR627677.
[6] Fang Y, Feng H, Chen Y. A robust interaural time diﬀer-
[20] Rix A W, Beerends J G, Hollier M P, et al. Percep-
ences estimation and dereverberation algorithm based on
tual evaluation of speech quality (PESQ)-a new method
the coherence function[J]. Applied Acoustics, 2018, 129:
for speech quality assessment of telephone networks and
126–134.
codecs[C]. Proceedings of the Acoustics, Speech, and Sig-
[7] Peng R, Tan Z H, Li X, et al. A perceptually motivated
nal Processing, 2001, 2: 749–752.
LP residual estimator in noisy and reverberant environ-
[21] Fall T H, Zheng C, Chan W Y. A non-intrusive qual-
ments[J]. Speech Communication, 2018, 96: 129–141.
ity and intelligibility measure of reverberant and derever-
[8] Zheng C, Peng R, Li J, et al. A constrained MMSE
berated speech[J]. IEEE Transactions on Audio Speech &
LP residual estimator for speech dereverberation in noisy
Language Processing, 2010, 18(7): 1766–1774.
environments[J]. IEEE Signal Processing Letters, 2014,
[22] Schwarz A, Kellermann W. Coherent-to-diﬀuse power ra-
21(12): 1462–1466.
tio estimation for dereverberation[J]. IEEE/ACM Trans-
[9] 赵红, 李双田. 改进的多级线性预测晚期混响抑制算法 [J]. 信
actions on Audio Speech & Language Processing, 2015,
号处理, 2014, 30(6): 674–682.
23(6): 1006–1018.
Zhao Hong, Li Shuangtian. Improved late reverber-
[23] Zheng C, Tan Z H, Peng R, et al. Guided spectrogram
ation suppression algorithm using multiple-step linear
ﬁltering for speech dereverberation[J]. Applied Acoustics,
prediction[J]. Journal of Signal Processing, 2014, 30(6):
2018, 134(5): 154–159.
674–682.

66 67 68 69 70 71 72 73 74 75 76