Page 12 - 《应用声学》2022年第6期

858 2022 年 11 月

目标声源的位置变化更具鲁棒性。仿真实验的结果
验证了该方法的有效性。
d

参 考 文 献
d
[1] Yoshioka T, Sehr A, Delcroix M, et al. Making machines
understand us in reverberant rooms: robustness against
图 8 传声器位置示意图 reverberation for automatic speech recognition[J]. IEEE
Signal Processing Magazine, 2012, 29(6): 114–126.
Fig. 8 Diagram of the microphone position
[2] Beutelmann R, Brand T. Prediction of speech intelli-
gibility in spatial noise and reverberation for normal-
ຉ־ηՂ ԝຉ־ηՂ hearing and hearing-impaired listeners[J]. The Journal of
ࣜదឭភ̡ͯᎶቊԫೝ฾ᄊԝຉ־ηՂ the Acoustical Society of America, 2006, 120(1): 331–342.
2.5 [3] 齐园蕾, 杨飞然, 杨军. 基于卡尔曼滤波的低复杂度去混响算
法 [J]. 应用声学, 2018, 37(4): 559–566.
2.0
Qi Yuanlei, Yang Feiran, Yang Jun. Kalman filter based
low-complexity dereverberation algorithm[J]. Journal of
PESQ४Ѭ 1.5 [4] 张雄伟, 李轶南, 郑昌艳, 等. 语音去混响技术的研究进展与
Applied Acoustics, 2018, 37(4): 559–566.
1.0
展望 [J]. 数据采集与处理, 2017, 32(6): 1069–1081.
0.5 Zhang Xiongwei, Li Yinan, Zheng Changyan, et al.
Speech dereverberation: review of state-of-the-arts and
0 prospects[J]. Journal of Data Acquisition and Processing,
1 2 3
2017, 32(6): 1069–1081.
ᡰሏ d/m
[5] Lebart K, Boucher J M, Denbigh P N. A new method
(a) PESQ४Ѭ
4.0 based on spectral subtraction for speech dereverbera-
tion[J]. Acta Acustica united with Acustica, 2001, 87(3):
3.5
359–366．
3.0 [6] Habets P E. Multi-channel speech dereverberation based
SRMR४Ѭ 2.5 on a statistical model of late reverberation[C]. IEEE In-
2.0
ternational Conference on Acoustics, Speech and Signal
1.5
Processing, 2005, 4: 173–176.
1.0 [7] Nakatani T, Yoshioka T, Kinoshita K, et al. Speech dere-
verberation based on variance-normalized delayed linear
0.5
prediction[J]. IEEE Transactions on Audio Speech and
0
1 2 3 Language Processing, 2010, 18(7): 1717–1731.
ᡰሏ d/m [8] Nakatani T, Yoshioka T, Kinoshita K, et al. Blind
(b) SRMR४Ѭ speech dereverberation with multi-channel linear predic-
图 9 声源距传声器不同距离时算法平均 PESQ 和 tion based on short time Fourier transform represen-
tation[C]. IEEE International Conference on Acoustics,
SRMR 得分
Speech and Signal Processing, 2008: 85–88.
Fig.9 The algorithm averages PESQ and SRMR
[9] Xiang T, Lu J, Chen K. Multi-channel adaptive derever-
scores when the sound source is different from the beration robust to abrupt change of target speaker posi-
microphone tion[J]. The Journal of the Acoustical Society of America,
2019, 145(3): EL250–EL256.
4 结论 [10] 李毅. 自适应滤波及滤波算法研究 [D]. 西安: 西北工业大学,
2003.
[11] Yoshioka T, Tachibana H, Nakatani T, et al. Adap-
本文研究了一种能够检测说话人位置变化的 tive dereverberation of speech signals with speaker po-
自适应语声去混响方法，该方法采用 RLS 算法实现 sition change detection[C]. IEEE International Confer-
滤波器系数的自适应估计更新，如果声源的位置发 ence on Acoustics, Speech and Signal Processing, 2009:
3733–3736.
生变化，该方法将检测这些变化并重新初始化滤波
[12] 向腾. 混响环境下的语音增强研究 [D]. 南京: 南京大学,
器系数的估计，使基于RLS的MCLP去混响方法对 2020.