Page 34 - 《应用声学》2022年第1期

30 2022 年 1 月

表 6 DWPE 与 SN-WPE 处理结果主观测听 [5] Rompaey R V, Moonen M. Distributed adaptive node-
实验偏好度对比 specific signal estimation in a wireless sensor network with
partial prior knowledge of the desired source steering vec-
Table 6 Comparison of preferences for the
tor[C]//2019 27th European Signal Processing Conference
subjective listening test between DWPE
(EUSIPCO). IEEE, 2019: 1–5.
and SN-WPE results [6] Markovich-Golan S, Gannot S, Cohen I. Distributed mul-
tiple constraints generalized sidelobe canceler for fully
DWPE SN-WPE Equal connected wireless acoustic sensor networks[J]. IEEE
偏好度 89.06% 3.13% 7.81% Transactions on Audio, Speech, and Language Process-
ing, 2013, 21(2): 343–356.
表 7 DWPE 与 Cen-WPE 处理结果主观测听 [7] Zhang J, Heusdens R, Hendriks R C. Rate-distributed
实验偏好度对比 spatial filtering based noise reduction in wireless acous-
tic sensor networks[J]. IEEE/ACM Transactions on Au-
Table 7 Comparison of preferences for the
dio, Speech, and Language Processing, 2018, 26(11):
subjective listening test between DWPE
2015–2026.
and Cen-WPE results [8] Li Z, Yiu K F C, Dai Y H, et al. Distributed LCMV beam-
former design by randomly permuted ADMM[J]. Digital
DWPE Cen-WPE Equal Signal Processing, 2020, 106: 102820.
偏好度 12.50% 26.56% 60.94% [9] 丁建策, 厉剑, 郑成诗, 等. 基于稀疏表示和特征加权的离格
双耳声源定位 [J]. 应用声学, 2019, 38(6): 917–925.
4 结论 Ding Jiance, Li Jian, Zheng Chengshi, et al. Off-grid bin-
aural sound source localization using sparse representation
and feature weighting[J]. Journal of Applied Acoustics,
本文将WPE算法应用到WASNs上，提出了一
2019, 38(6): 917–925.
种实时分布式运算的自适应去混响 DWPE 算法。 [10] Naylor P, Gaubitch N. Speech dereverberation[M]. Lon-
在该方法中，所有节点共享相同的参考信号，并交换 don: Springer, 2010.
[11] 吴礼福, 陶明明, 郭业才. 频域合成房间频率响应的人工混响
本地输出，进行并行分布式计算。文中以理论推导
方法 [J]. 应用声学, 2020, 39(2): 163–168.
的方式证明了 DWPE算法的收敛性，在第 3 节通过 Wu Lifu, Tao Mingming, Guo Yecai. Artificial rever-
仿真以及主客观实验，证明了DWPE算法在显著降 beration by synthesizing room frequency response in the
低计算和传输成本的同时，能够达到与集中式算法 frequency domain[J]. Journal of Applied Acoustics, 2020,
39(2): 163–168.
相当的去混响性能。此外，收敛速度与噪声测试实 [12] Schwarz A, Reindl K, Kellermann W. A two-channel re-
验结果表明，提出的方法比集中式方法具有更快的 verberation suppression scheme based on blind signal sep-
收敛速度和更高的稳定性。 aration and wiener filtering[C]//Acoustics, Speech and
Signal Processing (ICASSP), 2012 IEEE International
Conference on. IEEE, 2012: 113–116.
参 考 文 献 [13] Habets E A P, Benesty J. A two-stage beamforming ap-
[1] Estrin D, Girod L, Pottie G, et al. Instrumenting proach for noise reduction and dereverberation[J]. IEEE
Transactions on Audio, Speech, and Language Processing,
the world with wireless sensor networks[C]//Acoustics,
Speech and Signal Processing (ICASSP), 2001 IEEE In- 2013, 21(5): 945–958.
ternational Conference on. IEEE, 2001: 2033–2036. [14] Miyoshi M, Kaneda Y. Inverse filtering of room acous-
[2] Markovich-Golan S, Bertrand A, Moonen M, et al. Op- tics[J]. IEEE Transactions on Audio, Speech, and Lan-
timal distributed minimum-variance beamforming ap- guage Processing, 1988, 36(2): 145–152.
proaches for speech enhancement in wireless acoustic sen- [15] 郭颖, 彭任华, 郑成诗, 等. 偏度最大化多通道逆滤波语声去
sor networks[J]. Signal Processing, 2015, 107: 4–20. 混响研究 [J]. 应用声学, 2019, 38(1): 58–67.
[3] Doclo S, Moonen M, van den Bogaert T, et al. Reduced- Guo Ying, Peng Renhua, Zheng Chengshi, et al. Max-
bandwidth and distributed MWF-based noise reduction imum skewness-based multichannel inverse filtering for
algorithms for binaural hearing aids[J]. IEEE Transactions speech dereverberation[J]. Journal of Applied Acoustics,
on Audio Speech & Language Processing, 2009, 17(1): 2019, 38(1): 58–67.
38–51. [16] 厉剑, 彭任华, 郑成诗, 等. 球谐域自适应混响抵消与声源定
[4] Bertrand A, Moonen M. Distributed adaptive node- 位算法 [J]. 声学学报, 2019, 44(5): 874–886.
specific signal estimation in fully connected sensor net- Li Jian, Peng Renhua, Zheng Chengshi, et al. Dereverber-
works—Part I: sequential node updating[J]. IEEE Trans- ation and localization using adaptive reverberation cancel-
actions on Signal Processing, 2010, 58(10): 5277–5291. lation in the spherical harmonic domain[J]. Acta Acustica,