Page 28 - 《应用声学》2022年第6期

874 2022 年 11 月

表 6 最大稳态增益 [8] 江虹, 周上清, 刘鹏辉, 等. 一种基于对数函数的改进变步长
Table 6 Maximum stable gain LMS 算法研究 [J]. 科技风, 2020(19): 51–52, 57.
Jiang Hong, Zhou Shangqing, Liu Penghui, et al. Study on
an improved step size factor of variable LMS algorithm[J].
算法 无反馈抑制 LMS NLMS PNLMS 本论文算法
Technology Wind, 2020(19): 51–52, 57.
MSG/dB 31.36 39.91 48.03 48.88 51.87 [9] Duttweiler D L. Proportionate normalized least mean-
squared adaptation in echo cancelers[J]. IEEE Trans-
4 结论 actions on Speech and Audio Processing, 2000, 8(5):
508–518.
[10] 王侠, 梁瑞宇, 王青云, 等. 应用于助听器反馈抑制系统的变
本文针对助听器反馈抑制领域提出了一种基
步长归一化子带自适应滤波算法 [J]. 东南大学学报 (自然科学
于信噪比的自适应滤波算法，以解决语声信号的相 版), 2015, 45(3): 417–422.
关性较大，容易使自适应滤波器出现较大失调的问 Wang Xia, Liang Ruiyu, Wang Qingyun, et al. Variable
题。当信噪比高时，将自适应滤波算法步长控制在 step size normalized subband adaptive filter algorithm for
acoustic feedback cancellation in hearing aids[J]. Journal
一个较小值，减少算法失调；当信噪比低时，将自适
of Southeast University(Natural Science Edition), 2015,
应滤波算法步长控制在一个较大值，加快收敛速度。 45(3): 417–422.
通过实验仿真，本文提出的基于信噪比的 NLMS 算 [11] 章坚武, 余皓, 章谦骅. 改进的双曲正切函数的变步长 LMS
算法 [J]. 通信学报, 2020, 41(11): 116–123.
法较传统的 NLMS 算法在性能上分别降低约 1 dB
Zhang Jianwu, Yu Hao, Zhang Qianhua. Improved vari-
的平均稳态失调量、2 dB 的稳态失调范围，提高了 able step-size LMS algorithm based on hyperbolic tangent
4 dB 的最大稳态增益，同时具有更快的收敛速度， function[J]. Journal on Communications, 2020, 41(11):
实现了传统NLMS算法的优化。对比本文所提出的 116–123.
[12] Hellgren J. Analysis of feedback cancellation in hearing
算法与其他参考算法，验证了本文提出的优化算法
aids with filtered-X LMS and the direct method of closed
具有较低的失调量、较快的收敛速度，同时能够达 loop identification[J]. IEEE Transactions on Speech and
到较高的增益，均优于其他3种算法。 Audio Processing, 2002, 10(2): 119–131.
[13] 孙志文. 基于自适应滤波的声反馈抑制算法研究与实现 [D].
成都: 电子科技大学, 2019.
[14] 高鹰, 谢胜利. 一种变步长 LMS 自适应滤波算法及分析
参 考 文 献
[C]//第十九届中国控制会议论文集 (一), 2000: 162–166.
[15] Khoubrouy S A, Panahi I M S. An efficient delay less
[1] 凃磊, 任杰, 赵坚, 等. 助听器双重反馈抑制算法中的参数效应 sub band filtering for adaptive feedback compensation in
研究 [J]. 中国听力语言康复科学杂志, 2020, 18(5): 383–386. hearing aid[J]. Journal of Signal Processing Systems, 2016,
Tu Lei, Ren Jie, Zhao Jian, et al. The parameter effect of 83(3): 401–409.
the Dual-Mode feedback canellation algorithm for hearing [16] Morgan D R, Hall J L, Benesty J. Investigation of sev-
aids[J]. Chinese Scientific Journal of Hearing and Speech eral types of nonlinearities for use in stereo acoustic echo
Rehabilitation, 2020, 18(5): 383–386. cancellation[J]. Speech & Audio Processing IEEE Trans-
[2] Patronis E T. Acoustic feedback detector and automatic actions on, 2001, 9(6): 686–696.
gain control: U.S.A., US4079199[P]. 1978–03–14. [17] Joncour Y, Sugiyama A. A stereo echo canceller with
[3] Schroeder M R. Improvement of acoustic feedback stabil- pre-processing for correct echo-path identification[C]. Pro-
ity in public address systems[J]. The Journal of the Acous- ceedings of ICASSP, Seattle, 1998: 3689–3692.
tical Society of America, 1959, 31(6): 851–852. [18] Haykin S. Adaptive filter theory[M]. 5th Edition. Canada:
[4] Kawamura A, Matsumoto M, Serikawa M, et al. Sound Publishing House of Electronics Industry, 2014.
amplifying apparatus with automatic howling-suppressing [19] Martin R. Noise power spectral density estimation based
function: Japan, EP0599450[P]. 2001–11–21. on optimal smoothing and minimum statistics[J]. IEEE
[5] Kates J M. Feedback cancellation in hearing aids: results Transactions on Speech and Audio Processing, 2001, 9(5):
from a computer simulation[J]. IEEE Transactions on Sig- 504–512.
nal Processing, 1991, 39(3): 553–562. [20] Ni J, Li F. A variable step size matrix normalized subband
[6] Slock D. On the convergence behavior of the LMS and adaptive filter[J]. IEEE Transactions on Audio Speech &
the normalized LMS algorithms[J]. IEEE Transactions on Language Processing, 2010, 18(6): 1290–1299.
Signal Processing, 1993, 41(9): 2811–2825. [21] Spriet A, Proudler I, Moonen M, et al. Adaptive feedback
[7] Anand A, Kar A, Swamy M N S. An improved CLMS algo- cancellation in hearing aids with linear prediction of the
rithm for feedback cancellation in hearing aids[J]. Applied desired signal[J]. IEEE Transactions on Signal Processing,
Acoustics, 2018, 129: 417–426. 2005, 53(10): 3749–3763.