基于GSPAP的子带自适应声反馈消除算法

发布时间：2018-03-01 05:22

本文关键词： 声反馈抑制子带稳态增益高斯-赛德尔伪放射投影(GSPAP)　出处：《清华大学学报(自然科学版)》2017年07期 　论文类型：期刊论文

【摘要】：声反馈抑制需求鲁棒、高效的自适应滤波器。该文提出一种基于Gauss-Seidel伪放射投影(Gauss-Seidel pseudo affine projection,GSPAP)的子带自适应声反馈消除算法。通过加权重叠相加滤波器组进行子带划分,子带上采用参考信号的自相关矩阵取代能量对滤波器的自适应步长进行归一化;在自相关矩阵的迭代公式中引入基于输入信号和参考信号能量最大值的自整定系数,增强算法鲁棒性;选用二阶GSPAP迭代法对自相关矩阵解算求逆,以平衡算法性能与复杂度。实验结果表明:在相同滤波器长度的条件下,该文方法获得11~22dB的稳态增益增量,比时域归一化最小均方误差(normalized least mean square,NLMS)方法提升20%~55%。
[Abstract]:Acoustic feedback suppression requires robust and efficient adaptive filters. In this paper, a subband adaptive acoustic feedback cancellation algorithm based on Gauss-Seidel pseudo-radiative projection Gauss-Seidel pseudo affine projection is proposed. In the subband, the autocorrelation matrix of the reference signal is used instead of the energy to normalize the adaptive step size of the filter, and the self-tuning coefficient based on the maximum energy of the input signal and the reference signal is introduced into the iterative formula of the autocorrelation matrix. The robustness of the algorithm is enhanced, the second order GSPAP iteration method is used to inverse the autocorrelation matrix, and the performance and complexity of the algorithm are balanced. The experimental results show that the steady gain increment of 1122 dB is obtained under the condition of the same filter length. Compared with the normalized least mean square-NLMS method, the minimum mean square error in time domain is increased by 20% and 55%.
【作者单位】：集美大学信息工程学院海上通信与智能电子系统福建省高等学校重点实验室;清华大学信息技术研究院语音和语言技术中心;厦门莱亚特医疗器械有限公司;
【基金】：福建省科技计划工业引导性项目(2015H0027) 厦门市科技计划产学研协同创新项目(3502Z20163009)
【分类号】：TN713;TN912.3

【相似文献】