基于贝塞尔平滑的改进SRP声源定位算法

Improved SRP sound source localization algorithm based on Bezier smoothing

为增强可控响应功率(Steered Response Power,SRP)声源定位算法的抗噪性能,提出了一种基于贝塞尔平滑(Bezier Smoothing,BS)的改进SRP声源定位算法。所提算法结合多帧短时长、等间隔的阵列声音数据,对各组阵元对的互功率谱进行贝塞尔曲线拟合以减少随机噪声干扰,并根据各频点复数域内随时间变化的程度调整其在互功率谱中的权重,突出稳定信号的SRP占比。低信噪比条件下,与传统SRP算法相比,所提算法的定位角度的均方根误差在仿真测试中降低了39.7%。仿真与实验结果表明,相较传统SRP算法,所提算法提高了对随机噪声的鲁棒性。

To enhance the anti-noise performance of the steered response power(SRP)sound source localization algorithm,an improved SRP sound source localization algorithm based on Bezier smoothing(BS)is proposed.The proposed algorithm combines multiple frames of short-duration and equal-spaced array sound data,and performs Bezier curve fitting on the cross-power spectrum of each array element pair to resist random noise interference.The degree of adjustment of its weight in the cross power spectrum highlights the SRP proportion of the stable signal.Under the condition of low signal to noise ratio,compared with other traditional SRP algorithms,the root mean square error of the localization angle of the proposed algorithm is reduced by 39.7%on average in the simulation test,which shows that the proposed algorithm improves the robustness to random noise in comparison with traditional SRP algorithms.