| 时间: 2026-11-20 | 次数: |
阚阅,胡亚鹏,张腾飞,等.结合粒子群优化算法的单传声器阵列声源定位[J].河南理工大学学报(自然科学版),2027,46(1):147-156.
Kan Y, Hu Y P, Zhang T F,et al.Sound source localization with a single microphone array via particle swarm optimization[J].Journal of Henan Polytechnic University(Natural Science) ,2027,46(1):147-156.
结合粒子群优化算法的单传声器阵列声源定位
阚阅1, 胡亚鹏1, 张腾飞1, 史晨哲1, 查富生2, 高娃3, 刘欢1
1.河南理工大学 机械与动力工程学院,河南 焦作 454003;2.哈尔滨工业大学 机电工程学院,黑龙江 哈尔滨 150001;3.南京林业大学 家居与工业设计学院,江苏 南京 210037
摘要: 目的 为了解决基于到达时间差(TDOA)的单传声器阵列三维声源定位传统算法定位精度低、实时性差等问题,提出一种结合粒子群优化(PSO)的混合粒子群(Hybrid-PSO)声源定位算法。 方法 将传统广义互相关(GCC)声源定位方法与粒子群优化算法相结合,在保证实时性的基础上改善声源定位精度。首先,采用GCC算法估计出任意2个传声器之间的TDOA参数,根据获得的TDOA参数值构建含有声源位置参数的定位方程组,求解此方程组可获得声源位置粗略参数值;其次,利用该粗略位置参数,对PSO算法的搜索区域和搜索速度施加限制条件;最后,使用PSO算法在参数限制范围内进行搜索,得到精度更高的声源位置坐标参数。 结果 在相同的仿真条件下,与遗传算法、萤火虫算法、拟牛顿法、PSO算法相比,Hybrid-PSO算法拥有更好的实时性、准确度和鲁棒性。不同信噪比情况下,方位角 θ =30°,俯仰角 φ =40°,距离 r 为1~10 m的声源,最大定位误差分别为∆ r =1.63 m,∆ θ =0.81°,∆ φ =0.83°。距离 r =3 m,俯仰角 φ =30°,方位角 θ 为5°~85°的声源,最大定位误差分别为∆ r =0.09 m,∆ θ =0.72°,∆ φ =0.61°。距离 r =3 m,方位角 θ =40°,俯仰角 φ 为10°~80°的声源,最大定位误差分别为∆ r =0.19 m,∆ θ =2.17°,∆ φ =1.30°。 结论 Hybrid-PSO算法拥有更好的实时性、准确度和鲁棒性,在不同信噪比情况下均具有良好的定位精度。
关键词:粒子群优化算法;声源定位;传声器阵列;广义互相关方法;到达时间差
doi:10.16186/j.cnki.1673-9787.2024010024
基金项目:国家自然科学基金资助项目(U2013602,52105262);河南理工大学博士基金资助项目(B2019-51);河南省科技攻关项目(242102221041)
收稿日期:2024/01/08
修回日期:2024/09/16
出版日期:2026/11/20
Sound source localization with a single microphone array via particle swarm optimization
Kan Yue1, Hu Yapeng1, Zhang Tengfei1, Shi Chenzhe1, Zha Fusheng2, Gao Wa3, Liu Huan1
1.School of Mechanical and Power Engineering, Henan Polytechnic University, Jiaozuo 454003, Henan, China;2.State Key Laboratory of Robotics and System, Harbin Institute of Technology, Harbin 150001, Heilongjiang, China;3.College of Furnishings and Industrial Design, Nanjing Forestry University, Nanjing 210037, Jiangsu, China
Abstract: Objectives To address the issues of low localization accuracy and poor real-time performance in traditional algorithms for three-dimensional sound source localization using a single microphone array based on time difference of arrival (TDOA), a hybrid particle swarm optimization (Hybrid-PSO) sound source localization algorithm combining particle swarm optimization (PSO) is proposed.Methods The traditional generalized cross-correlation (GCC) sound source localization method is combined with the PSO algorithm to improve localization accuracy while ensuring real-time performance. First, the GCC algorithm is used to estimate the TDOA parameters between any two microphones, and a system of localization equations containing the sound source position parameters is constructed based on the obtained TDOA values. Solving this system yields rough estimates of the sound source position. Second, using these rough position parameters, constraints are imposed on the search region and search velocity of the PSO algorithm. Finally, the PSO algorithm is applied to search within the constrained parameter range to obtain more accurate sound source position coordinates.Results Under the same simulation conditions, compared with the genetic algorithm, firefly algorithm, quasi-Newton method, and standard PSO, the proposed Hybrid-PSO algorithm demonstrates superior real-time performance, accuracy, and robustness. Under different signal-to-noise ratios, for sound sources within the range of azimuth angle θ =30°, elevation angle φ =40°, and distance r =1~10 m, the maximum localization errors are ∆ r =1.63 m, ∆ θ =0.81°, and ∆ φ =0.83°. For sound sources within the range of distance r =3 m, elevation angle φ =30°, and azimuth angle θ = 5°~85°, the maximum localization errors are ∆ r =0.09 m, ∆ θ =0.72°, and ∆ φ =0.61°. For sound sources within the range of distance r =3 m, azimuth angle θ =40°, and elevation angle φ =10°~80°, the maximum localization errors are ∆ r =0.19 m, ∆ θ =2.17°, and ∆ φ =1.30°.Conclusions The Hybrid-PSO algorithm proposed in this paper exhibits higher localization accuracy and better robustness while guaranteeing real-time performance, and can meet the localization accuracy requirements under various SNR.
Key words:particle swarm optimization;sound source localization;microphone array;generalized cross-correlation method;time difference of arrival