基于高阶Ambisonics的2.5维近场声源合成

2.5D Near-Field Sound Source Synthesis Using Higher Order Ambisonics

针对空间声信号处理中二维声场合成与实际声源维数失配问题及近场声源合成中的近场失真问题,本文提出了一种2.5维声场合成方法和一种改进的维纳滤波的近场补偿方法.2.5维声场合成基于高阶Ambisonics的球谐函数展开.改进的维纳滤波方法用于补偿高阶Ambisonics合成2.5维近场声源时引入的近场失真.基于连续扬声器阵列假设理论,推导出了扬声器激励信号的一般表达形式.声场合成实验结果表明:与余弦正则化方法产生的相对误差0.23相比,所提方法合成2.5维近场声源的相对误差减少到了0.031;与高通滤波法相比,所提方法可以提供更大范围的有效听音区.

In the spatial acoustic signal processing, for the issue of dimensionality mismatch between 2-dimensional （2D） sotmd field synthesis and practical sound source, and the issue of near-field distortion in the near-field sound source syn- thesis,this paper presents a 2.5-dimensional （2.5D） sound field synthesis method and a modified Wiener filter method for near-field compensation. The 2.5D sound field synthesis is achieved using the spherical harmonics expansion of higher order ambisonics （HOA）. The modified Wiener filter is used to compensate the near-field distortion which arises when HOA is uti- lized to reproduce a 2.5D near-field sound source. Based on the concept of continuous loudspeaker array hypothesis, a general expression of the loudspeaker driving signal is derived. The experiments show that the proposed method reduced the relative er- ror of the synthesized 2.5D near-field from 0.23 to 0.031 in comparison with the cosine regularization method. Meanwhile,the proposed method provides a larger effective listening area compared with the high-pass filter method.