基于双向多途散射模型的舱室空间噪声源定位方法

Noise source localization is very significant in the areas of cabin sound field design and control and the detection or recognition of various kinds of source signals. Most current methods have higher requirements on localization devices and have difficulty to overcome the multiple scattering effect in rooms. This project provides a bidirectional multiple scattering model (BMSM) based on room acoustic modeling and time reversal processing, in which a meshless model and the ray-tracing algorithm are combined to calculate the sound field from the reference sources to measurement positions in the forward scattering model, and the time reversal processing together with the measurement of two microphones are applied to achieve the space matching in the backward scattering model. This method utilizes the multiple scattering effect actively and can restrain the impacts of reverberation with less cost. The nodes in the meshless model can be regarded as the space sampling points for the backward scattering model, thus the forward and backward models can be integrated very well. The two microphones can simulate the binaural localization and give the fuzzy source areas for the precise localization of BMSM, which can narrow the searching ranges and enhance the computation efficiency. The research can also be applied into such areas as the noise source localization, speaker tracing, abnormal sound monitoring and robot hearing system in various types of rooms, sub-station, vehicle and stadium, etc.

噪声源定位在舱室声场设计与控制、舱内各种声信号的探测或识别等方面具有良好的应用价值。现有方法大多对定位设备有较高的要求，而且难于解决室内多途散射效应的影响问题。本项目提出一种结合室内声场建模和时间反转处理的双向多途散射模型。其中，正向散射模型采用无网格法与声线跟踪法结合的混合法，计算参考声源到观测点的声场，反向散射模型基于双传声器测量声场和时间反转处理理论，完成与实际声源的空间匹配。本方法主动利用多途散射效应，能以更小的代价抑制混响的影响；由于无网格模型的节点可作为反向搜索中的空间采样点，从而能巧妙融合正向与反向散射模型；采用双传声器可以先根据双耳听觉机理实现初步定位，然后再利用双向散射模型进行精确定位，可大大缩小反向散射模型的空间搜索范围，提高计算速度。本项目的研究成果还可以广泛应用于各类居室、地铁站、交通工具、艺术欣赏场馆的噪声源定位、说话人追踪、异音监控及机器人听觉系统等领域。

针对舱室空间声源定位问题，提出一种结合声场建模和时间反转处理的双向多途散射模型。其中，正向散射模型结合无网格法与声线跟踪法，计算参考声源到观测点的声场，反向散射模型基于双传声器测量声场和时间反转理论，完成与实际声源的空间匹配。本方法主动利用多途散射效应，能有效抑制混响的影响；由于将无网格模型的节点作为反向搜索中的空间采样点，巧妙融合了正向与反向散射模型；采用双传声器可以先根据双耳听觉机理实现初步定位，再利用双向散射模型进行精确定位，能大大减少反向散射模型的空间搜索范围，提高计算速度。利用大量的实际房间定位实验和仿真房间定位实验研究了混响时间、声源数量、类别、声源与传声器的距离、声源频率、低频噪声、节点密度与定位精度的关系，结果表明：所提方法具有良好的抗混响性能；声源数量少时效果更好；对稳态声源的定位效果优于脉冲声源；声源与传声器距离满足一定的信噪比要求即可保证定位效果；宽频声源定位效果优于单频声源；低频噪声对定位结果影响很小；大多数条件下，网格间距为15cm时即可达到95%以上的定位正确率。本项目的研究成果可广泛应用于各类交通工具、居室、地铁站、场馆的噪声源定位、说话人追踪、异音检测及机器人听觉系统等领域。