Auralization of Urban Environments - Real-time Simulation of Diffraction

在DFG和FAPERJ的提案征集中，一个来自巴西的研究小组（UFRJ，里约热内卢联邦大学）和一个来自德国的研究小组（RWTH， Rheinisch-Westfälische亚琛工业大学）申请了一个联合项目。本提案的目标是一种先进的方法，在噪声传播和评估的应用中实现3D声音模拟（听觉化）：城市环境的听觉化。特别是，德国团队将主要关注城市环境中声音衍射的实时模拟。模拟方法将基于几何声学。它指的是标准的噪声映射计算，但它将通过数字滤波器表示的有限边缘的频率相关衍射扩展已建立的程序。第一步是根据场景几何来确定主要的传播路径。专注于这些贡献将减少从分析模型（Biot-Tolstoy, Medwin， Svensson等）推导衍射滤光器的计算工作量。控制计算努力的策略将通过实现近似来有效地模拟衍射（Maekawa， UTD， Svensson等）。这些基本的衍射对象在类似的基本运算中都有它们的算法对应部分，它们将使用共享内存并行化有效地求解。将这些基本计算分配到可用资源（多核处理器的单元、pc集群的节点）需要智能调度。为了简化几何输入数据的交换，德国和巴西的研究人员将采用既定的数据格式，如CityGML。通过将这些格式整合为声音传播模拟和听觉化的基础，可以直接使用现有的数据（例如地籍地图和大城市的数字表示）。巴西研究小组还将着重于适当的视觉捕捉和声学记录城市动态。在一项共同努力下，巴西和德国的成果将最终利用亚琛和里约热内卢非常不同的城市模式和噪音情况，在数字和心理声学上进行评价和比较。

Within the call for proposals from DFG and FAPERJ, a research group from Brazil (UFRJ, Universidade Federal Rio de Janeiro) and a research group from Germany (RWTH, Rheinisch-Westfälische Technische Hochschule Aachen) apply for a joint project. The objective of this proposal is an advanced methodology to achieve a 3D sound simulation (auralization) in the application of noise propagation and assessment: Auralization of Urban Environments. In particular, the German group will mainly focus on real-time simulation of sound diffraction in urban environments. The simulation methodology will be based on Geometrical Acoustics. It refers to standard noise mapping calculations but it will extend the established procedures by frequency-dependent diffraction at finite edges represented as digital filters. The first step will be identifying the major propagation paths based on a scene geometry. Concentrating on these contributions will decrease the computational effort to derive diffraction filters from analytic models (Biot-Tolstoy, Medwin, Svensson, etc.). A strategy to control the computation efforts will be pursued by implementing approximations to efficiently model diffraction (Maekawa, UTD, Svensson, etc.). These basic diffraction objects have their algorithmic counter parts in likewise basic operations, which will be solved efficiently using shared-memory parallelization. The distribution handling of these basic calculations to the available resources (units of a multicore processor, nodes of a PC-cluster) require intelligent scheduling. In order to simplify the exchange of geometrical input data, the German and Brazilian researchers will settle on established data formats, such as CityGML. By making it possible to integrate these formats as basis for the sound propagation simulation and auralization, already existing data can be used directly (e.g. cadastral maps and digital representations of large cities). The Brazilian research team will also focus on appropriately visually capturing and acoustically recording urban dynamics. In a joint effort, the results of the Brazilian and German accomplishments will be finally evaluated and compared numerically and psycho-acoustically using both the very different city models and noise situations in Aachen and Rio de Janeiro.