Abstract:
Over the last decades Virtual Reality (VR) technology has emerged to be a powerful tool for a wide variety of applications such as rapid prototyping, evaluation, therapy, or training tasks. State-of-the-art systems aim to reproduce virtual environments as realistic as possible in order to maximize the user’s feeling of immersion, presence and acceptance. Such systems deliver multiple sensory stimuli and provide an opportunity to act interactively, as reality is neither mono-modal nor static. For high quality auralizations (in analogy to visualization) of virtual environments, methods of Geometrical Acoustics (GA) are mostly applied to simulate the propagation of sound inside enclosures. By adapting acceleration algorithms from Computer Graphics, such as BSP- and Octrees, current implementations can manage the computational load of moving sound sources around a moving receiver in real-time – even for complex scenarios. While these data structures enable a fast intersection search of sound rays, they lack a good support of changing geometry due to their hierarchical nature. However, insertion, modification and extraction of geometrical objects are basic operations in many real-world experiences and should therefore be supported by the simulation as well. For this purpose the concept of Spatial Hashing was introduced, which also originates from Computer Graphics and is usually applied to collision detection tests of deformable objects. Here, the given geometry is encoded in a one-dimensional hash table, which supports geometry modification with a complexity of only 1, but also comprises the drawback of slower intersection tests when compared to BSP-/Octrees. This contribution describes the design, implementation and integration of a dynamic object controller in the real-time room acoustics simulation software RAVEN. RAVEN features a hybrid simulation approach using deterministic image sources for the direct sound and early reflections, and a stochastic ray tracer for the computation of the reverberant sound field. By adapting the concept of Spatial Hashing to the simulation algorithms, RAVEN is able to handle geometry modifications in real-time. The performance of the newly implemented data handling- and simulation routines is discussed and compared to that of Brute Force and BSP-based algorithms.