Physically-based auralization : design, implementation, and evaluation

摘要

The aim of this research is to implement an auralization system that renders audible a 3D model of an acoustic environment. The design of such a system is an iterative process where successive evaluation of auralization quality is utilized to further refine the model and develop the rendering methods. The work can be divided into two parts corresponding to design and implementation of an auralization system and evaluation of the system employing objective and subjective criteria.The presented auralization method enables both static and dynamic rendering. In dynamic rendering positions and orientations of sound sources, surfaces, or a listener can change. These changes are allowed by modeling the direct sound and early reflections with the image-source method. In addition, the late reverberation is modeled with a time-invariant recursive digital filter structure. The core of the thesis deals with the processing of image sources for auralization. The sound signal emitted by each image source is processed with digital filters modeling such acoustic phenomena as sound source directivity, distance delay and attenuation, air and material absorption, and the characteristics of spatial hearing. The digital filter design and implementation of these filters are presented in detail. The traditional image-source method has also been extended to handle diffraction in addition to specular reflections.The evaluation of quality of the implemented auralization system was performed by comparing recorded and auralized soundtracks subjectively. The compared soundtracks were prepared by recording sound signals in a real room and by auralizing these signals with a 3D model of the room. The auralization quality was assessed with objective and subjective methods. The objective analysis was based on both traditional room acoustic criteria and on a simplified auditory model developed for this purpose. This new analysis method mimics the behavior of human cochlea. Therefore, with the developed method, impulse responses and sound signals can be visualized with similar time and frequency resolution as human hearing applies. The evaluation was completed subjectively by conducting listening tests. The utilized listening test methodology is explained and the final results are presented. The results show that the implemented auralization system provides plausible and natural sounding auralizations in rooms similar to the lecture room employed for evaluation.