Spatial auditory information allows humans to separate sources from each other. Furthermore, source localisation can steer the focus towards events that are not in sight. 3D sound is hence an important feature of human hearing. To investigate the particular effects of sound perception, a controlled reproduction of spatial sound is needed in order to recreate an acoustic scene as accurately as possible. In contrast to headphone-based reproduction, spatial sound presentation via loudspeakers lowers the inhibition threshold as no device has to be attached to the listener. It also increases immersion into the scene as well as allowing communication between listeners or with a supervisor. The virtual acoustic scenes are typically simulated by auralization. The basis are computer simulations of room impulse responses (RIR), which can physically and perceptually be divided into different time sections: direct sound, early reflections and diffuse decay. Furthermore, different, loudspeaker-based reproduction methods exist with individual strengths and weaknesses. Preceding investigations hint, that for the perceptually relevant qualities of each time section one reproduction method is optimal, while the shortcomings of this reproduction method only applies for qualities that are relevant in another time section of the RIR.

The goal of the thesis is to implement a system that combines different reproduction methods to optimise the sound reproduction of virtual rooms. The user should furthermore be integrated into the system and the virtual room, i.e. speaking should result in a room dependent perception of reverberation. Additionally, the influence of the room the loudspeakers are placed in (listening room) should be compensated to avoid the requirement of an anechoic chamber for reproduction.

The creation of such a system is based on various steps. The individual systems have to be implemented and combined with the room acoustical simulation. The techniques implemented are binaural crosstalk cancellation (CTC), 2nd Order Ambisonics and Vector-Base Amplitude Panning (VBAP) due to their ability to provide full, three-dimensional source positioning. They are tested against each other in various perceptual qualities including localisation accuracy. To keep the original structure of the RIR, and ensure a correct transition between the reproduction techniques, their loudness has to be calibrated against a reference condition. To compensate the room, prior knowledge of the virtual room and the listening room is used. The virtual room is changed until the combination of virtual room and listening room complies with the unchanged virtual room.

Based on the findings, a suitable reproduction technique is selected for each section of the RIR. The final system is then again tested to verify that improvements in different perceptual qualities are audible and identify any trade-offs. The user integration, speaking into a virtual room, is evaluated.

Melden Sie sich hier an um Einladungen zu den Kolloquium-Vorträgen per E-Mail zu erhalten.
Register here to receive the invitations to colloquium talks via e-mail.

Zoom-Meeting-ID: 954 4073 3814
Passwort: 450783