S-Space College of Engineering/Engineering Practice School (공과대학/대학원) Dept. of Electrical and Computer Engineering (전기·정보공학부) Theses (Ph.D. / Sc.D._전기·정보공학부)
Sound Diffraction Filter Design for Parametric Processing of Acoustic Virtual Reality
가상현실 음향의 파라메트릭 처리를 위한 소리 회절 필터 설계
- 공과대학 전기·컴퓨터공학부
- Issue Date
- 서울대학교 대학원
- Sound diffraction; Sound occlusion; Diffraction filter; Perception boundary; Virtual reality; Artificial neural network
- 학위논문 (박사)-- 서울대학교 대학원 : 공과대학 전기·컴퓨터공학부, 2018. 2. 김성철.
- This dissertation investigates sound diffraction around a sound-occluding obstacle in acoustic virtual reality applications. The concept of ``diffraction filter'' is proposed to represent the changes in sound when it is diffracted around an obstacle. The diffraction filter in free field is obtained using the Fresnel-Kirchhoff diffraction formula and Babinet's principle. To the author's knowledge, this dissertation is the first to construct the transfer function of sound diffraction using the diffraction formula. The diffraction transfer function obtained by numerical computation of the diffraction formula is compared with that measured in the same environment. The comparison validates the numerically computed diffraction filter. As a result, the sound diffraction effect due to obstacles can be obtained from the numerical calculation without taking measurements one by one.
Researches have reported that humans can perceive the sound-occluding obstacles with auditory perception. Unfortunately, these studies have considered limited number of sound-occluding situations because their experiments were mainly based on the actual speakers and microphones. Consequently, systematic study on obstacle perception could not develop and the conclusions of the studies did not coincide. With the aid of recent developments in computational techniques, there have been several attempts to reproduce the sound-occluding effects in the digital domain. However, as these studies were mainly intended to achieve real-time sound effects in virtual reality, they failed to yield physically accurate results.
In these respects, this dissertation attempts to systematically investigate the auditory perception of obstacles by simulating the sound occlusion situations. Simulating various sound occlusion situations is difficult because it involves diffraction and wave interference phenomena. Interactions between sound source, obstacle, and the surrounding environment further impedes the precise analyses. Therefore, as the first step into the systematic study on the auditory perception of obstacles, this dissertation excludes the other environmental effects such as room reverberation, and considers only the diffraction around the obstacle. The diffracted sound is obtained by numerically computing the diffraction filter. In order to provide diversity of scenarios, the shape and size of the obstacle and the direction of the sound source are kept freely adjustable. Particularly, rectangular-shaped obstacles with different aspect ratio values are considered in order to approximately represent the various shapes of obstacles. Then, the minimum size of the perceptible obstacle size is investigated with various parameters. Experimental results exhibit that the minimum perceptible obstacle depends on the aspect ratio of the obstacle, direction of the incoming sound, hearing ability of individuals, and experience of professional music training. Moreover, the concept of ``perceived magnitude response'' is introduced to interpret the determination of minimum perceptible obstacle size.
In the next step, this dissertation investigates physical aspects of the diffraction filter. Magnitude response of the diffraction filter exhibits significant scalability with respect to the obstacle conditions. This scalability is likely to inspire the design of diffraction filters for virtual reality applications that are both physically meaningful and computationally efficient. With regard to this aspect, this dissertation proposes two practical approaches to apply a diffraction filter in acoustic virtual reality. The first approach utilizes the artificial neural network (ANN) structure to construct a lookup table of the diffraction filters for various sound-occluding situations. The other approach provides a physically meaningful model to characterize the magnitude response of the diffraction filter. This model further interprets the scalability of the diffraction filter. These two approaches will enable real-time reproduction of sound diffraction effect in acoustic virtual reality.