A virtual sound rendering is a technology, which places a virtual sound source at an arbitrary position in 3D space. Human vision is mostly confined to the area in the front and so humans depend heavily on the sense of hearing to gather information in area out of sight. Thus, a virtual reality system capable of producing 3D sound effects gives the user a much better sense of reality than the system without such capability. There are generally two ways to render a virtual sound. One is a binaural sound system utilizing a 2-channel loudspeaker set or a headphone. The other is a multichannel 3D sound system utilizing a multiple loudspeaker set.
There are two kinds of hot issues on the binaural sound rendering. One is the design method of crosstalk cancellation. The other is the loudspeaker arrangement for robust performance. The binaural sound system utilizing a 2-channel loudspeaker set requires crosstalk cancellation, which is generally designed in the form of an FIR filter by using an adaptive scheme. The adaptive algorithm implemented in the filter requires tuning parameters such as filter length, modeling delay and step size which are difficult to obtain. In Chapter 2, an effective design method of crosstalk cancellation is proposed which directly obtains the inverse filter with the desired accuracy and less time and effort.
The crosstalk can be imperfectly cancelled and the localization performance of the binaural sound system deteriorates if there exists an uncertainty such as head movement. In Chapter 3, loudspeaker arrangement method is proposed for robust performance of the binaural sound system. The effect of uncertainty on localization performance of the binaural sound system is theoretically investigated. New robustness measure is defined and the robustness is analyzed with respect to the span angle of two loudspeakers and the frequency of the sound source. The loudspeaker arrangement methods are suggested for robust binaural sound system according to head movement of a listener.
The multichannel 3D sound system generally depends on the amplitude panning method to render virtual sound. The amplitude panning methods are derived under the assumption that all loudspeakers are placed at equal distance from head center, which is often easily violated in real practice. In Chapter 4, the method to determine an appropriated loudspeaker output without the burden of position constraint is proposed.
To verify the feasibility of the proposed methodologies, computer simulations are carried out and the results are presented in the corresponding chapters. In Chapter 5, subjective tests are carried out for arrangement and driving method of two channel loudspeakers and driving method of multiple loudspeakers. Some examples of virtual 3D sound rendering systems for virtual reality simulators such as bicycle and helicopter simulator are introduced.