Research On Virtual 3D Sound Technology And Implementation Of A Realtime Processing System Based On DSP

"Virtual Reality" is a very active research area nowadays and virtual sound is one of the most important parts. With the methods from virtual sound, we could generate different feelings in the sound space. This could be valuable in many application fields.Based on the research on HRTF (Head-Related Transfer Function) and its related algorithms, this paper is to construct a system of virtual 3D sound system for earphone as well as loudspeaker, using the HRTF data published by CIPIC lab.First, this paper described the general structure of virtual 3D sound system. For realization of virtual 3D sound system, we reviewed features of human ear, based on which, we summarized the main cues that aid in human sound localization. Then, we briefly described the traditional human ear positioning theory, duplex theory, and introduced the HRTF theory. We also indicated the restrictions in the implementation of real time virtual 3D sound system.In order to resolve such restrictions, we proposed to use PCA (Principal Components Analysis) to modeling HRTF. This model could significantly reduce the memory required by DSP, which could avoid using external high speed SRAM, which could reduce dramatically the system cost of virtual 3D sound system.We also analyzed the errors introduced by PCA and we proposed using Normalized PCA, which reduced the error introduced by PCA. Under the same quality requirement, Normalized PCA could also reduce further the memory requirement of real time system.Based on the Normalized PCA method proposed in this paper, we developed a virtual 3D sound earphone system. At the same time, proposed a method to solve the HRTF personalization using maximum correlation between maximal ITD and anthropometric parameters. Then, we also proposed a realization of virtual loudspeaker system.Finally, we use DIOPSIS (a chip from ATMEL based on ARM7TDMI and 40bits floating point DSP mAgic) to realize a virtual 3D sound system based on the method we proposed. After some objective experiments, our system has significant benefits over other virtual 3D sound systems.