Characteristics Analysis And Efficiency Modeling For Hrtfs In The Realization Of Virtual Auditory Space

The technique of spatial hearing mainly studies the elementary principles of how human auditory system distinctly perceives sound source's spatial position, which has extensive applications in the field of psychoacoustics and computer science. In this research field, people have long ago taken cognizance of this fancy phenomenon; and the systemic description, HRTFs (Head Related Transfer Functions) or relevant HRIRs (Head Related Impulse Responses) in time domain, has been introduced to describe the entire spatial information for the recognition of sound's position, where interaural time difference (ITD) can be seen as the delay between lateral HRTFs and low-amplitude contralateral HRTFs in transmission time lags; interaural intensity difference (IID) can be their difference in magnitudes; and the important elevation cues, which are generated from the sound's reflection, dispersion and diffraction on head, shoulders and outer ear, are implied in the HRTFs'grotesque spectrum. Human auditory system utilizes the clues in HRTFs and compares them with the past auditory experience in brain memory to discern sound source's position. So far, people have constructed different mathematical models to simulate this delicate function, and try to give a reasonable interpretation for this phenomenon. From this, we can see that, in the research of spatial hearing and implementation of virtual auditory space, it is important to accurately model the latent acoustical clues in HRTFs or HRIRs related to certain position of sound source. And the dissertation just makes in-depth investigations on this problem.First of all, the chapter one summarizes the basic status and situation of spatial hearing in virtual reality research, and looks back to the pertinent psychoacoustical experiments and their conclusions in this area. Then based on the analysis of this subject, some primary research directions are provided. In chapter two, HRTFs'data packages used in this dissertation, including KEMAR and CIPIC, are detailedly introduced, especially their special measurement methods and anthropometric parameters. Meanwhile, some basic correlated characteristics among relevant HRTFs are also shown and analyzed in this part. And in chapter three, through statistical analysis, some qualitative judgment and quantificational estimation between anthropometric parameters and HRTFs'clues, such as ITDmax and the HRTFs'spectrum, are approximately made. In chapter four, based on the judgement of HRTFs'different importance for sound source's recognition in different spectrum spans, and associated with some perspectives of psychoacoustics in spatial hearing, improved genetic algorithm (GA) is applied to the approximation of HRTFs'zero-pole model and common-acoustical-pole/zero model. In the following chapter five and six, aiming at HRIRs'singular characteristics and their distributing characteristic for all the measured HRIRs in time domain, adaptive non-linear approximation algorithm in the field of wavelet transformation, also including wavelet packet, local cosine packet and bases pursuit algorithm, are proposed, whose results show better performance than traditional HRIRs'PCA approximation model. At last, in the chapter seven, as a conclusion, several shortcomings of VAS's realization for actual application are summarily pointed out; and some probably improved approaches are also brought forward for future study.