Research On Platform For Virtual Auditory Environment Real Time Rendering System

Virtual auditory or acoustic environment (VAE) recreates auditory perceptions or eventsas those would happen in real world by controlling acoustic environment artificially.Dynamically synthesizing binaural signals and reproducing them via headphones orloudspeakers is the most conventional way. VAE systems have been applied in various fields,such as the research of binaural hearing, multimedia and virtual reality, among others.A few systems have been set up in the world. However, limited system resource requiressome tradeoffs between performance and computational cost. Improving the performance ofVAE is still challenging. Moreover, no report was found on the hardware/software platformfor dynamic VAE in China. Thus, this dissertation carried out the investigations in thefollowing aspects:The first work is improving the algorithms in VAE. Here the far-field principalcomponents analysis (PCA) decomposition is extended to distance-dependent near-fieldHRIRs. An algorithm for efficiently synthesizing multiple virtual sources based on PCA isproposed, based on which the block diagram of the signal processing is designed. Comparedwith the conventional scheme, the computational cost is greatly reduced when the number ofvirtual sources is large. Moreover, HRIR interpolation, moving virtual source, and dynamicvirtual sources synthesis are realized by adjusting PCA-weights, time delay, and the scalingfactor1/r in PCA-based virtual source synthesis. These implementations are simpler and moreefficient than the conventional one, meanwhile, without audible artifacts.The second work is setting up the hardware/software system of VAE. To meet theperformance requirements of VAE system, a low-latency ASIO sound card, a high updaterate head tracker with six degrees of freedom, a high clock speed (2.83GHz)computer and aprofessional monitoring headphone are used. A software written in C++language performs thefunctions of human-machine interface such as data exchange with the head tracker anddynamic VAE processing. In order to use the system conveniently as a platform for futurebinaural hearing researches, several parameter interfaces are designed including HRTFparameters (both the conventional and PCA-based schemes), dynamic parameters,environment information and sound stimuli.The third work is testing three important parameters for dynamic performance of the VAEsystem, including the scenario update rate, system latency time, and the maximal number ofsimultaneous virtual sources (including direct and image sources for reflections). The meanlatency time is25.4ms measured by a simple self-designed apparatus. The scenario update rate is120Hz measured using both the high-precision timer function of Visual C++andcomputer serial port. The computation tasks are allocated to each CPU core using multithreadtechnology, and then the maximal number of virtual sources is evaluated base on the use rateof CPU(about90%). The current system exhibits following performance,(1) For theconventional scheme, it is able to synthesize up to280free-field virtual sourcessimultaneously.(2) For the PCA-based scheme, it is able to synthesize up to4500free-fieldvirtual sources simultaneously.The fourth work is validating the performance of our VAE system by psychoacousticexperiments. First, a method for reporting the direction and distance of the perceived virtualsource position using a tracker is proposed. The PC received and saved the direction anddistance data of a virtual source when the listener points the tracker to the perceived locationof virtual source. The proposed method demonstrates good repeatability as well as highaccuracy with less variance. Using the proposed reporting method, a mean judgment responsetime decreases from19.8s to7.2s. Compared with the conventional oral reporting method, theproposed method is simpler, more accurate and efficient. Psychoacoustic experiments areconducted to validate the performance of system. Both far-field and near-field HRIRs forKEMAR were used. Results show that, dynamic rendering of VAE, even withnon-individualized far-field HRIRs (HRTFs), can improve performance of VAE, such asalmost completely resolving the front-back confusion, reducing the percentage of up-downconfusion to some extent, and completely diminishing the inside-head effect. Moreover, theconventional and PCA-based schemes yield equivalent perceptual effects in localization.Near-field HRIR processing is also capable of controlling perceived virtual source distance tosome extent in a dynamic VAE.The fifth work is the application of VAE system. The contribution of dynamic cues,individual spectral cues and signal bandwidth on the location of a virtual sound source isstudied. The results show that dynamic cues are very important in location, enabling reducingthe front-back and up-down confusion rates significantly, completely diminishing theinside-head effect, and improving localization accuracy as well as externalization. Individualspectral cues greatly contributes to reducing up-down and front-back confusion rates,improving directional location accuracy, while without effect on distance perception. Signalbandwidth is closely related to high frequency spectral cues, and can reduce the up-downconfusion rate significantly as well as improve localization accuracy with broadband stimuli.However, the contribution of high frequency spectral cues to reducing front-back confusionrate will be masked by dynamic cues. These findings in the dissertation not only provide a platform for future binaural hearingresearches, but also supply foundation for the applications of VAE.