Application Of Minimum-phase Approximation On The Signal Processing Of Virtual Auditory Display

Virtual auditory display (VAD) recreates special sound scenes by synthesizing binauralsignals and later rendering them via headphone or loudspeaker. VAD has been widely used inthe fields such as scientific research, multimedia, virtual reality and electronics products etc.As mentioned, VAD can be reproduced through headphone or loudspeaker. To eliminate theaffect on binaural signals introduced in the process of reproduction, equalization on thebinaural signals is necessary, namely, the binaural signals are equalized by headphone to earcanal transfer function(HpTF)(in headphone reproduction) or cross-talk cancellation andtimbre equalization function(CTC)(in loudspeaker reproduction) in frequency domain. Inpractice, to design more compact filter, simplification of HpTF and CTC is a key point. Theminimum-phase approximation is a common simplification method. This paper focus oninvestigating the reasonability of minimum phase approximation in virtual audio byheadphone and loudspeaker-based VAD through both theoretical analysis and psychoacousticexperiments.In this paper, HpTFs of3circumaual headphones and20human subjects were measuredwith20repetitions for each subject. Based on these data, minimum phase characteristics ofHpTF were analysized using correlation function. The results show that, in the overall audiblefrequency domain(0-20kHz), minimum phase approximation of HpTF may lead to errorsbecause the mean correlation coefficient is only0.85. However, in the frequency range of0~12kHz, the minimum phase approximation of HpTF is reasonable with the meancorrelation coefficient reaching0.95. The above conclusions are also verified throughsubjective psychoacoustic experiment. Ten subjects participated in the experiment, which were performed at seven horizontal directions. The result shows that: for these threecircumaural headphones, minimum phase approximation of HpTF in the frequency range of0-12kHz causes no audible change. Thus, the minimum phase approximation of HpTF below12kHz is reasonable, and the corresponding filter design can be simplified. Since the mainsound localization cues are concentrated below12kHz, the minimum phase approximationdoes not significantly affect reproduction performance.On the other hand, for the cross-talk and timbre equalization function inloudspeaker-based VAD, minimum phase approximation of the direction-independentcommon part was proposed. Corresponding results of subjective listening experiment showthat the minimum phase approximation causes no audible change in loudspeakerreproduction.In conclusions, introduction of minimum-phase approximation in the equalization ofheadphone or loudspeaker reproduction is reasonable and feasible under certain condition.This approximation is able to reduce the effective length of equlization function in favour offilter design and hardware realization.