Individualization of Dynamic Binaural Synthesis by Real Time Manipulation of ITD

Virtual acoustic environments (VAEs) are commonly realized via dynamic binaural synthesis. Therefore, anechoic audio is convolved in real time with head related impulse responses (HRIRs) or binaural room impulse responses (BRIRs). Binaural filters are exchanged inaudibly and in real time according to the listener’s head movements, thus providing a realistic auditory experience. From Lord Rayleighs [1] duplex theory of hearing it is known, that interaural level (ILD) and time (ITD) differences are exploited for auditory localization. ITDs are being evaluated mainly below 1500 Hz, whereas above, envelope delays and ILDs determine localization. If the signal contains ambiguous temporal and spectral localization cues, the ITD tends to dominate the perceived sound source direction [2]. As it is mostly unfeasible to conduct individual binaural measurements, binaural data sets of dummy heads are commonly used. ILDs and ITDs of the binaural data sets therefore reflect the physical structure of a certain individual’s pinna, head and torso. ITDs differ individually due to varying head geometries. As a result, when listening to a VAE with non-individualized binaural data sets, degradation of localization accuracy and the perceived stability of the sound sources occur: If the head in the original data sets was smaller than the listeners head, a displacement of the sound sources in the same direction as the heads’ movement may be perceived during head movements; if on the other hand, the head was larger, an apparent movement in the opposite direction occurs [3]. This effect can be very annoying, especially as adaptation does not seem to occur. Therefore, a method for the customization of binaural synthesis by means of real time manipulation of the ITD contained within binaural impulse response data sets is being proposed. Additionally, based on the listener’s individual anthropometry and listening tests, a predictive model for adequate individual manipulation will be presented.