Anthropometric tuning of a spherical head model for binaural virtual acoustics based on interaural level differences

In this paper we propose a method to customize a spherical head model for binaural sound rendering based on the listener's anthropometry. Interaural level difference (ILD) information from a HRTF database is used to subjectively tune the radius parameter of the spherical model so as to best fit individual measures. Multiple linear regression on anthropometric data is performed, yielding a closed formula relating the three head dimensions to the ILD-optimized radius. The effectiveness of the proposed radius estimation method in predicting the correct ILD with a spherical model is compared to that of alternative methods from the literature. Results show that the average spectral distortion between experimental and predicted ILDs with our method is significantly lower than with other estimation methods for lateral source locations. The proposed customization approach provides substance towards the development and evaluation of personal auditory displays for binaural virtual acoustics.