Minimum-Phase HRTF Modeling of Pinna Spectral Notches using Group Delay Decomposition

Sandeep Reddy C,Rajesh M Hegde

DOI: https://doi.org/10.48550/arXiv.1711.01872

2017-11-06

Audio and Speech Processing

Abstract:Accurate reconstruction of HRTFs is important in the development of high quality binaural sound synthesis systems. Conventionally, minimum phase HRTF model development for reconstruction of HRTFs has been limited to minimum phase-pure delay models which ignore the all pass component of the HRTF. In this paper, a novel method for minimum phase HRTF modelling of Pinna Spectral Notches (PSNs) using group delay decomposition is proposed. The proposed model captures the PSNs contributed by both the minimum phase and all pass component of HRTF thus facilitating an accurate reconstruction of HRTFs. The purely minimum phase HRTF components and their corresponding spatial angles are first identified using Fourier Bessel Series method that ensures a continuous evolution of the PSNs. The minimum phase-pure delay model is used to reconstruct HRTF for these spatial angles. Subsequently, the spatial angles which require both the minimum phase and all pass components are modelled using an all-pass filter cascaded with minimum-phase pure-delay model. Performance of the proposed model is evaluated by conducting experiments on PSN extraction, cross coherence analysis, and binaural synthesis. Both objective and subjective evaluation results are used to indicate the significance of the proposed model in binaural sound synthesis.