Head-Related Transfer Function Interpolation with a Spherical CNN

TL;DR

This paper introduces a spherical CNN approach for interpolating HRTFs, leveraging spherical harmonics to better capture spatial features, resulting in more accurate high-resolution HRTF estimation from sparse data.

Contribution

It presents a novel spherical CNN method that uses spherical harmonics for effective HRTF interpolation, improving over existing deep learning and SH-based methods.

Findings

Outperforms existing SH and learning-based methods in interpolation accuracy

Effectively captures spatial acoustic features of HRTFs

Achieves accurate high-resolution HRTFs from sparse measurements

Abstract

Head-related transfer functions (HRTFs) are crucial for spatial soundfield reproduction in virtual reality applications. However, obtaining personalized, high-resolution HRTFs is a time-consuming and costly task. Recently, deep learning-based methods showed promise in interpolating high-resolution HRTFs from sparse measurements. Some of these methods treat HRTF interpolation as an image super-resolution task, which neglects spatial acoustic features. This paper proposes a spherical convolutional neural network method for HRTF interpolation. The proposed method realizes the convolution process by decomposing and reconstructing HRTF through the Spherical Harmonics (SHs). The SHs, an orthogonal function set defined on a sphere, allow the convolution layers to effectively capture the spatial features of HRTFs, which are sampled on a sphere. Simulation results demonstrate the effectiveness of the proposed method in achieving accurate interpolation from sparse measurements, outperforming the SH method and learning-based methods.