Recent developments in AR / VR applications have brought a renewed focus on efficient and scalable real-time HRTF renderers to alleviate compute constraints when spatializing many sound sources at once. To efficiently achieve a reasonable approximation of the full-sphere, the HRTF dataset is often linearly decomposed into a predetermined number of basis filters via methods such as Ambisonics, VBAP, or PCA. This paper proposes a novel HRTF renderer and decomposition technique that, when compared to previous methods, allows for greater accuracy of the HRTF approximation for an equivalent compute cost. This is achieved through a multi-layered optimization network architecture that minimizes a perceptually motivated error function to derive the basis filters. We will demonstrate the numerical accuracy of our technique as well as provide listening test results comparing our method to other linear decomposition methods of relative computational cost using both our internal and the publicly available SADIE HRTF datasets.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
