In this study we propose a novel approach for the latent representation of the Head Related Transfer Functions (HRTFs), which are understood to be dominant cues for the perception of sound localization. HRTFs represent the acoustic transformations that a sound signal undergoes between its point of origin in space and its receiving point inside the ear, due to interactions with the human outer body structures such as the torso, head and the pinna, thus making the HRTFs a function of both the subjects anthropometric features as well as of the source locations, for a given individual subject. In this work we propose to derive latent representations of the HRTFs by disentangling subject-specific features from the location-specific features. We posit that such latent representations hold promise for personalized HRTF predictions in various downstream applications. We validate the disentanglement of the features through visualizations, comparing the disentangled latent representations with those without the disentan-glement. Additionally, we also conduct a comprehensive analysis to understand the impact of disentanglement on the reconstructions.
(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.
