Speaker
Description
Head-related transfer functions (HRTFs) describe how the human head, torso, and pinnae filter sound arriving from different directions. They are unique to each listener and are an important part of binaural rendering, spatial audio, and virtual reality systems. HRTF research involves not only acoustic analysis but also the management of Spatially Oriented Format for Acoustics (SOFA) files, public datasets, and reproducible data pipelines for analysis and deep learning model training. In practice, this work often relies on custom scripts combined with existing software packages, resulting in heterogeneous workflows that make experiments harder to reproduce or extend. These challenges motivated the development of hrtfpykit, a Python library for HRTF research workflows. It includes a SOFA layer for reading, writing, and editing SOFA files. The hrtf layer represents HRTF data as a unified object containing synchronized impulse response and transfer-function representations and supports acoustic transformations and spatial and spectral selection. The plots layer provides visualization tools for HRTF inspection and comparison in the time and frequency domains. Finally, the datasets layer standardizes access to public HRTF datasets through declarative specifications and configurable transformations, enabling reproducible pipelines for training deep learning models, particularly for HRTF individualization.