HRTF Personalisation from Head Meshes
Neural HRTF models using geometric features from 3D head meshes.
At the Speech and Hearing Research Center, Peking University, I worked with Tianshu Qu on HRTF personalisation using 3D head meshes. The approach maps PCA-based geometric features to spatial principal component representations of HRTFs with deep neural networks.
The work involved processing three datasets with more than 400 subjects and evaluating the models against anthropometric feature baselines.