Dataset Augmentation and Dimensionality Reduction of Pinna-Related Transfer Functions

Augmentation de données et réduction de dimensionnalité de fonctions de transfert relatives à l'oreille (PRTFs)

Corentin Guezenoc (1, 2) , Renaud Seguier (2, 1)

1 CentraleSupélec
2 IETR - Institut d'Électronique et des Technologies du numéRique

Efficient modeling of the inter-individual variations of head-related transfer functions (HRTFs) is a key matter to the individualization of binaural synthesis. In previous work, we augmented a dataset of 119 pairs of ear shapes and pinna-related transfer functions (PRTFs), thus creating a wide dataset of 1005 ear shapes and PRTFs generated by random ear drawings (WiDESPREaD) and acoustical simulations. In this article, we investigate the dimensionality reduction capacity of two principal component analysis (PCA) models of magnitude PRTFs, trained on WiDESPREaD and on the original dataset, respectively. We find that the model trained on the WiDESPREaD dataset performs best, regardless of the number of retained principal components.

Domaines: Synthèse d'image et réalité virtuelle [cs.GR]; Son [cs.SD]; Modélisation et simulation

Corentin Guezenoc, Renaud Seguier. Dataset Augmentation and Dimensionality Reduction of Pinna-Related Transfer Functions. Audio Engineering Society Convention, May 2020, Vienna, Austria. ⟨10.17743/aesconv.2020.978-1-942220-32-9⟩. ⟨hal-02484088⟩