BezierPPM: a new parametric model of the human pinna

The external part of the human ear – the pinna – has a unique and intricate shape that varies significantly from person to person, making it difficult to accurately determine and recreate its individual properties. 

New SONICOM research introduces BezierPPM – a new parametric model of the pinna based on cubic Bézier curves and concave deformations. This parametric pinna model (PPM) represents the biological structure of the human pinna and enables its parametric modelling in many degrees of freedom. 

“In contrast to other PPMs, BezierPPM is inspired by the biological structure of the human pinna aiming at providing parameters directly related to the complex pinna geometry,” says Felix Perfler, PhD student at the Acoustics Research Institut of the Austrian Academy of Sciences and first author of the publication. “Our analysis shows that BezierPPM can accurately represent human pinnae, meaning that it has high potential for a range of applications, including biometric identification, the design of ear prostheses, and the personalisation of spatial hearing.” 

Capturing complex geometry 

Accurately capturing the complex morphology of the human pinna can prove challenging. Direct optical capture of pinna geometry potentially offers the required accuracy, but requires special equipment such as laser scanners and the result must be post-processed to remove artifacts resulting from the scanning process. This post-processing is usually performed manually and requires a considerable time and expertise. 

By representing the pinna geometry parametrically, this process can be simplified: parameters of parametric pinna models can be modified such that the modelled mesh matches the target pinna mesh, a process referred to as registration. 

BezierPPM offers a PPM that uses parameters closely linked to human ear structures, representing the ear geometry as cubic Beziér curves and including local modifiers of predefined concave areas. 

BezierPPM was evaluated by manually registering its parameters to 20 ears selected from various databases of digitized ears. Analysis shows that BezierPPM was capable of representing human pinnae very well, and that most of the inaccuracies were located on the back side of the ear, an area having a rather low relevance for most target applications. Further, BezierPPM was tested in a machine-learning setting to show that it can be applied in future AI-based algorithms. 

While the BezierPPM is ready to be used, further evaluation will be required for specific use cases such as the customization of ear prostheses or HRTF personalization. In particular, for applications in acoustics and binaural audio, evaluations in the psychoacoustic domain in terms of various spatial-hearing performance metrics when listening with numerically calculated HRTFs obtained from BezierPPM-based meshes is necessary. 

Read the full article to find out more, or access BezierPPM directly on GitHub.