Anthropometric Personalisation of Head-Related Impulse Responses: An application to the Scattering Delay Network and Higher Order Ambisonics
Author
Tissieres, Jason Yves
Term
4. Term
Education
Publication year
2018
Submitted on
2018-11-10
Pages
84
Abstract
Denne afhandling foreslår et lille sæt nøgleparametre for personlig og effektiv dynamisk binaural syntese til hovedtelefoner. Den præsenterer en billedstyret metode, der bruger to enkle fotos af hovedet til at personalisere den hovedrelaterede overføringsfunktion (HRTF) – den måde hoved og ører filtrerer lyd fra forskellige retninger. Metoden kombinerer en simpel sfærisk hovedmodel med øreplacering med HRTF-magnituder valgt fra en database ud fra antropometri (mål af hoved og ører). Rumsakustik forenkles med et scattering delay network (SDN), en beregningsmodel der gengiver tidlige (førsteordens) refleksioner med præcise rumlige ledetråde. Den foreslåede model sammenlignes med higher-order ambisonics (HOA), en udbredt 3D-lydteknik. I et virtual reality-lytteforsøg gav personaliserede HRTF'er bedre lokalisering og eksternalisering (oplevelsen af, at lyden er uden for hovedet) end generiske HRTF'er, og den tilpassede SDN-tilgang klarede sig godt i forhold til HOA.
This thesis proposes a small set of key parameters for personalized and effective dynamic binaural synthesis for headphones. It introduces an image-guided method that uses two simple head photos to personalize the head-related transfer function (HRTF)—the way the head and ears filter sounds from different directions. The method combines a simple spherical head model that includes ear position with HRTF magnitude data selected from a database to match a listener’s anthropometry (head and ear measurements). Room acoustics are simplified with a scattering delay network (SDN), a computational model that accurately spatializes early, first-order reflections. The proposed model is compared with higher-order ambisonics (HOA), a common 3D audio technique. In a virtual reality listening experiment, personalized HRTFs improved sound localization and externalization (hearing sounds as outside the head) over generic HRTFs, and the customized SDN approach performed favorably compared with HOA.
[This abstract was generated with the help of AI]
Keywords
Documents