Extracting significant features from the hrtf

TitleExtracting significant features from the hrtf
Publication TypeJournal Articles
Year of Publication2003
AuthorsRaykar VC, Duraiswami R, Davis LS, Yegnanarayana B
JournalProceedings of the 2003 International Conference on Auditory Display
Date Published2003///

The Head Related Transfer Function (HRTF) characterizes the au-ditory cues created by scattering of sound off a person’s anatomy.
While it is known that features in the HRTF can be associated with
various phenomena, such as head diffraction, head and torso re-
flection, knee reflection and pinna resonances and anti resonances,
identification of these phenomena are usually qualitative and/or
heuristic. The objective of this paper is to attempt to decompose
the HRTF and extract significant features that are perceptually im-
portant for source localization. Some of the significant features
that have been identified are the pinna resonances and the notches
in the spectrum caused by various parts of the body. We develop
signal processing algorithms to decompose the HRTF into compo-
nents, and extract the features corresponding to each component.
The support of NSF award ITR-0086075 is gratefully acknowl-