Matt Montag
HRTF: The Head-Related Transfer Function
What Works
- Successfully exploits the directional sensitivity of human hearing
- Simulate point sound sources in free space
|
What Doesn't Work
- Listener has to keep his/her head still or use head tracking to maintain the illusion of a fixed directional source
- Ideal reproduction must use in-ear headphones and account for transducer characteristics
- HRTF filtering alone produces a dry sound lacking "presence" because it ignores room acoustics
|
Audio Examples
original: the unfiltered 2-channel source audio; Rachmaninoff Symphonic Dances No. 1.
CIPIC: filtered with CIPIC HRTF 137. Left channel mapped to virtual source at 0° elevation, -45° azimuth; right channel mapped to 0° elevation 45° azimuth.
OFFICE: filtered with University of Oldenburg binaural room impulse responses for Office I. Left channel mapped to virtual source at 0° elevation, -45° azimuth; right channel mapped to 0° elevation 45° azimuth, distance of 1 meter.
|
What is it?
Humans can locate sounds in three dimensions in three primary ways: the difference in pressure between the two ears (interaural level difference, ILD), the difference in arrival time between the two ears (interaural time difference, ITD), and because the odd shape of the ear causes sound to be filtered differently depending on which direction it comes from. The head and ears diffract and reflect sound in a unique and consistent way on its path to the eardrums. The brain uses this filtering profile to localize sounds. The HRTF or head-related transfer function is a digital model of that physical filtering process. HRTFs are unique to individuals, as they depend on physical anatomy; torso, head, and outer ear (pinna) shape. The HRTF varies with the Azimuth and Elevation of incoming sound.
HRTF data for a given subject can be collected in several ways:
- Place tiny microphones in the subject's ear canal as close to the eardrum as possible. Emit an impulse signal at various locations around the user's head, recording the response received at the microphones.
- Same as above, except instead of emitting an impulse, emit a sine sweep to model the ear's response to frequency, instead of recording the impulse response train. This will capture the same information in the frequency domain instead of the time domain.
- Same as above, except place a sound emitter inside the subject's ear, and mount several microphones around the user's head to enable simultaneous measurement at many locations at once.
What is it used for?
An audio signal can be filtered with a given HRTF at a given Azimuth and Elevation to simulate an arbitrary source location. Although HRTFs are different for everyone, it has been shown that humans are tolerant to generalization and smoothing of HRTF filtering profiles.
HRTF filters are ideally reproduced by in-ear headphones; with the transducers occupying the same location in playback as the microphones do in HRTF recording.
Essential Articles selected from 5390 articles
- HRTF measurements of a KEMAR dummy-head microphone by B. Gardner, K. Martin (1994)
- The CIPIC HRTF database by V.R. Algazi, R.O. Duda, D.M. Thompson IEEE Proceedings (2001)
- A model of head-related transfer functions based on principal components analysis by D.J. Kistler, F.L. Wightman - The Journal of the Acoustical Society of America (1992)
- Role of spectral detail in sound-source localization by A. Kulkarni, H.S. Colburn - Nature (1998)
Matt Montag 2010