Spectral elevation cues predict human sound-localization performance via machine learning

Evidence

In particular, an accurate prediction can be achieved using artificial neural networks trained with large data sets from a single subject (Jin et al., 2000). Recently, a functional model that predicts listeners’ performance in such localization tasks has been proposed by Baumgartner et al. (2013). This model computes the probability of perceiving a sound coming from a certain elevation angle by comparing the internal representation of the sound as processed by the peripheral auditory system with an internal representation of the HRTF. — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 1687

Categorization

Category: Connection Points

• XREF: Connects machine-learning approaches to binaural hearing — a useful bridge toward the reader’s interest in acoustic communication applications like hearing aids and VR audio.

Related

• Neural networks trained on one subject’s data can predict human sound-elevation localization (similarity 0.90)
• Neural networks trained on single-subject data can accurately model human elevation localization (similarity 0.88)
• Elevation localization relies on pinnae spectral cues encoded as HRTFs (similarity 0.84)
• Elevation localization is a timbral recognition task driven by resonant peaks (similarity 0.84)
• Pinnae shape the sound spectrum to encode elevation and front-back position (similarity 0.82)
• Horizontal localization uses binaural cues while vertical localization does not (similarity 0.82)
• Sound localization by elevation is guided by plausibility, not just acoustics (similarity 0.81)
• Auditory elevation calibration requires visual confirmation of sound sources (similarity 0.81)
• Sound source location is encoded in waveform distortions between the ears (similarity 0.80)
• The neural mechanism extracting perceived distance from room response remains unknown (similarity 0.80)
• Ear separation on one axis makes horizontal binaural cues strongest (similarity 0.80)
• Large diverse datasets of human HRTF and ear geometry are now being collected (similarity 0.80)
• Spatial hearing is calibrated through learning to each listener’s head and pinna geometry (similarity 0.79)
• Auditory spatial perception emerges from weighted combination of localization cues (similarity 0.79)
• Elevation localization rests on spectral recognition of familiar sounds, not spatial cues. (similarity 0.79)

Broader Connections

• Sound localization from pinna cues is learned and specific to each person’s ear shape. #expands (0.82) — David M. Howard, Jamie Angus, Acoustics and Psychoacoustics, 732
• Localization improves as we learn our own ears’ direction cues through experience #expands (0.75) — David Howard, Jamie Angus, coustics and Psychoacoustics, 729