Using reverse correlation to derive spectral weights for median plane localisation

Pedro Llado Gonzalez*, Petteri Hyvärinen, Taeho Kim, Ville Pulkki

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionScientificpeer-review


The individual internal representation of the spectral cues that allow distinguishing between front and back is investigated by using the reverse correlation method. The stimuli were noise bursts presented randomly from either a front or back loudspeaker. For each trial, the spectrum of the noise was modified with a 1 ERB spaced gammatone filterbank with random attenuation applied for each band separately. Analysing the link between responses and spectral attenuation profiles with the reverse correlation approach results in a data-driven weighting of spectral cues that each subject relies on to distinguish front and back. The spectral localisation cue decoding was based on the positive spectral gradient method, which was motivated by neurophysiological findings. The results suggest that different listeners use different frequency regions to
distinguish between front and back. The spectral weights derived from the perceptual test were different than the ones derived from objective measurements alone, which may confirm the importance of the non-acoustic factors in localisation.
Original languageEnglish
Title of host publicationICA 2022 proceedings
PublisherAcoustical Society of Korea (ASK)
Number of pages8
Publication statusPublished - 2022
MoE publication typeA4 Article in a conference publication
EventInternational Congress on Acoustics - Hwabaek International Convention Center (HICO), Gyeongju, Korea, Republic of
Duration: 24 Oct 202228 Oct 2022
Conference number: 24

Publication series

NameProceedings of the ICA congress
ISSN (Electronic)2415-1599


ConferenceInternational Congress on Acoustics
Abbreviated titleICA2022
Country/TerritoryKorea, Republic of
Internet address


Dive into the research topics of 'Using reverse correlation to derive spectral weights for median plane localisation'. Together they form a unique fingerprint.

Cite this