PaperlayerView exact revision on arXiv

Source-linked AI summary

Beyond Localisation Accuracy: Sensorimotor Effects of HRTF Individualisation

Fulvio Missoni, Katarina C. Poole, Tim Murray-Browne, Andrea Canessa, Lorenzo Picinali

arXiv:2609.08422v1eess.AS

TL;DR

Static localisation tasks isolate hearing from natural multimodal action, while individual-level HRTF outcomes are unreliable. This study tested individualisation in aurally guided visual search and found faster anechoic responses, with the advantage absent under reverberation.

  • Problem

    Static localisation isolates hearing from natural multimodal action, and individual-level behavioural outcomes are unreliable, leaving limited evidence about HRTF individualisation in sensorimotor listening.

    Source: Introduction · pp. 1–2

  • Method

    Listeners searched for a visual target guided by a co-located sound while moving freely, comparing individual and non-individual HRTFs in anechoic and reverberant conditions.

    Source: Introduction · p. 2, Methods · p. 4

  • Results

    ∆RT = 200 ms: individualised HRTFs produced faster anechoic responses than non-individualised HRTFs, particularly for front–back stimuli; the advantage disappeared under reverberation.

    Source: Discussion and conclusions · pp. 10, 12

  • Takeaways & Limitations

    Dynamic listening tests should assess movement-related measures alongside localisation metrics because individualisation affected orienting behaviour even when later task outcomes remained similar.

    Source: Discussion and conclusions · p. 14

    Real-time 6DoF rendering required HRTF interpolation, which may have introduced magnitude errors and weakened the individualisation contrast.

    Source: Discussion and conclusions · p. 13

Abstract

from arXiv · show

Everyday listening requires the brain to integrate cues from the body, environment, other senses, and movement, continuously translating auditory information into action. Yet HRTF individualisation is still commonly assessed through localisation accuracy, which may not fully capture its effects on this sensorimotor process. Here, we investigate whether these effects can instead be revealed through behaviour in a more ecologically valid listening task. We used an aurally guided visual search paradigm in which listeners located a visual target using a co-located virtual sound while moving freely, comparing individualised and non-individualised HRTFs under anechoic and reverberant conditions. Performance was assessed through response times and measures of movement organisation. In anechoic conditions, individualised HRTFs produced faster responses than non-individualised HRTFs, with an average reduction of approximately 200ms and the clearest benefit for front-back source locations. This advantage was expressed primarily in movement initiation, whereas overall movement extent was only weakly affected. Under reverberant conditions, HRTF-dependent differences disappeared. These results suggest that HRTF individualisation can influence how listeners plan and initiate orienting actions even when differences in conventional localisation outcomes are limited. Assessing sensorimotor behaviour alongside localisation performance may therefore provide a more sensitive and ecologically relevant account of the perceptual benefits of HRTF individualisation.

Introduction

Static localisation established the HRTF’s role in spatial hearing, but individual-level outcomes are unreliable and findings from more realistic listening tasks have been inconsistent. This study uses aurally guided visual search to test whether HRTF individualisation affects response times and movement strategies across acoustic environments.

Introduction

  • Individual-level outcomes in static localisation are unreliable because non-acoustic factors explain more outcome variance than the HRTF itself.Static, blindfolded tasks have served as the primary benchmark and confirmed a group-level benefit.Source: Introduction · p. 1
  • More realistic listening studies have reported inconsistent HRTF benefits across free movement, speech-in-noise, and perceptual-plausibility tasks.Benefits have depended on restricted movement or limits on non-spatial cues; plausibility ratings have shown no significant difference between HRTFs.Source: Introduction · p. 2
  • Auditory cue changes may alter how listeners map perceived sound locations to actions without changing localisation accuracy.Movement measures may therefore reveal adaptation that accuracy alone does not capture.Source: Introduction · p. 2
  • Aurally guided visual search measures natural orienting as listeners locate a visual target using a co-located sound.The paradigm has linked longer search times and altered movement strategies to front–back confusion and degraded cues from hearing protectors.Source: Introduction · p. 2
  • The study tests whether individual HRTFs benefit performance in anechoic conditions, alter movement planning, and lose their advantage with early reflections.The authors hypothesise that room acoustics reduce spectral detail on which individualisation relies.Source: Introduction · pp. 2–3

Methods

The study compared individual and KEMAR HRTFs in anechoic and reverberant conditions with 12 participants performing a visually guided sound-localisation task in VR. Response and movement measures were extracted from task performance and head tracking, then analysed with mixed-effects models.

Participants and conditions

  • 12 participants with self-reported normal hearing completed individualised and non-individualised HRTF conditions, each in anechoic and small-room reverberation.The reverberant room measured 3 m × 4 m × 2.5 m.Source: Methods · p. 3

Equipment and setup

  • The VR experiment used over-ear Sennheiser HD599SE headphones, a Meta Quest 3 head-mounted display, two handheld Meta controllers, and real-time inside-out six-DoF tracking at 72 Hz.Acoustic scenes were rendered in real time using the Unity Wrapper module of the 3DTune-In Toolkit, using HMD head-tracking data.Source: Methods · p. 3

Stimuli

  • Listeners heard 12-second modulated white-noise bursts rendered with their own HRTF or a KEMAR HRTF, with direct sound alone or a reverberant virtual-loudspeaker rendering.The reverberant condition used impulse responses from a small trapezoidal meeting room; direct and reverberant sound paths had separate distance attenuation.Source: Methods · pp. 3–4

Experimental procedure

  • Participants reported the LED count on the sound’s target loudspeaker while searching among 28 virtual loudspeakers, completing 336 trials across four conditions.They received brief familiarisation with feedback using the individual HRTF in anechoic conditions, with selected positions presented once each.Source: Methods · p. 4

Behavioural and movement measures

  • Performance measures included correct responses and response time, while head rotations yielded movement-onset latency and total angular extent.Movement onset used yaw velocity; angular extent summed distances between local extrema, and frontal horizontal trials were excluded. Response-time estimation excluded some return movements using a 5%-of-peak-velocity threshold.Source: Methods · pp. 4–5

Behavioural data analysis

  • Generalized linear mixed models assessed response time, movement onset, and tilt extent; yaw extent used a linear mixed model.Models included HRTF, reverberation, and direction effects, with post hoc interaction contrasts adjusted using Tukey’s method.Source: Methods · pp. 5–6

Results

Individual HRTFs shortened response times in anechoic conditions, but not in reverberation. HRTF-related differences also appeared in direction-specific response times and movement measures, despite high response accuracy.

Response accuracy

  • Responses were 98.54% correct on average, with no significant effects of tested condition or stimulus location on accuracy.Accuracy variability across participants was low (SD = 1.24%).Source: Results · p. 6

Analysis of head-rotation behaviour

  • Movement-onset latency models had limited explanatory power, with fixed effects explaining marginal R2 = 0.087; those results were reported only in the Supplementary Information.The model also had a moderate within-group correlation (ICC = 0.36), consistent with large individual variability in movement onset.Source: Results · p. 6

Response time analysis

  • In anechoic conditions, response times were 0.67 log(s) with individual HRTFs versus 0.76 log(s) with non-individual HRTFs; no significant HRTF difference appeared in reverberation.The anechoic comparison was significant (p = 0.0002), while the reverberant comparison was not (p = 0.456).Source: Results · p. 6, Figure 1 · p. 8
  • In anechoic conditions, HRTF-related response-time differences were significant for back and front directions, but not lateral directions.The HRTF-by-direction interaction was significant (χ2(1) = 7.47, p = 0.03).Source: Results · p. 6, Figure 2 · p. 9

Analysis of head-rotation behaviour

  • Yaw extent showed no overall HRTF or acoustic-condition effect, although frontal trials had increased yaw with KEMAR; tilt extent showed a small individual-HRTF increase in reverberation.The direction-dependent yaw interaction was significant (p = 0.03); tilt showed significant HRTF, acoustic-condition, and interaction effects.Source: Results · p. 7, Figure 4 · p. 10
  • Movement direction at onset was more often correct with individual HRTFs, while acoustic condition and its interaction with HRTF had no significant effects.The HRTF effect was significant (χ2(1) = 5.9363, p = 0.01483; partial η = 0.15).Source: Results · p. 8, Figure 3 · p. 9
  • Head positioning was accurate on average but imprecise, consistent with participants using head movements for coarse repositioning and compensatory eye movements for final target selection.Azimuth and elevation variability were SD = 15.58° and SD = 11.98°, respectively.Source: Results · pp. 8–10

Discussion and conclusions

In the aurally guided visual-search task, individualised HRTFs improved response time in anechoic conditions, chiefly during movement initiation and particularly for front–back stimuli; this advantage disappeared under reverberation. The discussion interprets the pattern as context-dependent sensorimotor organisation and identifies rendering and movement constraints that qualify the findings.

Discussion and conclusions

  • The study treats localisation as a sensorimotor process in which auditory spatial estimates are translated into orienting actions, rather than solely as a perceptual outcome.The authors interpret the benefit as most apparent when individualised spectral cues remain informative and behaviourally relevant.Source: Discussion and conclusions · pp. 10, 12
  • A 200 ms response-time reduction with individualised versus non-individualised HRTFs occurred in anechoic conditions, especially for front–back stimuli.The advantage was modest, expressed primarily in movement initiation, and overall movement extent was only slightly affected.Source: Discussion and conclusions · pp. 10, 12
  • The study’s median response time was 1.98 s (SD = 1.04 s), comparable to prior VR binaural paradigms but longer than free-field reports.The authors relate the difference to the added demands of VR and binaural reproduction and note that they omitted training to assess relatively unadapted performance.Source: Discussion and conclusions · pp. 10–12
  • The anechoic advantage was concentrated in initial orienting, while dynamic sampling and visual feedback may reduce early differences as movement unfolds.The authors suggest non-individualised HRTFs can increase initial front–back ambiguity, delaying action onset, while later outcomes become more similar.Source: Discussion and conclusions · pp. 10, 12
  • Under reverberation, the individualisation advantage disappeared, possibly because early reflections and temporal smearing reduced the salience of spectral cues.The authors note that the reverberant condition individualized only the direct sound path, while the reflected field used a non-individualised BRIR.Source: Discussion and conclusions · pp. 10, 12, 13
  • Response time varied with both HRTF and reverberation, whereas movement planning was sensitive to HRTF but not reverberation; the authors present this as a tentative dissociation.They state that the data do not establish this interpretation conclusively and call for targeted investigation.Source: Discussion and conclusions · p. 13
  • Real-time HRTF interpolation and non-individualised rendering of reflected sound may have attenuated the observed effects, while restricted movement may have limited active exploration.The authors propose more freedom of movement and fully individualised rendering of direct and reflected sound as future directions.Source: Discussion and conclusions · pp. 13–14
  • The authors propose pairing movement-related measures with conventional localisation metrics and using eye tracking to examine gaze–head coordination.They also recommend richer acoustic interaction and greater movement freedom while maintaining experimental control.Source: Discussion and conclusions · pp. 13–14

Author information

The listed authors are affiliated with the University of Genoa and Imperial College London. The contribution statement assigns experimental design, implementation, data collection, analysis, interpretation, and manuscript work across the team; the authors declare no competing interests.

Authors and affiliations

  • Fulvio Missoni, Katarina C. Poole, Tim Murray-Browne, Andrea Canessa, and Lorenzo Picinali are affiliated with the University of Genoa and Imperial College London.Missoni and Canessa are listed with Genoa; Poole, Murray-Browne, and Picinali with Imperial College London.Source: Authors and affiliations · p. 17

Contributions

  • K.C.P. designed the paradigm; T.M.B. contributed to design and led implementation and software, while F.M. handled measurements, data acquisition, analysis, interpretation, and drafting.L.P. and A.C. contributed to discussion and interpretation; L.P. also critically revised the initial manuscript, and A.C. reviewed it.Source: Contributions · pp. 17–18

Competing interests

Loading 2609.08422v1…