Immersive Formats Evaluation

Reproducible analysis repository (with Danowski) for a cross-format objective evaluation of immersive audio quality metrics. Two binaural models, BAM-Q (perceptual quality) and BINAQUAL (localization similarity), are applied to 11 format variants derived from three professional recordings differing in genre and Ambisonic provenance, rendered under three reference anchors with a controlled test of stimulus alignment. Supplementary materials on Zenodo document the full render chain: REAPER and Pro Tools sessions, plugin presets for Dolby Atmos, Auro-3D, Sony 360 Reality Audio, and IAMF/Eclipsa, and the IEM loudspeaker layouts behind the 42pIKO variants.

Two properties of the evaluation setup affect the outcome more than most of the format differences themselves. Immersive renderers impose transport delays of up to 510 ms that are not quality degradations; left uncompensated, they make BINAQUAL – which compares phase representations – track delay rather than fidelity (ρ = -0.86), and correcting them shifts localization similarity by a median of 149.3%, while the modulation-domain BAM-Q metric is essentially unaffected (median 0.0%). Separately, under BINAQUAL the apparent advantage of one format family over another follows the reference anchor’s own family, reversing in every tested item when a channel-based anchor replaces a scene-based one – the BAM-Q composite never reverses. With aligned stimuli the two models are moderately correlated (r = 0.51) and format orderings are reproducible across content (Kendall’s W = 0.94, permutation p < 0.001). Companion paper submitted to Applied Acoustics (2026).

@article{mroz_latency_2026,
  title = {Latency and anchor dependence in objective binaural quality metrics: a cross-format evaluation of immersive audio},
  author = {Mróz, Bartłomiej and Danowski, Przemysław},
  journal = {Applied Acoustics},
  year = {2026},
  note = {Under review}
}