Research code

Repositories that reproduce the analyses and figures of my papers and workshop contributions. Each one is frozen at the state used for the publication, with the environment needed to run it.

bayesian_listener_notebooks

License Last commit

Notebooks, figure scripts and cached results for R. Barumerli, F. Brinkmann, E. Zanoni, A. Hoyer, L. Picinali and M. Geronazzo, Statistical validation and full-sphere extension of a Bayesian model for human static sound localisation, Acta Acustica 10, 89 (2026). One script regenerates every figure of the paper from cached results, and a dedicated notebook reproduces all the reported statistics. The model itself is the Bayesian Listener package.

fa2026-localisation-mixture

License Last commit

Companion code for J. Sztandera, L. Picinali and R. Barumerli, A Probabilistic Mixture Model for Evaluating Sound Localisation Performance in the Median Plane (Forum Acusticum 2026, Graz). The classical median-plane metrics (polar error, quadrant error and polar gain) each rely on a hard threshold, so none of them reads out a single property of the listener. The mixture model fits the whole polar response distribution instead and recovers front-back confusions, response scatter and the pull towards the horizon as separate parameters.

iwai2026-fep-spatial-hearing

License Last commit

Code for the extended abstract Framing Spatial Hearing within the Free Energy Principle: A Hierarchical Model of Perceptual Learning (Barumerli, Meyer-Kahlen and Picinali, International Workshop on Active Inference 2026). Sound-localisation learning with visual feedback is modelled as trial-by-trial belief revision: lateral precision, polar precision and hemifield disambiguation are updated by natural-gradient descent and compared with the manual-pointing data of Majdak, Goupell and Laback (2010).

  • Source code on GitHub, including the accepted extended abstract (PDF)

telluride2026-robot-awareness

Last commit

A collaborative project from the Telluride Neuromorphic Workshop 2026 on audio-based robot awareness of peripersonal space. A real-time pipeline detects an approaching drone from its acoustic signature, estimates whether it is looming or receding, and instructs a Boston Dynamics Spot robot to duck as a simulated collision-avoidance reflex.