Research & Projects
My research spans distributed audio signal processing and room acoustics, and I am transitioning into groove-based rhythm analysis. My current focus is the computational side of the GROOVE project at RITMO, University of Oslo.
GROOVE: Mapping, Modeling, and Perceiving the Combinatorics of Groove-Based Rhythms
Hosted at the RITMO Centre, University of Oslo and funded by the Norwegian Research Council, GROOVE (2026–2029) is a three-year project led by PI Guilherme Schmidt Câmara, with Anne Danielsen and Olivier Lartillot as supporting professors. It develops the first comprehensive framework for systematically identifying, formalizing, and perceptually validating groove archetypes. These archetypes are defined as the recurring multi-layered rhythmic combinations that characterize groove-based music across Afro-diasporic traditions such as funk, soul, reggae, samba, and hip-hop.
As Postdoctoral Research Fellow, I lead the computational component (WP2): developing a hybrid pipeline that combines traditional MIR methods with machine learning to perform audio source separation on music recordings, extract multi-dimensional features (onset timing, dynamics, pitch and tonal accents), and automatically categorize groove pattern combinations. I also contribute to the theoretical framework (WP1) and the perceptual evaluation (WP3). One of the main goals of the project is to produce an open-source toolbox for groove analysis, aimed at setting a new standard for reproducibility in groove studies within computational musicology.
Audio signal processing & acoustics
-
Distributed Multichannel Wiener Filter (dMWF)
Alternative to the DANSE framework for MWF-based distributed signal estimation, enabling distributed beamforming without depending on (possibly many) algorithmic iterations. Submitted to IEEE Transactions on Audio, Speech and Language Processing (2026).
-
Fast-converging topology-independent distributed adaptive signal estimation
Distributed audio signal estimation algorithms (based on the DANSE algorithm) for noise reduction in ad-hoc wireless microphone networks. The key contribution: addressing practical limitations of state-of-the-art algorithms in terms of convergence speed and robustness to changing network topology, so algorithms remain functional as nodes join, leave, or fail. Developed the TI-DANSE+ variant with convergence guarantees.
-
TI-dMWF box
A Python framework for reproducible simulations of distributed multi-microphone processing algorithms in configurable acoustic environments. Designed for exhaustivity and flexibility: swap acoustic scenarios, network topologies, or algorithm variants without restructuring the pipeline. Included: dMWF, TI-dMWF, DANSE, TI-DANSE, and TI-DANSE+.
-
Mitigating sampling rate offsets in distributed signal estimation
State-of-the-art distributed audio signal estimation algorithms assume perfect synchronicity between nodes, which rarely holds in practice due to imperfect crystal oscillators. The resulting sampling rate offsets (SROs) can significantly degrade algorithm performance. Developed an SRO estimation and compensation mechanism embedded in the DANSE framework, restoring robust operation under realistic asynchrony.
-
Reverberation room design optimization (ZAPPA project)
Developed a reverberation room design optimization framework using finite element method (FEM) acoustic simulations, as part of the ISO 354 revision taskforce. The project led to the construction of a real, in-use reverberation room at the headquarters of ROCKWOOL A/S in Hedehusene, Denmark, and contributed significantly to the discussions around the standard.
-
Fault detection via vibro-acoustical data for electrical transformers
At Oktogrid (Copenhagen): DSP algorithms detecting electrical transformer faults from MEMS microphone signals. The work led to an issued US patent.