Annotated Bibliography
A guide to the research papers collected in sources/, organized by theme.
Physics of Sound & Signal Analysis
End-to-End Probabilistic Inference for Nonstationary Audio Analysis
Wilkinson et al., 2019 — 1901.11436v5.pdf
arXiv:1901.11436
Proposes a unified Gaussian process framework that jointly performs time-frequency analysis and nonnegative matrix factorization for audio signals. Uses spectral mixture kernels with nonstationary priors and a state-space formulation for scalable inference (linear in time steps). Directly relevant to building principled, probabilistic audio analysis pipelines that don’t rely on ad-hoc spectrogram stages.
Why it matters for Chaos Hearing: This paper shows how to replace the traditional “compute spectrogram → extract features” pipeline with a single probabilistic model. It connects to our interest in understanding the structure of sound — not just its surface representation — and opens the door to audio analysis methods that can adapt to nonstationary, real-world signals (music, speech, environmental sound).
Object-Based Synthesis of Scraping and Rolling Sounds Based on Non-Linear Physical Constraints
Agarwal et al., 2021 — 2112.08984v1.pdf
arXiv:2112.08984
Presents a source-filter model for synthesizing sustained contact sounds (scraping, rolling) with physically and perceptually meaningful parameters. Key innovations include nonlinear contact force constraints, naturalistic normal force variation, and location-dependent impulse response morphing. Perceptual experiments validate realism.
Why it matters for Chaos Hearing: This is physical sound synthesis done right — grounded in mechanics but validated by human perception. It bridges the physics thread (how sound is produced) with the cognition thread (how sound is heard). The nonlinear constraints echo the chaotic dynamics we study in auditory models.
Mathematical Foundations
Superselection Structure of Massive Quantum Field Theories in 1+1 Dimensions
Müger, 1997 — 9705019v1.pdf
arXiv:hep-th/9705019
Analyzes the superselection structure of massive QFTs in 1+1 dimensions using algebraic quantum field theory (Haag duality, split property). Shows that certain representation-theoretic structures are vacuous for massive theories but rich for massless/conformal theories.
Why it matters for Chaos Hearing: The algebraic structure of wave phenomena — particularly the distinction between massive and massless theories — provides deep mathematical analogies for understanding sound propagation, resonance, and the structure of auditory representations. The conformal invariance of massless theories connects to scale-invariant properties of pitch perception and harmonic structure.
Cognition of Hearing & Neuroscience
bioRxiv 2026.03.12.711349v1
2026.03.12.711349v1.full (1).pdf
(Auditory neuroscience / cognitive hearing — paper details to be annotated upon review.)
Provisional relevance: Likely connects to the cognitive thread — auditory processing, neural correlates of sound perception, or related neuroscience of hearing.
SSRN 5827765
ssrn-5827765.pdf
(Social science / cognitive science of hearing — paper details to be annotated upon review.)
Provisional relevance: Likely connects to the broader cognitive and social dimensions of hearing — perception, musical experience, or auditory phenomena in clinical populations.
Themes Across the Collection
The papers cluster around a central question: How does sound go from physics to perception?
Physical Production Signal Structure Neural Processing
(mechanics, nonlinear → (spectral analysis, → (auditory scene analysis,
dynamics, synthesis) GP models, TF reps) imagery, hallucination)
The Chaos Hearing project lives in the arrows between these stages — understanding the transformations, building computational models of them, and designing interfaces that let people interact with sound at each level.