Skip to content

Research Context

Jump to bottom
Alexander Refsum Jensenius edited this page Jul 17, 2026 · 3 revisions

ambiscape is a research instrument first: it was built within the AMBIENT project (fourMs / RITMO, University of Oslo) — which studies how indoor environments sound and look, and how they affect the people in them — to answer questions about how rooms sound over long durations. Its feature set mirrors that programme.

The corpora it grew on

  • StillStanding365 — 365 daily ~8-minute standstill sessions through 2023, each in a different room, five modalities (FOA audio, 360 video, chest IMU, heart rate, daily log). Its Zenodo deposit publishes non-identifying 1 Hz acoustic features — the schema ambiscape deposit now exports — plus directional energy arrays. Depth: one year, one person, one region.
  • An intercontinental database of indoor environments — long-form opportunistic sessions (hotel rooms, airport halls, a train, domestic interiors, a performance) collected while traveling, including full overnight recordings. Breadth: many building cultures, climates, and mains frequencies, one protocol. This corpus drove most of ambiscape's design: streaming extraction, state-transition analysis, taxonomy figures, the ISO layer.

Findings that shaped the tool

  • Diffuseness stratifies indoor life. Median ψ separates transit halls (0.84–0.90) from lived interiors (0.6–0.8) from machine-dominated rooms (≈ 0.2) — and the machine-dominated condition turned out to be the most common indoor state observed. Hence ψ and R are first-class descriptors.
  • State transitions carry the information. A ventilation unit switched off releases, within seconds, everything it masked (birds, bells, water, footsteps) — a direct lo-fi → hi-fi traversal in Schafer's terms. Hence transition detection, transition segments, and the protocol advice to stage transitions.
  • Rooms have rhythms at every scale: refrigerator duty cycles (~24 min), the city's diurnal low-frequency envelope (nightly minimum in the small hours), day/night machine states. Long-duration descriptors — duty cycles, diurnal envelopes, state-alternation periods — are candidate ingredients for a still-missing standard metric of environmental rhythm.
  • A-weighted single numbers understate LF-dominated scenes (vehicle interiors, HVAC rooms) by up to ~15 dB relative to unweighted levels — why the summary reports both.

Standards positioning

Descriptor conventions follow ISO 1996; psychoacoustic indicators follow ISO/TS 12913-3 (via MoSQITo); the perceptual layer of ISO 12913-2 (Method-A affective ratings) is supported by protocol, not by software — no instrument can supply it. See Design Decisions.

Related tooling and lineage

  • ambiviz — spatial visualization (AEM, anglegram, directogram); ambiscape follows its plot naming.
  • Musical Gestures Toolbox — the fourMs motion-analysis lineage ambiscape's reporting style descends from.
  • MoSQITo, PANNs, silero-vad — the optional metric/ML dependencies.

Citing

Until a paper is out, cite the software: fourMs (2026), ambiscape (version as installed), https://github.com/fourMs/ambiscape — and this wiki's pages for method rationale.

Clone this wiki locally