Sonic Drift

Sonic Space Shift Analysis (Workflow)

– Method & Score (IP: Artist): Conceptual framework and Spatiotemporal Score for compact-city scenario testing; geocoded data model; GIS cartography; mixed-media installation.
– AI Analysis (IP: CERTH): Selection and integration of AER / USS algorithms; statistical analysis; comfort-over-time traces; domain inference (biophony / anthropophony / technophony).
– Joint Addition (Co-developed): Subsonic & Ultrasonic Substrate category to include infra-/ultra-sonic ranges (reported as LAeq dB where applicable).

Geocoding & Geospatial Workflow

– Geocoding: All recordings (active & passive) are linked to score positions and times; walk paths and rooftop logger sites are encoded as layers (points / lines / polygons) with observation notes. (Artist)
– GIS Authoring: QGIS (open-source) is used for viewing, editing, analysis, and generating print-ready GIS / SVG comfort grids; ethnographic observations and AI outputs are integrated into planner-ready layers. (Artist)
– Cartographic Products: Comfort grids, domain presence maps, VA (Valence–Arousal) traces, and site overlays for Garden / Helihaven / Masui. (Artist)

Audio Recording Methods & Tools

– Passive (rooftops) and Active (street-level walks) using contact, shotgun, and omnidirectional microphones, and a high-quality recorder. (IP split: field method & Score by Artist; passive data logging and monitoring device network system  by CERTH)
– Scheduling: Sunrise / sunset / midnight aligned with the Score to capture temporal change.

(Privacy-by-Design)– GDPR Rooftop Logger (IP: CERTH):
On-device, low-energy voice obfuscation in real time to protect privacy while preserving environmental integrity.

Hardware: Raspberry Pi 4B + AudioMoth (mic) + power bank (portable).

– CDA-North Active and Passive Recording Samples (scored geocode tags) — IP: Artist
– CDA Multitrack Set (incl. open-source audio editing workflow) — IP: Artist
– CDA-North Time based labels of passive recording samples, AER / USS Labels & GDPR software configurations — IP: CERTH

Statistical Analysis (Metrics)
(Analysis and metric design conducted by CERTH. Metric selection and artistic relevance reviewed and curated in dialogue with the Artist.)

1. Stress / Calm Levels (AER, VA model):
Measures perceived affect on a 2D Valence–Arousal diagram, indicating how stressful or calming a recording segment is. Values are normalized on a 0–1 scale.

2. Sonic Space Presence (USS, domain inference):
Source separation estimates domain presence across biophony / anthropophony / technophony; intensity calibrated on a 0–1.414 range.

3. Subsonic & Ultrasonic Substrate (Co-developed):
Frequency analysis identifies subsonic / ultrasonic components reported as LAeq dB; included at the artist’s request to account for vibrational substrates affecting acoustic dis/comfort.
Outputs are aggregated for the three pilot spaces (Garden, Helihaven, Masui) and visualized as dis/comfort-over-time traces, spectrograms, and GIS overlays by the artist.

Audio Essay (Feedback Installation)

Audio Essay Production: Three stereo mixes developed from active and passive field recordings (mastering by Christophe Albertijn)

Dubplate Listening Deck: Three hand-cut dubplates (one per pilot site) presented on a modular headphone setup, allowing cueing, looping, and reversal to emphasize temporal and affective textures. (cuts by Dubplate – Lester Michiels)

Printed Components: Riso-printed Spatiotemporal Score and GIS-derived comfort maps designed as interactive surfaces for public annotation. (Sleeve design and print by Chez Rosi – Axel Claeys).

Feedback & Reflection Tools: On-site listening journals and online listening cards collect participant reflections on acoustic comfort, urban density, and affective perception of space, linking sonic aesthetic experience to planning and design dialogue (Artist).

Presentation of the Audio Essay as installation at the final ReSilence Exhibition, BASE – MILANO (IT)

Process Diagram
Methodology Refinement, Prototype, & Installation Building

Project Cycle
Site → AI → Re-synthesis diagram

Eco-Acoustic Environment Analysis: The Eco-Acoustic Environment Analysis interface provides a unified environment for environmental sound assessment by combining three complementary computational pipelines: source separation, eco-acoustic domain classification, and comfort (valence–arousal) estimation. The interface enables users to decompose complex audio scenes, interpret their ecological composition, and assess their perceptual comfort characteristics in a single, integrated workflow. — IP: CERTH

Functional flow
The eco-acoustic analysis workflow consists of three interconnected modules executed in the following order:

1. Source Separation:
The system invokes the universal sound separation module  with predefined configuration and model checkpoints. The uploaded audio is decomposed into up to independent stems, representing distinct sound contributors such as human speech, engine/mechanical noise, natural environmental sounds, and background components. These stems are displayed in a structured list with download buttons for independent inspection.

2. Domain Presence Estimation:
Each separated stem is mapped to one of four eco-acoustic categories: Biophony, Geophony, Anthropophony, and Technophony. Classification is performed using filename-inference heuristics and consistent label-mapping rules. The interface displays a donut chart summarizing the proportion of each eco-acoustic domain.

3. Comfort / Affective Analysis:
The comfort estimation module computes a continuous valence–arousal curve using an AST -based perceptual inference model. The output includes a Mel-spectrogram of the input audio and a comfort curve, allowing interpretation of acoustic stressors and perceptual impact.