AI at urban scale: sound and machine learning working together to detect large-scale patterns between humans and non-humans

A data set always has some direct or indirect temporal scope. It can be an instantaneous snapshot of many things happening in a blink of an eye, or one thing happening over the years. Machine cognition free us from the limitations inherent to our perception apparatus, enabling us to develop systems that work upon data sampled from inhumanly vast or minuscule timescales. This form of sampling not only documents but makes it possible the rendering of causal patterns normally incomprehensible to us.

Discipline: architecture, soundscape ecology
Role: data visualization, software development, machine learning
Team: Olesia Kovalenko, Antonia Burchard-Levine, Evgenia Vanyukova, Olga Cherniakova
Research Partner: Strelka Media Institute
Year: 2019

Responsive image

SOL is an experimental sensing network platform deployed at an urban scale. Its design utilizes the sound frequency spectrum and inexpensive microphones distributed throughout the city to capture sound events at unimaginable timescale: from micro-sounds to macro patterns unknown to human perception. The sound events are analyzed and classified under hundreds of categories, including various species of birds and thousands of anthropogenic sounds such as cars, and pedestrian activity. Classified events together with their spatial data are used to derive meaningful relationships between actors and events in the city.

Compared to the radio frequency spectrum, sound’s frequency space is currently free from any form of regulation. This, in turn, has major effects on how various species (from birds to insects) can adapt to the urban environment. Platforms for research on urban sensing can facilitate the emergence of new urban practices that address the preservation of biodiversity in highly human-populated areas. By bringing research methodologies applied in the field of soundscape ecology, SOL sheds new light on the role of sound as a carrier of rich information networks between different agents in the urban environment.

SOL (Sonic Ontological Language) emerged out of a working collaboration between policymakers, architects, and urban developers in Moscow during The New Normal research program in 2019.

Responsive image
Responsive image
Responsive image