@article {Park:2017:0736-2935:5389,
title = "Urban Soundmapping at the Edge",
journal = "INTER-NOISE and NOISE-CON Congress and Conference Proceedings",
parent_itemid = "infobike://ince/incecp",
publishercode ="ince",
year = "2017",
volume = "255",
number = "2",
publication date ="2017-12-07T00:00:00",
pages = "5389-5400",
itemtype = "ARTICLE",
issn = "0736-2935",
url = "https://ince.publisher.ingentaconnect.com/content/ince/incecp/2017/00000255/00000002/art00048",
author = "Park, Tae Hong and Yoo, Minjoon and Dye, Chris and You, Jaeseong and Buranintu, Varatep and Rekesh, Dima and Leonard, Isaac",
abstract = "In this paper, we report on Citygram [1]-[5] and IBM's recent partnership in soundmapping with a focus on capturing urban noise pollution via edge compute paradigms. Our noise sensor network is built on plug-and-sense design principles to maximize citizen-scientist participation and
sensor network growth. We thus provide both standalone hardware solutions as well as pure software solutions. The pure software solutions operate entirely on standard web browsers that run a single codebase on popular operating systems including OS X, Windows, Android, and Linux OS. On the
other hand, the standalone hardware solutions run continuously as dedicated sensing devices and are based on inexpensive, configurable single-board Raspberry Pis (RPi) that are applicable to various environmental sensing situations. The RPi forms the most fundamental computational base unit
and utilizes IBM Blue Horizon edge compute service. Together, Citygram and Horizon sensor solutions enable efficient workload expansion, autonomous operation, updates, and edge analytics. Both hardware and software node solutions capture, process, and analyze data at the "edge" and have been
developed to address issues associated with traditional sensor network designs: (1) ease of deployment, (2) network density, (3) reliability, (4) cost, and (5) dynamicity. The sensor network is particularly efficient as it is event-triggered and pushes high value data to the cloud only in
the presence of salient acoustic events. The sensor nodes perform automatic acoustic event detection and classification using a combination of procedural analytics, machine learning, and cognitive-based Deep Learning.",
}