“Noise pollution is one of the topmost quality of life issues for urban residents in the U.S. with proven effects on health, education, the economy, and the environment,” said Juan Pablo Bello, director of the Music and Audio Research Lab (MARL) at the NYU Steinhardt School of Culture, Education, and lead investigator on the SONYC initiative. “Yet, most cities lack the resources for continuously monitoring noise, the technology for understanding how individual sources contribute to noise pollution, the tools to broaden citizen participation in noise reporting, and the means to empower city agencies to take effective, information-driven action. SONYC will help address these shortcomings.”
The SONYC project will use a distributed network of both sensors and individuals for large-scale noise monitoring. During the project’s initial phase, expected to take one year, approximately 100 sensors will be installed on NYU buildings at locations around Manhattan and Brooklyn to record street sounds as a way of “teaching” the sensors to recognize and differentiate between different types of noise (e.g., jackhammers, sirens, music, yelling, barking, etc.).
“SONYC sensors will need a new type of battery-powered computing node to support and even relearn diverse node recognition algorithms in situ while consuming very low power,” said Anish Arora, a professor of computer science and engineering from The Ohio State University, who has built numerous sensor networks. “By combining accurate and robust machine listening with large-scale, albeit subjective human complaint data, we expect to provide reliable information to support decision making.”
To ensure privacy, sound will be collected in 10-second snippets with random gaps in between. While sample snippets were reviewed by independent acoustical consultants Cerami & Associates and deemed to be unrecognizable as conversation, signs will be posted near sensors to alert passersby that sound recording is taking place. Audio will only be recorded from each sensor for a total of four months over the course of a year to make sure daily, monthly and seasonal variations are accounted for. Access to all recordings will be restricted and limited to researchers.
During phase two, the sensors will no longer record audio. Instead, they will use machine listening technology to recognize individual sound sources and produce statistical reports on sound levels and types. That data will be compared with 311 complaints and other citizen reports of noise to form an “acoustic model” of the city. Using a combination of data mining and predictive models, the project plans to work with City enforcement officials to more strategically identify and mitigate noise.
“Phase two of the project will usher in a distinct set of challenges as we attempt to reliably interpret, verify, and visualize the data we receive from both machines and people,” said Claudio Silva, a professor of computer science and engineering at the NYU Tandon School of Engineering and SONYC team member. “Because SONYC will continuously monitor and enable noise pollution to be analyzed, it can guide efforts to quantify the noise and understand its effects, thus providing actionable knowledge that can help City officials and residents address the problem.”
In addition to professors Bello, Silva, and Arora, the team also includes R. Luke DuBois and Oded Nov from NYU’s Tandon School of Engineering, as well as Justin Salamon and Charlie Mydlarz, senior research scientists at NYU’s Center for Urban Science and Progress (CUSP).
"SONYC is the culmination of a three-year research pilot that has brought together some of NYU's best research areas: machine learning for audio, big data, user experience design, data visualization, and citizen science,” said DuBois. “I'm proud to work with my colleagues at Tandon, Steinhardt, and CUSP on this project and look forward to making a meaningful contribution to understanding and mitigating noise pollution in New York City."
“New York City is the perfect environment for testing this system, not only because it’s the largest, densest, and noisiest city in the country, but also because it has a strict noise code and an already active citizen noise reporting system,” said Nov, associate professor of technology management and innovation at NYU Tandon. “Providing New Yorkers with customized web and smartphone applications, combined with innovative machine-listening technology, will significantly impact both the quality and quantity of data that we’ll be able to deliver to our city partners.”
The SONYC project will work closely with city agencies and industry in both research and implementation. Identifying noise events and designing and testing data-driven interventions will be done in cooperation with the New York City Department of Environmental Protection. The New York City Department of Health and Mental Hygiene will work with the team on using SONYC to study the public health effects of noise. The Downtown Alliance will provide access to infrastructure and logistical support needed to deploy the SONYC sensor network in Lower Manhattan, while the acoustic consulting team of the global design and engineering firm Arup will work closely with the project’s team in testing and developing its sensing technologies.
SONYC will also implement a summer program in partnership with the Center for K-12 STEM Education at NYU’s engineering school. This three-week program, to be part of the Center’s Science of Smart Cities initiative, will use project-based learning pedagogy to train K-12 students and teachers from the NYC public school system in Cyber-Physical Systems science, technology and engineering, with SONYC as a focal point.
SONYC is a collaboration between NYU Center for Urban Science and Progress (CUSP); NYU Tandon School of Engineering; NYU Steinhardt School of Culture, Education, and Human Development; and The Ohio State University.