Noise pollution is one of the banes of modern life. Its effects are broad and insidious, ranging from sleep and work disruption to hearing impairment and psychological damage.

So in recent years, a number of countries, such as Germany and the U.K., have begun to measure noise pollution over large metropolitan areas on a systematic basis. This is a tricky task. Part of the problem is that noise levels can vary throughout the day and night and over relatively small distances. And yet a noise pollution map must capture this variation over vast areas. Consequently, these maps are time-consuming and expensive to make.

Today, Rajib Rana at the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia’s national science agency, and a few pals say they have developed a cheaper, faster, and better way to make noise pollution maps. Their idea is to crowdsource the data from smartphones that record sound levels as their owners wander the streets.

Although it sounds simple, this turns out to be a task with some subtle difficulties. Smartphones can provide accurate measurements of ambient sound as well as a location and time stamp to show exactly where the measurement was taken. The difficulty is in ensuring that the measurements are useful recordings of ambient noise and not other sounds that might distort the readings. Unwanted noises include conversations nearby, which will sound loud but are unlikely to contribute significantly to ambient noise, or the sound of keys or other paraphernalia jangling in a bag or the muffled readings that a phone inside a pocket might produce.

So Rana and co have developed a number of tests that the smartphone employs to determine whether it can take a decent reading. The first is a GPS measurement to determine whether the phone is inside or outside, since they are only interested in readings taken outside.

Next, the phone listens for ambient conversations. If it finds any it waits until they are finished before taking a reading.

Rana and co are also only interested in readings taken when the phone is handheld and not inside a bag or pocket, where ambient noise would be muffled. They say it’s possible to automatically sense the phone’s position using various built-in sensors such as the proximity sensor and accelerometer. These generate a unique signature when the phone is handheld, which the software uses to determine whether it can take a reading. Rana and co say they can detect handheld use with an accuracy of 84 percent.

Once these criteria are all satisfied, the phone takes an ambient sound level reading and stores it along with the location and time. Finally, it sends the data to a central servier when it next enters a wifi zone.

A central server then begins to assemble a map from all the crowdsourced readings. These guys have tested their new approach, which they call Ear-Phone, on a number of Nokia and Android phones on the streets of the Australian capital, Canberra. And they’ve compared the results with ground truth data that they recorded using conventional sound level meters at the same time.

They say that the crowdsourced data is remarkably accurate and allows them to reconstruct the ground truth data, even when up to 40 percent of the original data points are missing. “Ear-Phone can accurately characterize the noise levels along roads from incomplete samples,” they say.

There is one downside—battery life. In particular, GPS measurements are a significant juice drainer. They say their system runs for around five hours on a standard smartphone, which is not nearly long enough for most people. The app would drain any phone before lunchtime.

Rana and co say improvements are possible by optimising the phones but this is certainly a factor that will need more thought.

Nevertheless, Ear-Phone is a potentially important step forward for noise monitoring. “Our work demonstrates that it is feasible to use mobile phones for environmental sensing applications such as noise pollution monitoring,” they say.

With clever incentives for those who take part, it’s not hard to imagine how this approach could build accurate noise maps over much larger areas than is possible using conventional methods. And that could help decision makers put in place the necessary measures to mitigate problems more quickly.

Useful stuff. These guys ought to be shouting from the roof tops about their new app—just not too loudly.

Ref: arxiv.org/abs/1310.4270: Ear-Phone: A Context-Aware Noise Mapping using Smart Phones