Through map4noise, you can monitor the noise exposure of you and your community. See what kinds of noise are threatening the health of you, your family and friends. Map for noise around you, and do MORE before the increasingly serious noise pollution makes any effort too late. Your contribution protects not only your right to information, but also that of your fellow citizens.

We answer an interesting question "how to real-timely improve a noise classifier via crowdsourced noise labels". For a broader setting, our solution sheds a light on the possiblity of real-timely training a classification model via mobile crowdsourcing. The real-time classification paper download

Key features:

· Crowdsource Sensing

· Incentive Mechanism

· Human as Sensor

· Task Allocation

· Context Aware

· Real-time Learning

Application download (Android platform): map4noise.apk

Variously clinic evidences [1][2][3][4][5] indicate the exposure to environmental noise can cause both auditory and non-auditory health effects like hearing loss, annoyance, sleep disturbance, cardiovascular disease, cognitive impairment and even psychological symptoms. What is worse, noise pollution is pervasive in modern life. Approximately, 5% of the world population is suffering from noise-induced hearing loss [6]. In the European Union states, just the traffic noise makes 54% of population (56 million) exposed to unhealthy acoustic climate [7]. Noise also results great economic cost. In the US, noise-induced hearing loss is the most frequent occupational disease. About 22 million Americans are working in an environment with hazardous noise, and about $242 million dollars are spent on the compensation for occupational hearing loss per year [8].

A noise map can present measured or predicted noise exposure levels over a specified area. This visualized noise distribution will raise citizens' awareness of environmentally acoustic quality and also enhance the ability of city planners to successfully manage noise issues and further reduce noise-induced health threats [9]. In the strategy of controlling the severe noise pollution in urban areas, the European Parliament and Council issued the EU Environmental Noise Directive [10] to require all EU member states to prepare and publish strategic noise maps.

Since the release of EU Noise Directive, numerous pioneering efforts [11][12][13][14][15] have been made to optimize the method of acoustic climate assessment, in Europe. Generally, the technical implementations can be classified into two categories, namely the noise simulation and the noise measurement. The computational predication is a feasible and inexpensive approach. The procedure of performing the simulation requires input parameters including noise source data (e.g. traffic flow and composition, average speed) and propagation environment (e.g. 3D digital terrain, meteorological conditions). Even though the simulation has merits, it also has obvious shortcomings such as the lack of accuracy and temporal dynamics which would be significant to evaluate annoyance and sleep disturbance. In contrast, the approach based on measurement can provide higher temporal solution. A grid of stationary devices is necessary to provide measured sound pressure and associated data. For example, Lille, France, deployed a noise measurement network, which is consisted by more than 80 distributed devices [16]. The noise measurement network can monitor the acoustic climate in real-time by fusing the data from every devices, and provide alerts for potential noise threats with small temporal granularity. However, to achieve higher spatial resolution, a much denser sensor network will be necessary, which greatly increases the difficulty of deployment and the cost of maintenance.

Although the traditional noise measurement network is neither feasible nor affordable for a large geographic area [17], the emerging sensor-enabled smart phones introduce a novel and promising technical approach to "install" a noise measurement network based on the mobile sensing [18][19]. Inspired by that, a few works [20][21][22][23][24][25][26] have been done on the topic of the mobile sensing based noise climate monitoring. The New York City 311 service [27] could also be one since it utilizes "human as a sensor" to record noise events instead of quantified noise data. From the perspective of us, these pioneering works are far from full-fledged crowdsensing applicaitons, since they only implemented several basic services such as measuring the personal and/or community exposure, and didn't emphasize on solving the known challenges of crowd-source sensing such as incentive design, data quality, task allocation, resource constrain and privacy [28][29][30].

We propose to design a noise monitoring system based on the crowdsource sensing with a focus on solving the known challenges. Comparing with the traditional implementations, the crowd sensing can greatly enhance temporal and spatial resolution for environmental noise monitoring, which makes the real-time noise alert possible. Incentive mechanism will be applied to encourage the participatory of users. The idea of "human as sensor" will be used to collect qualitative data that are hard to be automatically classified. Also, context-awareness designs will not only provide essential data such as location, which is significant when fusing microphone samplings, but also be used to evaluate the data quality of users. To solve the issue of sparse data, both the task allocation mechanism and the acoustic propagation model will be utilized.