Methods for Sensing Urban Noises
Tong Liu,Yu Zheng,Lubin Liu,Yanchi Liu,Yanmin Zhu
2014-01-01
Abstract:Many people live in cities are suffering from noise pollution, which may impair the physical and mental health. A first step towards the understanding of urban noises is to measure real noise levels. In this technical report, we introduce two categories of methods to measure urban noise levels, discussing their advantages and disadvantages. Finally, we apply one of mobile phone-based methods to an in-the-field study that explores the noise situation in New York City (NYC) [10]. The data collected and the mobile client used in the study are available for download at [2]. INTRODUCTION The rapid progress of urbanization is leading to serious noise pollution [11], which may cause stresses, sleep losses, high blood pressures and even heart attacks [4]. Tackling noise pollution has been attracted a wide range of attention [5][6] [8][9]. For example, European Union planned to build up three-dimensional noise maps of all major cities by 2007 [4]. In addition, since 2011, NYC has opened the platform where people can call 311 to complain whatever makes them uncomfortable, including noises [1]. To build a noise map based on real noise measurements [5][9] [10], it is necessary to collect the real-measured noise data of different places. In this paper, we introduce a few measuring methods, discussing their applications to different scenarios that are concerns with money, time and human resources. Using one of the proposed methods, we conducted a study to measure the real noise level of 36 locations in Manhattan. The results of the experiment are used to explore the correlation between real-measured noise levels and people’s complaints about urban noises [10]. METHODS In this section, we introduce two categories of sensing methods to collect real-measured noise levels, according to the measuring devices. One is based on professional sound level meters; the other is using mobile phones. Professional Device-Based Methods The first category of methods employs a professional noisemeasuring device, named sound level meter [3]. A standard sound level meter, as shown in Figure 1 A), gives a readout of equivalent continuous sound level in decibels (dBs), converted from the voltage signal sensed by its microphone. Figure 1. Sound level meter Figure 2. A network of sound level meters A straightforward method is to deploy professional devices permanently in different locations with a certain density e.g., a device per 0.25 square kilometers. The data collected by these devices is then transmitted to a backend system through wired or wireless communication channels, as shown in Figure 2. While this method can produce a long period of high quality noise measurements, it requires a high cost of deployment and maintenance, e.g. power supply, a cover protecting the device from rain, and data transition channels. Additionally, the coverage of this approach is also a concern as we cannot really deploy such sets of devices anywhere. Without deploying these professional devices permanently, another method is to employ many people, each of which holds a professional device to simultaneously measure the noises in different locations. While this method is more flexible than a fixed deployment, carrying a sound level meter, which is about 30×10 centimeters, is not a fun for a user. Thus, this method is difficult to scale up to thousands of users. The coverage of the locations that can be sensed through this approach is still a problem. Mobile Phone-Based Methods To solve the inconvenience of the aforementioned methods, mobile phones are used as substitutes whose microphones are employed to detect the sound power of the surrounding environment. Though the accuracy is slightly lower than a professional sound level meter, the wide availability of mobile phones in end users significantly increases the spatial and temporal coverage of sensing spaces. 1) In an ideal mobile phone-based method, many people hold a mobile phone to measure their ambient noises and then send the noise data back to a centralized place for a further processing. According to some studies [7][9], however, the noise level measured by a mobile phone has a certain deviation from the value measured by a professional sound level meter. Moreover, different mobile phones have *The paper was done when the first author was an intern in Microsoft Research under the supervision of the second author. The mobile app was developed by the third author in Shanghai Jiao Tong University. different deviations. Consequently, the data collected by different phones is not directly comparable. To overcome the problem, a calibration is needed for each phone to obtain accurate noise data. This is quite time consuming and reduce the feasibility of a large-scale deployment. Some calibration algorithm has been proposed to do a selfcalibration. For instance, an intervention–free calibration algorithm [7] uses a linear function to model the relationship between phone measurements and real values, as shown in Equation 1: