Most of the sound around us is accidental; much of it is unpleasant. But the effect of excessive noise goes beyond simple annoyance. To name a few effects, noise pollution can disturb sleep, interfere with complex task performance, and modify behavior [1]. In children, it can cause impaired reading comprehension, long term memory loss, and high blood pressure [1]. Noise pollution will become an increasing threat to residents in cities with rising populations. The first step to mitigating it is to monitor the levels of noise pollution at each point in the city. Our group has designed a network of sensors, optimized with IoT, to monitor noise pollution.

The sensor consists of three main components: the microphone, analog-digital converter, and microcontroller+4G module. The sound data from the environment is transduced through the microphone, which is then converted and transmitted over cellular networks. Cloud signal processing methods will be used to separate each signal into individual sources of noise (i.e. car honks). The intensity of each individual signals in the five sensors that read the highest intensity will be used to triangulate each source, and noise modeling techniques will be used to model noise levels throughout open space.

The sensor incorporates a Knowles SiSonic microphone [3] with an ADS1115 ADC module to amplify the data [4]. This will take full advantage of the SARA-R410M-02B I2C capability [2]. Which will allow for data transmission through cellular networks. A TL-5930/T battery, at 3.6V with a capacity of 19 Ah [5] will power the module.

We provide a quick back of the hand calculation to prove the feasibility of such a power source. The ADC and microphone seem to draw a typical 270 uA together [3,4]. The SARA-R410M-02B consumes 0.6 mA in the absolute lowest power case [2]. Therefore, in the absolute lowest power case, the system should be able to last just under 2 and a half years. This means in one year there is an excess of around 11400 mAh which can be used for transmission. According to the datasheet, if we transmit at the lowest possible power, we can do so at 100 mA. In the best case scenario, therefore, we will be able to transmit for 47 seconds in every hour. Keeping a factor of safety of a little more than 2, assume that we can get 20 seconds in every hour. Then, using the average speed of 17 Mbps for Canada [6], we should be able to send 340 Mb every hour. Since a minute of monaural data takes about 5 MB [7], Therefore, out of every hour, we should be able to get approximately on the order of 8 minutes of sound data.

All data processing will be done in a cloud. A given signal from a sensor can be expressed as a superposition of individual signals with the addition of signal noise (all sounds that do not count as the auditory noise we are trying to monitor). Signals corresponding to the auditory noise we are monitoring can be filtered out using machine learning techniques to detect the presence of each individual auditory noise. Signal processing techniques, including cross-correlation performed on the signal and its spectrogram, can be used to adjust for phase and pitch shift. After filtering out each auditory noise from each sensor, the sensors which recorded the highest level of sound can be used to triangulate the source of said sound.

The Time Difference of Approximation method would be used to locate these sources [8]. The method uses 5 sound receivers to triangulate the location of the source in 3D space. In free space, the location is determined by solving a system of linear equations, which can be fitted with the least squares or gradient descent method. With obstacles, represented by a computer model of the city, the shortest path must first be determined, this will be done using Dijkstra's algorithm. When the noise sources are located, first principles based sound modeling techniques will be used to determine sound pressure as a function of position (discussed at length in [8]), which can then be expressed in decibels.

The network can allow us to monitor noise levels in an entire city given a relatively sparse network of sensors. The data processing can allow us to separate the signals and identify different sources. The sensors are resilient, and only isolate sources of noise that the city wants in their analysis. By analyzing the different sources of noise at different time, city planners can then put forth targeted plans to remedy the noise pollution in those areas. Even beyond the scope of noise monitoring, this robust system can allow us to pinpoint specific urban activities around the city at specific times.

[1] S. Stansfeld and M. Matheson, "Noise pollution: non-auditory effects on health" 10.1093/bmb/ldg033.
[2] U-Blox, “LTE Cat M1 / NB1 and EGPRS modules” https://www.u-blox.com/sites/default/files/SARA-R4-N4_DataSheet_%28UBX-16024152%29.pdf
[3]Knowles, “Zero-Height SiSonicTM Microphone,” https://media.digikey.com/pdf/Data%20Sheets/Knowles%20Acoustics%20PDFs/SPU0410LR5H-QB_RevH_3-27-13.pdf
[4] Adafruit, “ADS1115 16-Bit ADC“ https://www.adafruit.com/product/1085
[5] Tadiran Lithium Batteries http://www.tadiranbat.com/assets/tl-5930.pdf
[6] Opensignal, “State of Mobile Networks,” https://www.opensignal.com/reports/2016/01/canada/state-of-the-mobile-network

Inspiration

As something that we can personally attest to, having a clean atmosphere to focus on our studies away from the clutter and confusion in cities is crucial for the well-being and success of students. Our increasingly crowded world is gnawing away at our mental health and rendering urban areas unappealing. Our motivation in designing this solution is to address an issue in our lives that often goes unmentioned and receives little importance. The auditory environment of any community, be it a university campus, a city district or a residential neighborhood is of great influence on the quality of life of its community members. Our system is designed to provide an elegant solution that is not only pre-emptive but can be used as a proactive source of data to improve the quality of life from a locational to the regional scale. We value that our system is cost-effective and expandable to many environments without being intrusive and infringing on the privacy of citizens. Not only are we truly inspired by the impact that our solution will have in transforming smart cities, but by how policymakers and citizens will have real-time low-cost access to data about their environment. As we progress towards the future, citizen engagement and involvement of important stakeholders will result in the disappearance of the issues surrounding mental health in tomorrow’s smart cities.

Voting

Voting is closed!