Audio Engineering Society

Convention PaperPresented at the 137th Convention

2014 October 9–12 Los Angeles, USA

This Convention paper was selected based on a submitted abstract and 750-word precis that have been peer reviewedby at least two qualified anonymous reviewers. The complete manuscript was not peer reviewed. This conventionpaper has been reproduced from the author’s advance manuscript without editing, corrections, or consideration by theReview Board. The AES takes no responsibility for the contents. Additional papers may be obtained by sending request

and remittance to Audio Engineering Society, 60 East 42nd Street, New York, New York 10165-2520, USA; also seewww.aes.org. All rights reserved. Reproduction of this paper, or any portion thereof, is not permitted without directpermission from the Journal of the Audio Engineering Society.

The design of urban sound monitoring devices

Charlie Mydlarz1,2, Samuel Nacach2, Tae Hong Park2 and Agnieszka Roginska2

1NYU CUSP, New York, NY, 11201, USA

2NYU Steinhardt, New York, NY, 10012, USA

Correspondence should be addressed to Charlie Mydlarz (cmydlarz@nyu.edu)

ABSTRACTThe urban sound environment of New York City is notoriously loud and dynamic. As such, scientists,recording engineers, and soundscape researchers continuously explore methods to capture and monitor suchurban sound environments. One method to accurately monitor and ultimately understand this dynamic en-vironment involves a process of long-term sound capture, measurement and analysis. Urban sound recordingrequires the use of robust and resilient acoustic sensors, where unpredictable external conditions can havea negative impact on acoustic data quality. Accordingly, this paper describes the design and build of aself-contained urban acoustic sensing device to capture, analyze, and transmit high quality sound from anygiven urban environment. This forms part of a collaborative effort between New York University’s (NYU)Center for Urban Science and Progress (CUSP) and the NYU Steinhardt School’s Citygram Project. Thepresented acoustic sensing device prototype incorporates a quad core Android based mini PC with Wi-Ficapabilities, a custom MEMS microphone and a USB audio device. The design considerations, materialsused, noise mitigation strategies and the associated measurements are detailed in the following paper.

1. INTRODUCTION

Noise pollution is an increasing threat to the well-being and public health of city inhabitants [17, 4, 3].Large advances have been made in noise predictionover the last few decades, with applications utiliz-ing Geographic Information Systems (GIS) technolo-gies and sophisticated noise transmission modeling

[18, 7, 5]. However, the complexity of sound prop-agation in urban settings and the lack of an accu-rate representation of the distribution of the sourcesof this noise have led to an insufficient understand-ing of the urban sound environment. The presentedproject aims to continuously monitor and ultimatelyunderstand these urban sound environments. Itis a multidisciplinary collaborative effort between

Mydlarz et al. The design of urban sound monitoring devices

New York University’s (NYU) Center for Urban Sci-ence and Progress (CUSP) and the NYU SteinhardtSchool’s Citygram Project [11, 13, 14, 8, 12]. Theimpetus of the Citygram project is focused on thelack of sufficient mapping paradigms for non-ocularenergies in urban settings. These energies, namely,sound can have a profound effect on a cities inhabi-tants and the key to understanding this effect firstlylies in the measurement of this energy. NYU CUSP’sinterests are focused on the noise of New York City,including how it impacts on the health of the city’spopulation, correlates with urban problems rangingfrom crime to compromised educational conditions,and how it affects real estate values. While a num-ber of past studies have focused on specific contextsand effects of urban noise [6, 10, 21, 16, 15, 1], nocomprehensive city-wide study has been undertakenthat can provide a validated model for studying ur-ban noise in order to develop long-lasting interven-tions at the operational or policy level.

The project is currently using NYC as a “lab” withthe aim of creating a model that can be utilizedand implemented in other cities around the world.With its population, its agency infrastructure, andits ever-changing urban soundscape, NYC providesan ideal venue for a comprehensive study and under-standing of the problem of urban noise. To achievethis goal an initial network of low cost acousticsensing devices were designed and implemented tocapture long-term objective acoustic measurementsfrom strategic locations throughout the city usingwireless communication strategies. These prototyperemote sensing devices (RSD’s) currently incorpo-rate a quad core Android based mini PC with Wi-Fi capabilities, and a custom MEMS microphone,whose characteristics are detailed in an accompanay-ing paper [9]. Acoustic data captured from each sen-sor node is comprised of a standard set of low-levelaudio descriptors for use in analysis, online mappingand visualization. The initial goal is to develop acomprehensive cyber-physical system 1 that providesthe capability of capturing, analyzing and wirelesslystreaming environmental audio data, along with itsassociated acoustic features and metadata - includ-ing automatic source identification.

The EU Directive (2002/49/EC) [2] resulted in the

1Network connected, distributed computing systems mon-itoring physical phenomena

production of noise maps for major urban areasacross Europe in 2007, to inform strategic planningfor noise control. However, many authorities acrossEurope found it difficult to use these noise mapsfor any kind of mitigation or action planning, duein part to a lack of confidence in the output datareflecting reality and the lack of any temporal vari-ation. The solution to this relied on more extensivemeasurement initiatives, which were prohibitivelyexpensive.

An example of a commercially available RSD is theLibelium Waspmote 2. The Waspmote can be con-figured as a “Smart cities” acoustic sensing devicewith an externally attached microphone and inter-nal audio capture and transmission capabilities. Thesystem is designed to capture audio and extract ba-sic level features for transmission to a central servervia Wi-Fi or cellular networks. The unit is ruggedi-zed and expandable via the multiple sensor inputports on the exterior of the case. Whilst this is avery capable solution to urban noise monitoring, itdoes not provide the processing power required foradvanced acoustic feature extraction in-situ. Withstarting prices of $620 as of June 2014, the unitwould also necessitate a prohibitively large equip-ment budget if deploying densely across a large area.The development of a low cost RSD solution is there-fore required for extensive urban noise monitoring.However, to adequately achieve this, the proposedsystem must firstly be physically resilient to theyear round environmental conditions of NYC, andsecondly, be robust to varying network connectivitysituations. Discussed in this paper are the measure-ments, design considerations, and noise mitigationstrategies employed, with respect to the sensor hard-ware to create the initial prototype RSD.

2. HARDWARE DESIGN AND TESTING

2.1. Computing and audio I/OThe projects sensor network is based around a con-sumer computing platform where low cost and highpower are of paramount concern. The design phi-losophy is based on the creation of a network thatprovides dense spatial coverage over a large area,through the deployment of inexpensive and physi-cally resilient sensors. The Citygram project con-sidered and tested numerous hardware platforms for

2http://www.libelium.com/products/waspmote/

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 2 of 9

Mydlarz et al. The design of urban sound monitoring devices

Fig. 1: Tronsmart MK908ii Android based mini PC

use in the projects RSDs [14]. These included Alixboards, Raspberry Pi, and a range of smart phones.The criteria for selection were: audio capture capa-bilities, processing power and RAM, low power con-sumption, a flexible OS, onboard storage, wirelessconnectivity, I/O options, robustness, low cost andsetup simplicity. The Android mini-PC platformwas eventually adopted as the hardware of choice asit best met the RSD selection criteria. A number ofthese mini-PCs incorporate an in-built microphone,however, the recording quality proved to be low interms of the microphone itself and the addition ofauto gain control in the signal chain on some de-vices.

At the core of the projects static remote sensing de-vice (RSD) is a Tronsmart MK908ii mini PC run-ning the Android 4.2, Linux based operating system.These small (87mm x 46.5mm x 14.3mm) and versa-tile devices shown in Figure 1 are priced at $70 as ofJune 2014 and provide a quad core Rockchip CortexA9 processor clocked at 1.6GHz, a quad core Mali400 GPU, 2GB of DDR3 RAM, 8GB of NAND flashstorage, USB I/O, Bluetooth 4.0 and Wi-Fi 802.11b/g/n connectivity. The computing power offeredby these units allows for complex digital signal pro-cessing to be carried out on the device, alleviatingthe need to transmit large amounts of audio data forprocessing on the project’s servers.

Whilst the device does not include an inbuilt micro-phone, USB I/O allows for the inclusion of a USBaudio device to handle all ADC work, thus providingthe means to connect a custom microphone solution.The USB audio device chosen for this application

Fig. 2: Bud Industries NBF-32004 ABS NEMA caseprofile

had to be compatible with Linux based Android de-vices, low in price, provide input gain control anda clean signal path. The device selected was theeForCity USB audio interface which retails for $4 asof June 2014. It provides a single microphone inputchannel with low noise and a software adjustable in-put gain stage. A swept sine signal was recordedvia the USB audio device, then compared with theinput signal which showed negligible differences.

2.2. CasingIn constructing the urban sound sensor device forlarge-scale deployment and fabrication, cost, envi-ronmental conditions, acoustic effects, and overallefficiency were considered. To move forward withthe prototype and properly account for these condi-tions, the first critical component specified was thecasing, responsible for housing all the sensor’s com-ponents. The casing utilized was a NEMA 1, 2, 4 &4X certified ABS plastic housing measuring 15cm x10cm x 7cm.

The casing provides environmental protection for thesensor components and allows for mounting on dif-ferent surfaces via attached metal brackets. The caseincludes a hermetically sealed, latch closed door al-lowing access to the sensor components inside. Itsprofile is shown in Figure 2. Additionally, the topand front facing panels of the casing were coatedwith impact absorbing felt to reduce structure bornerainfall noise transmitted to the microphone.

2.3. Microphone portThe MEMS microphone [9] was mounted on the un-derside of the casing, pointing down. In an effortto maximize frequency response and minimize di-rectivity, two microphone ports of 5mm and 10mm

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 3 of 9

Mydlarz et al. The design of urban sound monitoring devices

were drilled and tested under semi-anechoic condi-tions at the research laboratory at NYU Steinhardt.Figure 3 compares the frequency response effects ofeach port to the original MEMS response.

Figure 3 shows the minimal variation in frequencyresponse between the 5mm and 10mm port sizes.The boxed microphone exhibits a raised responsebetween 1kHz and 6kHz, corresponding to the firstfew dimensional resonances of the plastic enclosure.The rise in response after 10kHz, however, is a resultof the Helmholtz resonance created by the micro-phone’s inner chamber and PCB port [19]. Conse-quently, this results in an increase of perceived sibi-lance in the recordings, which can be filtered out inpost processing or on the RSD itself.

To determine the directivity of the unit with vary-ing port sizes, the sensor device was subjected totwo 3 second sine sweeps from 20-20,000Hz at 15 ◦

angle increments from 0-90 ◦ in azimuth and eleva-tion, with symmetry about 0 ◦ assumed. Speakerand room response were factored out by subtract-ing the same condition response of an EarthworksM30 measurement microphone assumed to be flat infrequency response from 20-20,000Hz. As expectedthe larger port size revealed marginally less directiv-ity which was more pronounced at high frequencies.Unfortunately, due to the proposed mounting loca-tions for these devices - lampposts and tree sidesat a height of roughly 3.5m - directivity will be in-troduced. Inevitably, the unit’s directivity will be-come more prominent as a result of the lamppost ortree’s shadowing effect. Though the sensor itself isomni-directional in nature, the physical attributes ofthe system limit its acoustic field of view to roughly90 ◦ in elevation from -90 ◦ to 0 ◦ and approximately270 ◦ in azimuth from 0 ◦ to 135 ◦ and 0 ◦ to -135 ◦,as shown in Figure 5.

To reduce the effect that this shadowing will haveon the capture of the overall soundfield, the place-ment of multiple RSD’s will need to be carefully con-sidered in order to provide sufficient coverage of aspaces sound environment.

2.4. Mounting conditionsAs previously mentioned, the microphone wasmounted on the underside of the casing, facing down.To alleviate structure borne sound and vibrationfrom rainfall, a sheet of vibration dampening vis-

0°

90°-90°

135°-135°

0°

-90°

Fig. 5: Mounted sensor acoustic field of view

coelastic urethane was placed between the micro-phone board [9] and the interior of the casing. Aswith the selection of the 10mm microphone port, twosheets of 0.1inch and 0.04inch in width, were testedand measured to select the less acoustically influen-tial design.

The addition of the 0.1inch sheet had the effect ofextending the port depth, resulting in an increasedresonance at high frequencies, as shown in Figure 6.Due to this, the thinner sheet of 0.04inch was used,shifting the resonance up to the more acceptable17kHz region. A further stage of testing to revealthe vibration reduction properties of this materialwill be carried out.

2.5. Wind and weather protectionA major concern in environmental acoustic monitor-ing is wind shielding, given the dynamic changes inwind currents. In order to mitigate the effect thismay have on the recorded audio signals, an electricfan was used to blow air across the microphone portin such a way that the maximum wind noise was pro-duced on the signal. Recordings were made withoutprotection and then using: craft felt, nylon tight ma-terial (two layers) and a section from a standard mi-crophone windshield. These materials were mountedflush on the outside of the housing, covering the mi-crophone port. RMS levels were calculated for a 3second portion of each recording where the signalwas in a steady state in terms of its level. Inter-estingly the microphone windshield produced morefrequency coloration than the other materials testedas shown in Figure 7, which had a minimal effect onfrequency response.

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 4 of 9

Mydlarz et al. The design of urban sound monitoring devices

102

103

104

−20

−10

0

10

20

Frequency (Hz)

Relative response normalized at 1kHz (dB)

5mm port

10mm port

Unboxed

Fig. 3: Comparison of device frequency response when unboxed and boxed with 5mm and 10mm port size

Felt provided the least amount of wind attenuationat 7dB compared to the windscreen and nylon re-duction of 9dB. Based on these findings, the double-layered nylon was chosen for our application, dueto its relatively high wind noise attenuation, low ab-sorbency, ease of mounting, and low cost. Excess wa-ter retention and potential freezing when soaked intoa windshield can have a profound effect on acousticsignal quality [20], which can be mitigated when us-ing a thin layer of windscreen such as the nylon se-lected for this application. In addition, a hydropho-bic coating was applied to the port area of the casingto provide further weatherproofing around the sensi-tive microphone area. This coating repels rainwaterso as to not allow any raindrops to collect on theunderside of the housing and potentially soak intothe microphones nylon wind shield.

2.6. Device powerOne of the benefits of deploying this system in an ur-ban environment is the potential for continuous 24/7power supply. However, though mounting the sen-sors on urban lamp posts is ideal in terms of locationand height, the poles themselves may only be pow-ered during nighttime hours. Because the first proto-type sensor unit relies on a constant 120v groundedpower supply, this means a stage of battery poweringis necessary to ensure a continuous 24/7 data streamcan be produced from the deployed sensors. The cur-rent draw from the sensor unit ranges from roughly200mA at idle to 750mA when processing and trans-mitting data. As such, a battery solution that could

provide at least 8000mAh to keep the sensor unit op-erational for the daylight periods was implemented.Furthermore, to account for any extraordinarily de-manding periods of operation, a 12000mAh lithiumpolymer battery pack was selected to provide the ex-tra supply needed for these unexpected situations.Further testing will be carried out on the batterypowered RSD to ensure its functionality and stabil-ity when deployed in varying environmental condi-tions, discussed in Section 3.

2.7. RFi mitigationThe internal components of the acoustic sensor hadto be strategically placed to minimize radio fre-quency interference (RFi) from the Android miniPC’s Wi-Fi antenna. Any low level audio compo-nents were located at the maximum distance fromany RF components, as well as the AC power sup-ply. Additionally, the length of the low level audiocabling was minimized to reduce the effects of RFiinterference on the system as shown in Figure 8.

Nonetheless, at the minute interval of data uploadto the server, RFi would taint the signal, due to thehigh power upload stage. To solve this, a small portwas drilled on the face of the casing to maximizethe distance between the antenna and audio com-ponents. In addition, the inside of the casing wascoated with grounded aluminum tape, thus creatingan RFi shield. After testing, this reduced RF in-terference. In future work, however, we propose theuse of a metal casing with an antenna port, ensur-

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 5 of 9

Mydlarz et al. The design of urban sound monitoring devices125 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

250 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

500 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

1000 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

2000 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

4000 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

8000 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

16000 H

z

0°15°

30°

45°

60°

75°

90°−90°

−75°

−60°

−45°

−30°

−15°

−20−15−10−5 0 5

10mm

5mm

Fig. 4: RSD directivity polar plots varying azimuth and port diameter

ing a fully sealed and RFi resistant housing. Lastlythe audio ground and power ground were matchedto eliminate excess RFi and PSU noise.

3. FUTURE WORK

In collaboration with the Brookhaven National Lab-oratories, a stage of environmental testing will beperformed on the complete sensor unit to ensureits resilience and consistent functionality under ex-treme variations in temperature. High ambient tem-peratures and direct sunlight falling on the casing

will raise internal temperatures, as well as the heatgenerated by the mini-PC itself, which is minimalas it only operates at upto 5 watts running at fullload. The current plastic prototype will be subjectedto temperatures in excess of 50 ◦C to firstly testfor hardware functionality and also to identify anyvariations in microphone sensitivity and frequencyresponse. Battery performance will also likely beaffected by lower (sub zero Celsius) temperatures,something that will have to be tested for to en-sure sustained operation of the lamppost mountedRSDs in winter periods. Heat ventilation is also an-

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 6 of 9

Mydlarz et al. The design of urban sound monitoring devices

104

−20

−10

0

10

20

30

Frequency (Hz)

Relative response normalized at 1kHz (dB)

0.1 inch

0.04 inch

Fig. 6: Effect of 0.1inch and 0.04inch vibration dampening sheet on sensor frequency response (6kHz-20kHz)

104

−20

−10

0

10

20

30

Frequency (Hz)

Relative response normalized at 1kHz (dB)

None

Felt

Nylon

Screen

Fig. 7: Effect of different wind shield strategies on sensor frequency response (3kHz-20kHz)

other key consideration that will be considered. Afan based system would obviously not be applicabledue to the noise they produce, however, a passivevent cut into the casing that relies on the wind toreplace the internal air within the housing may pro-vide an adequate solution. An aluminum casing hasalso been considered to further attenuate RFi noiseon the signal chain by providing a surrounding earthshield and common ground for all components. Theissue of solar heat gain from incident sunlight willalso be increased when using a metal case, however,the color that the casing is painted will have moreof an effect on this, so lighter shades will be used ifpossible.

A further stage of lab tests will be carried out to re-veal the complete audio systems response to varyingsignal levels in comparison to a calibrated Type 1sound level meter. This will identify any nonlinear-ities in level response and allow for the calibrationof each RSD before deployment. The varying envi-ronments these RSDs will be deployed in may alsorequire the addition of an automatic gain controlwhich can adapt to the ambient level of an urbanspace. For example, a busy roadside RSD may re-quire the USB audio device’s input gain to be re-duced to avoid any clipping on the resultant signal.The amount of attenuation applied could be learntover time by the system and would be stored along-

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 7 of 9

Mydlarz et al. The design of urban sound monitoring devices

Fig. 8: Internal prototype sensor layout (1 = miniPC, 2 = PSU, 3 = USB audio device, 4 = MEMSmicrophone)

side each level measurement.

The current stage of the project is focused on smallscale deployments in urban parks in NYC. Thesewill allow the project sensors and infrastructure tobe rigorously tested under outdoor weather condi-tions, providing valuable insights into Wi-Fi connec-tivity, equipment malfunction/damage, external sys-tem performance and power supply issues. The ob-jective acoustic data obtained through the project’ssensor network will be used to investigate connec-tions between spatial and temporal acoustic char-acteristics and existing geolocated datasets, such ascrime statistics, weather patterns, school attainmentmetrics, municipal/census data and public socialnetwork feeds, and real estate statistics which canprovide rich quantitative and contextual location-based information. Through a process of subjec-tive data collection in and around the deployed sen-sor locations, the effects and influences of the urbansoundscape can also be investigated in terms of itshuman perception and objective characteristics.

4. CONCLUSIONThe design of a resilient urban acoustic sensor pre-sented in this paper revealed, through a process ofacoustic measurements, the effects of differing portsizes, wind-shielding strategies, mounting design and

power supply conditions. In conclusion a weather-ized housing has been selected that exhibits minimalfrequency coloration on the recorded audio signalsat differing angles of incidence. Additionally, a windshielding solution has been devised resulting in areduction of wind noise with the added benefit oflow absorbency under humid conditions. Lastly, aseries of carefully arranged components and powersupplies, including a battery, were positioned withinthe plastic casing to minimize RFi, AC interference,and to generally isolate the audio components fromany detrimental noise.

5. REFERENCES

[1] A. Calixto, F. B. Diniz, and P. Zannin. Thestatistical modeling of road traffic noise in anurban setting. Cities, 20(1):23–29, 2003.

[2] EU Directive. Directive 2002/49/EC of the Eu-ropean parliament and the Council of 25 June2002 relating to the assessment and manage-ment of environmental noise. Official Journalof the European Communities, 189(12), 2002.

[3] E.A.M. Franssen, C.M.A.G. Van Wiechen,N.J.D. Nagelkerke, and E. Lebret. Aircraftnoise around a large international airport andits impact on general health and medicationuse. Occupational and Environmental Medicine,61(5):405–413, 2004.

[4] H. Ising and B. Kruppa. Health effects causedby noise : Evidence in the literature from thepast 25 years. Noise and Health, 6(22):5–13,2004.

[5] J. Kang. Urban sound environment. CRC Press,2006.

[6] J. Kang and M. Zhang. Semantic differentialanalysis of the soundscape in urban open publicspaces. Building and environment, 45(1):150–157, 2010.

[7] B. Li, S. Tao, R.W. Dawson, J. Cao, andK. Lam. A GIS based road traffic noise pre-diction model. Applied Acoustics, 63(6):679 –691, 2002.

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 8 of 9

Mydlarz et al. The design of urban sound monitoring devices

[8] M. Musick, J. Turner, and T.H. Park. Interac-tive auditory display of urban spatio-acoustics.In ICAD: 20th International Conference on Au-ditory Display, 2014.

[9] C. Mydlarz, S. Nacach, E. Rosenthal, M. Tem-ple, T.H. Park, and A. Roginska. The im-plementation of mems microphones for urbansound sensing. In AES 137th Convention, LosAngeles, USA, October 2014.

[10] M.E. Nilsson and B. Berglund. Soundscapequality in suburban green areas and city parks.Acta Acustica united with Acustica, 92(6):903–911, 2006.

[11] T.H. Park, B. Miller, A. Shrestha, S. Lee,J. Turner, and A. Marse. Citygram one: Vi-sualizing urban acoustic ecology. Proc. DigitalHumanities, Hamburg, Germany, 2012.

[12] T.H. Park, M. Musick, J. Turner, C. Mydlarz,J.H. Lee, J. You, and L. DuBois. Citygramone: One year later . . . . In 40th InternationalComputer Music Conference, 2014.

[13] T.H. Park, J. Turner, C. Jacoby, A. Marse,M. Musick, A. Kapur, and J. He. Locative soni-fication: Playing the world through citygram.In ICMC: International Developments in Elec-troAcoustics. ICMC, 2013.

[14] T.H. Park, J. Turner, M. Musick, J.H. Lee,C. Jacoby, C. Mydlarz, and J. Salamon. Sens-ing urban soundscapes. In EDBT/ICDT Work-shops, pages 375–382, 2014.

[15] R.K. Rana, C.T. Chou, S.S. Kanhere, N. Bu-lusu, and W. Hu. Ear-phone: an end-to-end participatory urban noise mapping system.In Proceedings of the 9th ACM/IEEE Interna-tional Conference on Information Processing inSensor Networks, pages 105–116. ACM, 2010.

[16] H. Slabbekoorn and M. Peet. Ecology: Birdssing at a higher pitch in urban noise. Nature,424(6946):267–267, 2003.

[17] S. Stansfeld, M. Haines, and B. Brown. Noiseand health in the urban environment. Reviewson environmental health, 15(1-2):43–82, 2000.

[18] C. Steele. A critical review of some traffic noiseprediction models. Applied Acoustics, 62(3):271– 287, 2001.

[19] J.W Weigold, T.J. Brosnihan, J. Bergeron, andX. Zhang. A MEMS condenser microphone forconsumer applications. In MEMS 2006 Istanbul19th IEEE International Conference, pages 86– 89, Istanbul, Turkey, 2006.

[20] R. Wright and G. Goulamhoussen. Influencesof weather, damage and ageing on the perfor-mance of outdoor microphone windshields. InProceedings of the Internoise, pages 1–10, 2011.

[21] W. Yang and J. Kang. Soundscape and soundpreferences in urban squares: a case study insheffield. Journal of Urban Design, 10(1):61–80, 2005.

AES 137th Convention, Los Angeles, USA, 2014 October 9–12

Page 9 of 9