The design and calibration of low cost urban acoustic sensing devices. SONYC Sounds Of New York City

Transcription

1 The design and calibration of low cost urban acoustic sensing devices SONYC Sounds Of New York City C. Mydlarz NYU CUSP C. Shamoon NYC DEP M. Baglione, M Pimpinella The Cooper Union cmydlarz@nyu.edu

2 Sounds of NYC Motivations Many 311 complaints are about NYC$311$Noise$Complains$(20103current)$ Loud%Music/Party% Loud%Talking% Noise:%Construc9on%Before/A=er% Hours%(NM1)% Car/Truck%Music% Noise Noise,%Barking%Dog%(NR5)% Noise:%Construc9on%Equipment%(NC1)% People%Created%Noise% Engine%Idling% Noise:%Jack%Hammering%(NC2)%

3 Sounds of NYC Project overview Sensor Network & Citizens Citizen Science City Agencies Machine Listening City Experts PHYSICAL Noise Mission Control CYBER Data Analytics & Visualization

4 Sounds of NYC Applications Map acoustic heartbeat of the city Provide 24/7 real-time objective data on the NYC sound environment Soundscape data analysis to inform noise pollution patterns Exploratory research investigating trends between noise patterns and existing geo-located datasets Working with city agencies to optimize resource allocation Consumer and commercial applications in the identification of suitable acoustic environments, e.g. consumer decision making

5 Sounds of NYC Sensor purpose and use Each acoustic sensing device continuously monitors its surrounding sonic environment Advanced DSP capabilities onboard Low and high level descriptors extracted Data transmitted over Wi-Fi Initial deployment of >100 nodes across NYU campus buildings

6 Software Core sensor applications Capture app Capture audio & save to disk in chunks Compress audio using FLAC codec Encrypt audio Feature extraction Source identification Controller app Secure & robust audio/feature upload Status ping & command execution Wi-Fi awareness and adaptation Clock synchronization

7 Hardware Computing core Tronsmart MK908ii Quad-core 1.6GHz ARM A9 Cortex processor 2GB RAM 8GB expandable internal storage Wi-Fi, Bluetooth, USB I/O Very low cost at $50 Optimized DSP routines Android based but flashed with Linux Picuntu

8 Hardware MKI sensor breakdown $ Omni analog MEMS microphone $ USB CODEC $ Linux based mini PC $ Apple low noise PSU

9 Hardware MKI sensor hardware Advanced, low cost ( $100) sensing device Analog MEMS microphone Quad core mini PC, 2GB RAM, 8GB int. Can produce data to OSHA/ANSI/IEC standards

10 Measurements Overview Analog MEMS microphone and USB CODEC (DUT) subjected to subset of: IEC (2013) Sound Level Meter Periodic Tests Can a low cost MEMS solution produce type 2 level acoustic data? Is this enough for city agencies tasked with enforcing noise complaints? Type 1 Larson Davis 831 used as reference To determine DUT s ability to perform within type 2 specs an adjusted tolerance was calculated: type 2 - type 1 tolerance = adjusted tolerance ±1dB ±1dB ±2dB

11 Measurements Overview Carried out under low level (<20dBA) anechoic conditions at the Cooper Union Vibration & Acoustics laboratory Consistent environmental conditions All reported values an average of four repeated measurements Compensation for MEMS & CODEC frequency response

12 Measurements IEC Sound Level Meter Periodic Tests Self generated noise Minimum SPL the system can reliably detect: 29.9dBA Dynamic range: 88.1dBA Signal to noise ratio: 64.1dBA Long term stability 30min 1kHz sine 94dBA Type 2 specs state max start-end of ±0.2dBA DUT showed difference of 0.07dBA Acoustic level response linearity Linearly increasing sine waves at octave frequencies from 31.5Hz to 8kHz DUT operates at type 2 levels above 41.2dBA between 31.5Hz to 8kHz

13 Measurements Acoustical signal tests of a frequency weighting Ability to produce reliable dba output at different frequencies Steady state sine waves at octave frequencies from 31.5Hz to 8kHz IEC Sound Level Meter Periodic Tests

14 Measurements Toneburst response Testing the response to transient SPLs IEC Sound Level Meter Periodic Tests Subjected to 4kHz sinusoidal tonebursts, varying in duration from 1000ms down to 0.25ms

15 Measurements Overview Findings Analog MEMS microphone able to meet subset of IEC periodic sound level meter tests at type 2 level PSU noise raises overall noise floor of system MEMS microphones extremely consistent Further work Perform IEC test set including directionality of entire sensor node Long term external comparison testing Digital microphone solution

16 Sounds of NYC MKII sensor hardware Advanced, low cost ( $100) sensing device Low noise digital MEMS microphone Quad core mini PC, 2GB RAM, 8GB int. Can produce data to OSHA/ANSI/IEC standards

17 Hardware Development of a fully digital solution Very small single board solution USB powered Very high PSU noise rejection High sensitivity Hot swappable Total cost of parts: $7.00 Digital microphone ST MP34DB01 PDM I 2 S USB audio Digital MEMS PDM to I 2 S converter I 2 S to USB digital audio bridge Core computing device

18 Hardware MKII sensor breakdown $ Omni digital MEMS microphone $ USB audio bridge $ Linux based mini PC $3.00 Power supply unit

19 Sounds of NYC Summary The Sounds Of New York City Project is building a novel cyber-physical system including a scalable network of low cost sensors for large scale, realtime urban acoustic monitoring Our devices are capable of producing calibrated and accurate SPL data Each sensor node has advanced processing capabilities for in-situ feature extraction and sound source classification Our Noise mission control will be utilized by decision-makers at city agencies to strategically deploy the human resources at their disposal to mitigate noise

20 SONYC Sounds of New York City Real-time acquisition & mapping of spatio-temporal, urban acoustic data Deployment of remote sensing devices based around consumer technologies Standards capable calibrated SPL data acquisition Strategic partnerships with city agencies

21 SONYC Sounds of New York City Real-time acquisition & mapping of spatio-temporal, urban acoustic data Deployment of remote sensing devices based around consumer technologies Standards capable calibrated SPL data acquisition Strategic partnerships with city agencies