OnoSensor, a system developed by UMRAE, is an environmental acoustic sensor that costs less than €500 and consists solely of readily available off-the-shelf hardware and a fully featured, 100% open-source software solution. As part of the “Rail4Earth” research project, 50 of these sensors were deployed across France and collected no less than 1 TB of data (acoustic and audio indicators).
This platform aim to define an open source hardware and software stack for the deployment of a large set of advanced noise sensors.
Installed on a Raspberry Pi (3+, 4, 5) device.
- Computation of noise indicators LZeq LAeq LCeq with iec_61672_1_2013 spectrum analysis.
- Recognition of 1Khz 94dB calibration device and storage of microphone sensitivity
- Embedded sound source classification using tensorflow light and Yamnet with 521 tags
- Optional recording and encryption (public/private key) of short audio sequence triggered by sound classification tags
- Storage and transfer of all data into bulked compressed json files, processed by big-data analysis open source platform Elastic-Search/Kibana.
- Visualization through Kibana
- All services on server(s) through a straightforward Docker compose configuration.
- Custom html public dashboard with Python FastApi
Stack with Raspberry Pi 3B+ with PiJuice and 4G/GPS Header Telit
Microphone on Tripod with RPI in a case
Web dashboard that display sensors connection status
Web dashboard with integrated audio player for encrypted audio recordings
Kibana dashboard with various analysis
Kibana map with location of sensors