Measuring and recognising sound with smart sensors
The system combines a microphone, sound detection algorithm and microcontroller running locally on TinyML. It continuously measures sound, stores peaks and anomalies in .wav format and automatically generates a log file with time, location and reliability of the measurement.
The solution runs on a Seeed ReSpeaker connected to a Raspberry Pi, complemented by a Real Time Clock and storage on USB. By running anomaly detection locally, the system saves power and storage space. Only relevant sound clips are captured. In short: a plug & play solution with minimum cost and maximum focus.
This changed the use of
TinyML to the nuisance
Automatic detection of noise pollution based on abnormal peak patterns, recognition of noise origin (direction), proven deployment in industrial environments and potential to reduce complaints by 20% within a year.
The combination of localisation, logging and live feedback enables Tata Steel to intervene more precisely and quickly in nuisance situations. This not only results in fewer complaints, but also contributes to a better relationship with local residents and internal safety.
more info or register?
This solution shows that complex challenges, can be addressed smartly, affordably and scalably with TinyML. It also offers opportunities for smaller companies with noise-sensitive processes: from monitoring to predictive maintenance. Fieldlab Edge AI helps you step by step, from idea to working prototype. Want to know more? Contact the Fieldlab Edge AI for Smart Industry at Andre Gerver.
CONTACT
Andre Gerver
📧 a.gerver@techport.nl