Acoustic anomaly detection
Microphones and audio models detect leaks, bearing wear and abnormal sounds from pumps and plant.
The problem
Some faults are audible long before they show on other sensors, but nobody is listening.
Data we need
- Audio from pumps, pipes and plant
- Labelled fault recordings
- Operating conditions
How it works
- Place acoustic sensors
- Record normal and fault conditions
- Train anomaly models
- Run on edge devices
- Alert maintenance teams
What it changes
Targets are agreed with you before we build, and measured on your data before release.
- Faults heard early
- Leaks located faster
- Low-cost sensors added to existing assets
Speech and audio
Transcription, acoustic monitoring and voice interfaces for noisy real-world environments.
- Speech-to-text
- Acoustic classification