About
The Meter Whisperer
AI-powered vibration intelligence for detecting abnormal water activity.
What it does
The Meter Whisperer listens to your water meter using a clip-on accelerometer. Mechanical activity inside the meter — caused by water movement — produces vibration. The accelerometer converts those vibrations into a signal, and a trained machine-learning model turns that signal into a calm, actionable explanation.
How it works
- Capture. A clip-on accelerometer picks up vibration from the water meter.
- Extract. Time-domain and frequency-domain features are computed from the signal window.
- Classify. A Random Forest model classifies the activity as normal or a specific leak type.
- Explain. The app surfaces severity, confidence, probable reasons, and suggested actions.
Supported leak types
- Normal — no abnormal activity detected.
- Tap / faucet leak — intermittent use or drips.
- Washing machine leak — cyclical fill/drain activity.
- Toilet leak — persistent pulsing from refill cycles.
- Pipe / fitting anomaly — loose mounts or pressure changes.
Data format
Recordings are CSV files with the columns:
time_s, acc_x_m_s2,
acc_y_m_s2, acc_z_m_s2,
and acc_abs_m_s2. A sample file is available from the
live-preview page.
Prototype — real-world performance requires validation with labelled sensor recordings.