Electrical Appliance Identification Through Signal Processing of Electrical Wave Signals

Clients

Techstack

Purpose

Knowing which appliances are drawing power usually means metering each one, which is intrusive and costly. This thesis set out to identify appliances from a single electrical feed — reading the shape of the current alone — and to prove it could tell five appliance types apart.

Description

We built a prototype that samples the current an appliance draws and a pipeline that turns those raw waveforms into images a deep-learning model can classify. The model was trained on data we gathered ourselves, with the full workflow covering collection, training, prediction, and flagging of unrecognized devices, and its accuracy documented through standard evaluation. Teaching a machine to recognize a device from its electrical fingerprint is the kind of signal-processing-meets-AI problem we like.