HONS 06/07
Detecting Advertising in Radio using Machine Learning
Robin Muller-Cajar
Department of Computer Science
University of Canterbury
Abstract
We present an algorithm that can distinguish between advertising and music without
understanding it, by extracting key attributes from a radio audio stream. Our
method combines advanced filtering of an audio stream with machine learning algorithms
to recognise the filtered variables. The result is lightweight enough to run
on an embedded processor, and could thus be used to create a device that gives the
listener the ability to filter advertising from radio broadcasts.