Beat Detection for Automated Music Transcription: An Exploration of Onset Detection Algorithms.
Context:
This was completed as the project requirement for my Masters in Electrical Engineering. This was a solo project under the advisement of Dr. Mark Fowler
Abstract/Summary:
This project attempts to answer part of the question: Can music be reconstructed into sheet music from a recording through signal processing? This is a multifaceted problem that entails instrument identification, instrument isolation, pitch detection, dynamics etc. This project will explore a subset of this area, i.e., the fascinating world of beat detection, more commonly called onset detection.
While there are many methods that can isolate onsets, this project focuses only on frequency domain detection functions. There is, however, some discussion of time domain methods to give a taste of insight to frequency domain methods. The discussion is concluded with the results of several onset detection algorithms to give an empirical comparison between algorithms.