The University of Auckland

Project #109: A Little Midnight Music – Sound Processing and Transcription App



Learning music can be hard, especially music theory. Imagine you want to write your own music. Unfortunately, the process of coming up with the right notes and then noting them down before you forget can be long and tiresome. Alternatively, if you have a song you really want to learn but for the life of you, you can't find the music sheets anywhere. You may not know all the chords and notes to the song you want to learn and trying to figure these all out by ear can be tedious and difficult. You’ll have to go into a specialised environment or ask someone who understands music theory better to compare for you. The aim of this project is to develop a tool that uses machine learning models to analyse the individual notes of classical piano music and automate the transcription of the notes to digital music sheets. This would greatly simplify the process of reproducing the music being performed, so students of music theory are able to automatically transcribe their favourite music into piano sheets they can learn off of. In addition, clients that wish to compose their own music can use the tool to transcribe their music. The final product will be comparable to the app Yousician.



•          An investigation into currently used spectral analysis techniques

•          An investigation into music theory and identifying chords and notes

•          Development of machine learning models to identify individual notes in music signals

•          Development of an automatic documentation tool that uses these notes to form a transcription of the music

•          An investigation into the efficacy of the tool in terms of its ability to recreate the notes for the music

•          An investigation into the usage of the tool as a mobile application

•          Formal analysis and assessment of the tool









Lab allocations have not been finalised