Isabelle Lin

How accurately can a machine learning model identify violin intonation errors in audio recordings?

Started May 6, 2025

Abstract or project description

The project consists of a program in which users can input an audio clip of their violin playing, a pdf of the sheet music and the tempo, then receive feedback on the intonation of the notes played in their audio. The program uses an existing pitch tracking machine learning algorithm to detect the pitches of each note in the audio. It also uses an existing software to convert the sheet music into a digital format. The difference between the audio pitch and the correct pitch is then used to produce feedback for the user through an existing AI model, which lists which notes were out of tune and whether each note was sharp or flat, as well as practice tips for the user. This program can be used by both beginner and advanced violinists as a practice tool to improve their intonation when playing, as well as develop their ability to recognize off-pitch notes. The result of the project includes a website that can be locally hosted and a research paper.