Peter
Edvardsson
AI for Musicians - Performance Audio Evaluator STEM
Abstract profile. Full document pending author claim.
Authors:
Peter Edvardsson
Date Created:
Not specified
Course Title:
Professor:
Not specified
About Paper:
This project aims to use digital technology to transform how musicians practice, posing particular utility to intermediate music learners. Private tutoring costs force such students to balance improvement with expenditure, necessitating an infrequency of lessons that leaves many hours of unsupervised practice in-between. To bolster the productivity of this practice time, our research aims to develop and deploy a mobile app that tracks the performance of a string player through given sheet music, providing visual in-score feedback of their pitch accuracy, intonation, and tempo. We explore a score-alignment based approach to recorded performance assessment, deriving reference audio from the score by synthesizing its MIDI data or adapting a professional recording. Using the Dynamic Time Warping algorithm, we align this reference with live performance audio. The resulting warping path enables us to estimate each note's unknown performance timestamp based on the known reference timestamp. Leveraging the calculated score position alongside fundamental frequency estimation algorithms, we determine whether the performance is too sharp or flat at any given note, evaluating the intonation of the performer. We visualize this feedback through in-score annotations. The empirical accuracy of our approach can be measured manually. We calculate alignment error by hand-labeling the onsets of live notes and taking their difference from the warping-path-informed onset predictions. Our intonation calculations can be checked against the live spectrogram. Furthermore, we intend to conduct user studies at the end of summer to receive feedback from other musicians assessing the usability, accuracy, and learning impact of our research. Keywords: Signal Processing; Score Following; Performance Error Detection; Musical Performance; Music Processing † Presenting Undergrad Author; ‡ Contributing Undergrad Author; * Undergrad Acknowledgment
Source:
Purdue University / 2025
Topics:
No topics listed
Co-authors:
Peter Edvardsson