A supervised learning AI model for automated holistic vocal performance feedback

Scalable and effective music education requires giving fast and accurate feedback on student audio performances. Manual feedback from teachers is often subjective and, therefore, sometimes inaccurate. Current technological feedback mechanisms evaluate whether a student is correct on a single note rather than the entire music piece, lacking cumulative or numerical feedback. This paper presents a machine learning model for automatically grading vocal music recordings cumulatively on pitch and rhythm, given a reference piece of music. The model predicts a numerical grade for the performance of a reference piece of music and employs a correspondence algorithm to provide granular feedback (e.g., pitch mismatch, timing errors, missed notes). When tested on the MAST Melody Dataset, the ML model achieved more than 80% accuracy in scoring performances where human judges had consensus. Additionally, analysis revealed 20.85% outlier samples in human grading, highlighting subjectivity in manual assessment. The proposed system demonstrates the feasibility of objective vocal performance evaluation, while exposing limitations in current grading practices..