Reconstruction of a detection system for piano practice: Conceptualized model for performance errors

This study aimed to investigate the types of errors that adult beginners in piano playing make during their practice and how these errors can be classified as parameters for detection. Eight adults (six males, mean age = 42.5 years), beginners in keyboard playing, participated in the experimental study. All participants were asked to practice a short musical piece for 3 min on an electric piano after watching a demo. This 3-min practice session followed by a performance was repeated five times. After each session, each participant took part in a semi-structured interview. All performance data, including practice sessions, were recorded using two cameras and exported as MIDI audio data. The errors identified in the performance data were categorized as score-based (pitch substitutions, deletions, and timing errors) or non-score-based (insertion errors). The total number of errors from all participants across the performance trials was 540, of which 71.9% were categorized as insertion errors, consistently the highest category. The total number of errors decreased drastically from the first to the second session. The analysis of data from practice sessions indicated that participants tended to shift from a segmented strategy (e.g., hands separate) to an integrated strategy (e.g., run-through) over time.