Traditional pianos are inherently non-portable, restricting everyday accessibility and on-demand creativity. Existing portable alternatives, largely vision-based with external cameras, suffer from limited range, occlusion, and unreliable contact detection. We present PianoBand, a wrist-worn system integrating an IMU, a miniature under-wrist RGB camera, and a printed keyboard sheet augmented with fiducial markers for reliable key mapping on any flat surface. Powered by a lightweight real-time IMU–vision pipeline, PianoBand enables high-fidelity piano interaction, supporting single notes, multi-finger chords, flexible fingering, dynamic velocity, and preliminary articulation techniques. Technical evaluation showed robust tap detection (over 99% accuracy) and accurate fingertip localization (8.90 pixels error), enabling precise note mapping. A comparative user study (N=15) further evaluated system performance, reporting high note accuracy, comparable to roll-up pianos and outperforming an XR piano, along with high ratings for portability, expressivity, and extensibility. Expert interviews highlighted broad application opportunities for piano-based experience and music creation, suggesting future design directions.
ACM CHI Conference on Human Factors in Computing Systems