Real-time Recognition of Guitar Performance Using Two Sensor Groups for Interactive Lesson

Abstract

The accurate recognition of guitarist performance is challenging compared with other instruments because a guitar player typically plays several notes at once and uses both hands in different ways. In this paper, we propose a sensor-based guitar that consists of two groups of sensors. One sensor is used to recognize the fingering positions of the fretting hand, and the other is used to detect the guitar strings that are played by the picking hand. We design an embedded system for accurate sensing and propose a data analysis mechanism to precisely figure out the played pitch and the duration of notes using the sensed data. We realize our scheme as a high-quality prototype that detects guitarist performance with accuracy sufficient for the transcribing of a performance. We also present real application examples such as a rhythm game for interactive lessons and a music sharing feature with user created musical scores.