Electric Guitar Player Classification Using Motion Sensor Data and Information Quantity

This study proposes a novel method for classifying electric guitar players based on picking motion data captured by a wrist-mounted motion sensor. We apply an information quantity-based approach that quantifies similarity between symbolic representations of time-series motion data. A symbolic transformation using Symbolic Aggregate approXimation (SAX) enables this conversion. The method is compared to a baseline classification using Dynamic Time Warping (DTW) with a 1-Nearest Neighbor (1-NN) classifier. Experiments were conducted using motion data collected from 12 participants across multiple guitar lessons. The proposed method achieved a maximum classification accuracy of 81%, demonstrating competitive performance with DTW while offering significantly faster computation. These results indicate that individual differences in picking motion can be effectively captured by information-based methods, particularly as players gain experience. This technique shows promise for applications in performance analysis and real-time feedback systems.