Gesture Analysis and Recognition of Bow-Stringed Instruments using Motion Capture System Qualisys

TitleGesture Analysis and Recognition of Bow-Stringed Instruments using Motion Capture System Qualisys
Publication TypeMaster Thesis
Year of Publication2013
AuthorsChen, J.
AbstractGesture analysis of instrumental performance has been a hot-discussed topic for a long time. Previous work mainly focus on violin performance. In this paper, we use a optical motion capture system Qualisys to capture gesture data of violin, viola and cello. For each instrument, we record 5 techniques (détaché, martelé, spiccato, staccato and staccato down) with 3 dynamics (piano, mezzo forte and forte). After calibration, bow descriptors computation and score-alignment, we use a vector of bow velocity to represent a stroke. We then categorize strokes according to their techniques, dynamics and durations. After analysis of each category, we propose five features to differentiate them. Using LIBSVM and 4-fold cross validation, we perform several classification tasks from different angles to classify them. Result shows that our method proves to be effective to differentiate strokes of different techniques with different dynamics.
Final publicationhttps://doi.org/10.5281/zenodo.1162455
intranet