Musical instrument recognition has recently received growing attention from the research community and music industry. It plays a significant role in multimedia applications. Many approaches have been proposed to classify musical instruments. Particularly, the articulation refers to the style in which a song's note is played. In this paper, we propose a new avenue for musical instrument classification into two categories: Pizzicato and Sustain articulations. New features derived from chromagram contours are investigated by using the classical invariant moments. A comparison with a reference system using a feature vector constructed from 38 feature parameters and using k-NN classifier is provided. The standard RWC database is used for all experiments.
展开▼
