USING DEEP FEATURES FOR AUDIO CLASSIFICATION

This extended abstracts describes our system for audio classification and tag classification submitted to MIREX 2010. Our system uses features learned with a deep belief network (DBN) and uses support vector machines (SVMs) classifiers on these features. 1. OVERVIEW In this paper, we describe our system for audio classification and tag classification submitted to MIREX 2010. This system relies on a deep belief network to extract features. The system is explained in more details in [2]. We describe how we use this network in section 2. In section 3, we describe the classification module. We conclude in section 4.