Sayed Armin Hosseini - Isfahan University of Technology, Isfahan, Iran
Mohammad Ali Montazeri - PhD, Isfahan University of Technology, Isfahan

In this paper we propose a new algorithm and introduce a data structure for music pattern mining. In the proposed method, we search both vertical and horizontal patterns in some pieces of music and classify each piece to specific classes. A learning process based on fuzzy neural networks is developed so that in the learning phase the system is trained with the vertical and horizontal patterns. In the test phase a new musical piece is introduced to the system. The system locates the patterns based on their characteristics and classifies the piece of music applying its dynamic rules and employing its knowledge base. The structure of this paper is as follows. First we briefly introduce a stochastic analysis of music and introduce a new mathematical model and develop its data structure. Then we will demonstrate the proposed algorithm for music pattern mining. Simultaneously a case study has prepared for explaining these concepts. This is an innovative practical way that can be used both in multimedia systems and in computer-aided music composition systems.

Computer-aided Music Analysis, Music Theory, Fuzzy Neural Networks, Machine Learning, Classification Algorithms, Pattern Recognition