A Machine Learning Approach for Analyzing Musical Expressions of Digital Piano Performance of Online Learning System

• Main Authors: Pao-Te Tsai, 蔡寶德
• Other Authors: Kuo-Liang Ou
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/07560546224059306608

碩士 === 國立新竹教育大學 === 資訊科學研究所 === 97 === This paper proposed a machine learning approach for analyzing teachers’ expert knowledge of classifying students’ piano performance into approximate expression categories. In traditional piano learning environment, students are usually confused when learning the expressive performance because of teachers’ subjective intention difference on the same performance. In this thesis, teacher models were built by analyzing teachers’ classification rules. By replaying students’ performances and read teachers’ suggestions in graphical and textual modes which are generated automatically by the teacher model, students could understand the nuance of performance features on each expression. Nine teachers and ten students joined this experiment. Forty six piano performances were recorded for constructing the teacher model. The average accuracy of teacher model for classifying performance expression is 89.1%. It only takes 9.84 seconds to build the teacher model and 0.1 second to automatically classify each recorded performance. This thesis proposed a highly accurate and fast-processing-rate analyzing system to assist teaching and help students understanding musical expression.