Proceeding of the past Conferences /

Below you can read the papers of the past Conferences


New Music Concepts 2021

Deep Learning based Detection of GPR6 GTTM Global Feature Rule of Music Scores

Part of the New Music Concepts book series (NMC volume 8)
ISBN: 978-88-944350-2-3

Author(s): Yan-Ru Lai, Alvin Wen-Yu Su

Abstract: Rules such as phrasing, articulation and intonation are important to musical performance. Local boundaries of a piece of musical score are closely related to these rules. For a computer program performance model, it is desired to detect such local boundaries to manipulate the above rules. The grouping pref- erence rules (GPRs) proposed in the generative theory of tonal music (GTTM) were proven to be effective in this respect. In the past two decades, computer automatic detections of GPRs have been studied. Recently, machine learning techniques were proposed for accurate localization of GPR2 and GPR3. GPR6 is another important feature for local boundary detection. The major difficulties to localize GPR6 lie in the insufficient amount of labeled dataset if deep learning models are used. In this paper, an algorithm is proposed to generate tens of thou- sands of scores with reliable GPR6 labels. These automatically generated scores are used as the pre-training dataset for the bidirectional long short-term memory (BLSTM) networks. Then, 267 manually labeled data set is used to test the model. The experimental results show that the proposed method is significantly superior to the existing ATTA model.

Keywords: A Generative Theory of Tonal Music, Music Boundary Detection, Grouping Preference Rule, Deep Learning, Bidirectional Long Short-Term Memory