SymCHM—An Unsupervised Approach for Pattern Discovery in Symbolic Music with a Compositional Hierarchical Model

Abstract

This paper presents a compositional hierarchical model for pattern discovery in symbolic music. The model can be regarded as a deep architecture with a transparent structure. It can learn a set of repeated patterns within individual works or larger corpora in an unsupervised manner, relying on statistics of pattern occurrences, and robustly infer the learned patterns in new, unknown works. A learned model contains representations of patterns on different layers, from the simple short structures on lower layers to the longer and more complex music structures on higher layers. A pattern selection procedure can be used to extract the most frequent patterns from the model. We evaluate the model on the publicly available JKU Patterns Datasetsand compare the results to other approaches.

An Analysis of Rhythmic Patterns with Unsupervised Learning

“…To improve its performance in learning complex rhythms, we will extend the model with accent encoding in part structures. Additionally, we will extend the model with pitch information, thus merging it with the compositional hierarchical model SymCHM [25] used for pattern discovery in symbolic music. While the SymCHM discovers repeating melodic patterns in data, it ignores the rhythmic information.…”

“…To bridge the gap between both types of approaches, we propose an unsupervised learning approach for modeling rhythm. The model is derived from the compositional hierarchical model (CHM) that was previously applied to tasks such as multiple fundamental frequency estimation [24] and pattern discovery from symbolic music representations [25]. The presented model-the compositional hierarchical model for rhythm analysis (RhythmCHM)-unsupervisedly learns a multi-layer hierarchy of rhythmic patterns ranging from simple structures on lower layers to more complex pattern compositions on higher layers.…”

“…In this section, we present the compositional hierarchical model for rhythm modeling (RhythmCHM). The compositional hierarchical model for music information retrieval was first introduced by [27] and has been applied to several MIR tasks, including automated chord estimation and multiple fundamental frequency estimation [24], as well as finding repeated patterns and sections in symbolic music representations [25].…”

A Comparison of Human and Computational Melody Prediction Through Familiarity and Expertise

“…In music theory and information retrieval, music patterns and their repetitions have been studied by a number of groups, both in theory (e.g., Lerdahl and Jackendoff, 1983;Margulis, 2013) and computational approaches (e.g., Marsden, 2010;Ren, 2016;Pesek et al, 2017b). Several tasks emerged throughout the years in the Music Information Retrieval Evaluation eXchange (MIREX), which is a community-based framework for formal evaluation of algorithms and techniques related to music information retrieval (MIR) (Downie, 2008).…”

Special Issue on “Sound and Music Computing”

“…To validate the database they also introduce a classifier that performs better than a traditional Mel-frequency-cepstral-coefficient classifier. The article by Pesek et al [11] introduces algorithmic concepts for modeling and detecting recurrent patterns in symbolically encoded music. Given a monophonic symbolic representation of a piece of music, the algorithm outputs a hierarchical representation of melodic patterns using an unsupervised learning procedure without the need of hard-coded rules from music theory.…”