Unveiling the Hierarchical Structure of Music by Multi-Resolution Community Detection

Abstract

Human perception of musical structure is supposed to depend on the generation of hierarchies, which is inherently related to the actual organisation of sounds in music. Musical structures are indeed best retained by listeners when they form hierarchical patterns, with consequent implications on the appreciation of music and its performance. The automatic detection of musical structure in audio recordings is one of the most challenging problems in the field of music information retrieval, since even human experts tend to disagree on the structural decomposition of a piece of music. However, most of the current music segmentation algorithms in literature can only produce flat segmentations, meaning that they cannot segment music at different levels in order to reveal its hierarchical structure. We propose a novel methodology for the hierarchical analysis of music structure that is based on graph theory and multi-resolution community detection. This unsupervised method can perform both the tasks of boundary detection and structural grouping, without the need of particular constraints that would limit the resulting segmentation. To evaluate our approach, we designed an experiment that allowed us to compare its segmentation performance with that of the current state of the art algorithms for hierarchical segmentation. Our results indicate that the proposed methodology can achieve state of the art performance on a well-known benchmark dataset, thus providing a deeper analysis of musical structure.

The Harmonic Memory: a Knowledge Graph of harmonic patterns as a trustworthy framework for computational creativity

“…where 𝑥 𝑖 ∈ X is a column vector; 𝑛, 𝑚 ∈ [0 : 𝑑 − 1]; 13 is a normalisation factor (the maximum TPS value); and the subtraction from 1 is used to obtain a similarity score from a distance measure. Self-similarity matrices have been extensively used for structure analysis, due to their ability to reveal nested structural elements [27,17] As can be seen from Figure 2, block-like structures are observed when the underlying sequence shows homogeneous features over the duration of the corresponding segment. Often, such a homogeneous segment is followed by another homogeneous segment that stands in contrast to the previous one.…”

Deep Learning for Musical Form: Recognition and Analysis

“…Classical music form analysis is a difficult task well-known to musicians and audio researchers alike, given the overall complexity of musical form design and the theoretical knowledge required to analyze pieces of music both at the phrase-and part-level, and large form classification. While there have been several various attempts utilizing novelty methods on audio features [6], community detection algorithms [1], and neural networks [7], none have proven to be sufficient for truly complex analysis tasks. This issue is especially true for more advanced musical forms such as sonata-allegro form where repetition of musical features is not nearly as clear as simpler Recurrent Neural Networks (RNNs) [7], resulting in an upper limit to the accuracy a model could achieve regardless of the training material [5].…”

“…This article presents a discussion on the challenge of automatic musical structure detection in audio recordings and the issue of most current algorithms being only able to produce flat segmentations which lack the ability to segment the music across multiple levels to reveal the piece's hierarchical structure [1]. The authors propose a new approach based on multi-resolution community detection and graph theory to create a new unsupervised learning task, which can perform both boundary detection and structural grouping without the need for specified constraints that limit the output segmentation.…”

“…• Plagal Cadence (PC) -A final sounding cadence, typically a IV-I progression which often follows a PAC but is less important and conclusive than an authentic cadence. df.loc[indx] = ["", "", c_flname, c_sngdur, c_slmmls[0], c_slmmls [1], c_scmcos[0], c_scmcos [1], 251.…”

Supervised Metric Learning for Music Structure Features

“…Devising a suitable feature for MSA is challenging, since so many aspects of music-including pitch, timbre, rhythm, and dynamics-impact the perception of structure [4]. Some methods have aimed to combine input from multiple features [5], but this requires care: MSA researchers have long been aware that structure at different timescales can be reflected best by different features (see, e.g., [6]).…”