Towards the characterization of singing styles in world music

Abstract

In this paper we focus on the characterization of singing styles in world music. We develop a set of contour features capturing pitch structure and melodic embellishments. Using these features we train a binary classifier to distinguish vocal from non-vocal contours and learn a dictionary of singing style elements. Each contour is mapped to the dictionary elements and each recording is summarized as the histogram of its contour mappings. We use K-means clustering on the recording representations as a proxy for singing style similarity. We observe clusters distinguished by characteristic uses of singing techniques such as vibrato and melisma. Recordings that are clustered together are often from neighbouring countries or exhibit aspects of language and cultural proximity. Studying singing particularities in this comparative manner can contribute to understanding the interaction and exchange between world music styles.

Finding Meanings in Low Dimensional Structures: Stochastic Neighbor Embedding Applied to the Analysis of Indri indri Vocal Repertoire

“…The t-SNE embedding and the use of MLP allowed an efficient analytical performance: Our results indicate that it was possible to automatically identify vocal types by using a dataset consisting of high-dimensional vector representations of objects, assigning similarities between those objects as conditional probabilities [10]. Still, although both t-SNE [15,16,17,18,19] and neural networks [50,51] are widely used to analyze acoustic characteristics in a wide range of research fields, ours represents the first attempt to combine these kinds of computational tools and apply them to the identification of vocal repertoire in nonhuman primates. Our findings support what was found in a previous analysis on indris’ vocal repertoire [39].…”

“…Since its introduction, due to its flexibility, efficiency, and accuracy, various studies successfully applied the t-SNE and its extensions to the visualization and the classification of different kinds of objects: Paintings [11], single nucleotide polymorphisms (SNPs) [12], data collected by computer-aided diagnosis systems (CADx) [13], and high-dimensional cytometry data in mouse tumors [14]. t-SNE has also been employed in several studies investigating a wide range of acoustic aspects: To solve problems in the estimation and characterization of pitch content in musical audio [15], to examine similarities among words and phrases in natural language processing [16], to visualize relevant selected features of audio data [17], to characterize singing styles and to discriminate vocal and non-vocal contours [18], and to perform a dimensionality reduction in the building of an efficient technique of speaker recognition [19]. Still, this promising technique has hitherto rarely been applied to the study of animal behavior in general (stereotyped behavior of freely moving fruit flies, Drosophila melanogaster ) [20], and never to investigate animals’ vocal behavior.…”

Addressing The Confounds Of Accompaniments In Singer Identification

“…Code available at https://github.com/bill317996/ Singer-identification-in-artist20. 1 Features extracted from the melody contour have been shown useful in many other MIR tasks [23,[25][26][27]. However, we note that most existing work used hand-crafted features, rather than features learned by a neural network.…”

Towards Cross-Cultural Analysis using Music Information Dynamics

“…In [5], hand-crafted audio features and traditional machine learning techniques are used to construct a similarity measurement to find outliers among music examples from various cultural origins and analyze their characteristics. In [6], hand-crafted vocal features are devised to study different singing styles from different cultures, and in [7], the same work is extended to classify concert non-western music recordings. Various non western or folk music corpus are evaluated and compared in [8].…”