Complexity of Vocal Vibrato in Opera and Jazz Recordings: Insights From Entropy and Recurrence Analyses

Gerardo Acosta Martínez; Helena Daffern

doi:10.1016/j.jvoice.2023.11.020

Complexity of Vocal Vibrato in Opera and Jazz Recordings: Insights From Entropy and Recurrence Analyses

J Voice. 2023 Dec 22:S0892-1997(23)00380-6. doi: 10.1016/j.jvoice.2023.11.020. Online ahead of print.

Authors

Gerardo Acosta Martínez¹, Helena Daffern²

Affiliations

¹ AudioLab, School of Physics, Engineering and Technology, University of York, York, United Kingdom. Electronic address: gerardo.acosta@york.ac.uk.
² AudioLab, School of Physics, Engineering and Technology, University of York, York, United Kingdom.

PMID: 38142185
DOI: 10.1016/j.jvoice.2023.11.020

Abstract

Introduction: Vibrato is an oscillation in frequency, intensity, and timbre of the singing voice. Previous studies have found a relationship between its periodicity and perceived quality. The diversity of vibrato enriches the music and singer's performances but create challenges for quantifying and capturing the characteristics that contribute to achieving these expressive goals. Vibrato tones have been addressed using rate, extent, jitter, and shimmer; however, these do not necessarily capture relevant complex time-varying features. This paper applies techniques from disciplines that specialize in periodicity and complexity to provide insight into vibrato characteristics not yet understood.

Objectives: This study aimed to assess whether nonlinear metrics are relevant features in characterizing and illustrating differences in vibrato behavior in opera and jazz singing, as well as considering the relationship of nonlinear metrics to other vibrato parameters.

Methods: Vibrato tones from published music material of world-class singers from opera and jazz were analyzed with entropy, recurrence, and the established parameters of rate, extent, jitter, and shimmer. Dimensionality reduction was employed to consider the relationship and significance of each of the metrics in collectively characterizing vibrato.

Results: The principal component, explaining 40% of variability, had positive weights of determinism and line length derived from recurrence while having negative weights of rate, shimmer, and sample entropy. Using these components, the vibrato tones from opera compared to jazz singing were found to be more regular and had lower rate and extent, and it was possible to spotlight singers and notes with high periodicity.

Conclusion: Our study shows that nonlinear metrics applied to vibrato tones provide a valuable tool for observing and quantifying regularity in vibrato tones. The results of this study highlight the potential for more detailed descriptions of vibrato characteristics that may support categorization of individuals, genres, or musical expression in the future and could be applicable to pedagogical techniques.

Keywords: Entropy; Periodicity; Recurrence; Singing; Vibrato.