Timbre perception in violins: spectral and formant influences of string materials in a controlled study

TL;DR

This study explores how different violin string materials affect the sound's timbre, linking subjective perceptions to acoustic features like spectral energy and formant frequencies.

Contribution

The study introduces a controlled experimental setup and a Gradient Boosting Tree model to predict timbre perception from acoustic parameters.

Findings

Gut and nylon strings are perceived as darker and rougher, while steel strings sound brighter and more defined.

Timbre perception is negatively correlated with formant frequencies, and Sharpness is influenced by high-frequency energy.

A Gradient Boosting Tree model outperformed linear regression in predicting subjective timbre impressions (R² ≈ 0.47).

Abstract

String instrument timbre is influenced by a complex interplay of string material, instrument body characteristics, and playing technique. However, the perceptual effects of different string materials and their relationship with acoustic parameters remain incompletely understood. This study investigates how violin string materials (gut, nylon, and steel) affect subjective timbre perception and their correlation with acoustic features. Violin sounds were recorded using an automated bowing machine, ensuring consistency, and analyzed for spectral energy distribution and formant frequencies. Subjective impressions were collected using a semantic differential (SD) method, and factor analysis identified three key perceptual dimensions: Texture, Sharpness, and Power. Statistical analysis revealed that string material significantly influenced perceived timbre, with gut and nylon strings associated with darker and rougher tones and steel strings producing brighter and more defined sounds. Correlation analysis showed that Timbre perception was negatively associated with formant frequencies, while Sharpness was influenced by high-frequency energy levels, particularly in the mid-to-high range. Power exhibited weaker correlations with acoustic parameters, suggesting additional influences from temporal dynamics and instrument-specific characteristics. A Gradient Boosting Tree (GBT) regression model was employed to predict subjective impressions from acoustic parameters, demonstrating superior performance over linear regression methods (R² ≈ 0.47 for Timbre and Sharpness). Sensitivity analysis further confirmed that formant frequencies and spectral balance played a critical role in shaping timbre perception. These findings provide valuable insights into the perceptual mechanisms of violin timbre, benefiting musicians, instrument makers, and computational timbre modeling research.

Figures