Table 2 Standardized beta (β) coefficients from a regression analysis predicting model attention from key acoustic features.

From: Interpretable deep learning reveals distinct spectral and temporal drivers of perceived musical emotion

Acoustic feature

Prediction of valence (β)

Prediction of rousal (β)

Tempo (BPM)

0.09

0.55***

Spectral Centroid (Brightness)

0.31**

0.18*

Spectral Flux (Timbral Variation)

− 0.11

0.48***

Harmonic Complexity

− 0.43***

0.15

Rhythmic Regularity

0.26**

0.35***

  1. Note: β represents the standardized regression coefficient. * p <.05, *p <.01, **p <.001.
Back to article page