Explainability guided feature selection for speech emotion recognition using SHAP and permutation importance

S I M M Raton Mondol, Sangmin Lee · 2026

Read Explainability guided feature selection for speech emotion recognition using SHAP and permutation importance by S I M M Raton Mondol, Sangmin Lee (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13615201.txt.

Description

S. I. M. M. Raton Mondol and Sangmin Lee present a 2026 method for teaching computers to read emotion from speech while keeping the process open to inspection. Emotion systems often swallow hundreds of sound features at once, which makes them slow, prone to overfitting, and hard to explain. Their approach ranks features two ways, by a game-theory style credit score called SHAP and by how much performance drops when a feature is scrambled, then merges the ranks.

A penalty for features that echo each other keeps the final set lean without throwing away sounds that carry real feeling. Tests ran on four well-known emotional speech collections and three kinds of classifier. Trimming to features that truly matter could make emotion software lighter and more trustworthy in settings like call centres and mental health tools.

Catalog record

CREATORS I M M Raton Mondol, Sangmin Lee
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE36973
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13615201

Browse the full library · About the Tower of Babel