An instrumental intelligibility metric based on information theory

TL;DR

This paper introduces SIIB, a new information-theoretic metric for speech intelligibility that accounts for talker variability and dependencies, showing high correlation with human judgments on degraded speech.

Contribution

The paper presents SIIB, a novel monaural intrusive intelligibility metric based on information theory that improves upon existing metrics by considering variability and dependencies.

Findings

SIIB correlates highly with human speech intelligibility scores.

SIIB outperforms existing metrics on noisy and processed speech.

The metric effectively captures information transfer between talker and listener.

Abstract

We propose a monaural intrusive instrumental intelligibility metric called speech intelligibility in bits (SIIB). SIIB is an estimate of the amount of information shared between a talker and a listener in bits per second. Unlike existing information theoretic intelligibility metrics, SIIB accounts for talker variability and statistical dependencies between time-frequency units. Our evaluation shows that relative to state-of-the-art intelligibility metrics, SIIB is highly correlated with the intelligibility of speech that has been degraded by noise and processed by speech enhancement algorithms.