A short-time objective intelligibility measure for time-frequency weighted noisy speech

TLDR

Existing objective speech‑intelligibility measures work for many degradations but are less suitable for time‑frequency weighted noisy speech, such as after noise reduction or speech separation. The paper introduces an objective intelligibility measure that correlates highly (ρ = 0.95) with intelligibility of both noisy and TF‑weighted noisy speech. The method computes an intermediate intelligibility score over ~400 ms TF regions using a simple DFT‑based TF decomposition. The measure achieves a ρ = 0.95 correlation, outperforms three advanced objective measures, and is available as free Matlab code.

Abstract

Existing objective speech-intelligibility measures are suitable for several types of degradation, however, it turns out that they are less appropriate for methods where noisy speech is processed by a time-frequency (TF) weighting, e.g., noise reduction and speech separation. In this paper, we present an objective intelligibility measure, which shows high correlation (rho=0.95) with the intelligibility of both noisy, and TF-weighted noisy speech. The proposed method shows significantly better performance than three other, more sophisticated, objective measures. Furthermore, it is based on an intermediate intelligibility measure for short-time (approximately 400 ms) TF-regions, and uses a simple DFT-based TF-decomposition. In addition, a free Matlab implementation is provided.