@article{mckinney2022,
Author = {A. F. McKinney and B. Cauchi},
Title = {Non-Intrusive Binaural Speech Intelligibility Prediction From Discrete Latent Representations},

Journal = {IEEE Signal Processing Letters},
Year = {2022},
Pages = {987-991},
Doi = {10.1109/LSP.2022.3161115},
type = {article},
Abstract = {Non-intrusive speech intelligibility (SI) prediction
from binaural signals is useful in many applications. However,
most existing signal-based measures are designed to be applied to
single-channel signals. Measures specifically designed to take into
account the binaural properties of the signal are often intrusive
– characterised by requiring access to a clean speech signal – and
typically rely on combining both channels into a single-channel signal before making predictions. This letter proposes a non-intrusive
SI measure that computes features from a binaural input signal
using a combination of vector quantization (VQ) and contrastive
predictive coding (CPC) methods. vector quantized contrastive
predictive coding (VQ-CPC) feature extraction does not rely on any
model of the auditory system and is instead trained to maximise the
mutual information between the input signal and output features.
The computed VQ-CPC features are input to a predicting function
parameterized by a neural network. Two predicting functions are
considered in this letter. Both feature extractor and predicting
functions are trained on simulated binaural signals with isotropic
noise. They are tested on simulated signals with isotropic and real
noise. For all signals, the ground truth scores are the (intrusive)
deterministic binaural short-time objective intelligibility (STOI).
Results are presented in terms of correlations and mean-squared
error (MSE) and demonstrate that VQ-CPC features are able to
capture information relevant to modelling SI and outperform all the
considered benchmarks – even when evaluating on data comprising
of different noise field types.}
}
@COMMENT{Bibtex file generated on }