An Information Theoretic Approach Towards Assessing Perceptual Audio Quality using EEG

TL;DR

This paper introduces an information theoretic model using EEG to objectively assess perceptual audio quality, especially under time-varying distortions, by measuring brain responses as a communication channel.

Contribution

It presents a novel approach modeling the brain-audio interaction as a nonlinear, time-varying channel and introduces a low-complexity method for mutual information computation.

Findings

The approach successfully quantifies subjective audio quality perception.

Results are consistent across various music sequences and distortion types.

EEG-based mutual information correlates with perceived audio quality.

Abstract

In this paper, we propose a novel information theoretic model to interpret the entire "transmission chain" comprising stimulus generation, brain processing by the human subject, and the electroencephalograph (EEG) response measurements as a nonlinear, time-varying communication channel with memory. We use mutual information (MI) as a measure to assess audio quality perception by directly measuring the brainwave responses of the human subjects using a high resolution EEG. Our focus here is on audio where the quality is impaired by time varying distortions. In particular, we conduct experiments where subjects are presented with audio whose quality varies with time between different possible quality levels. The recorded EEG measurements can be modeled as a multidimensional Gaussian mixture model (GMM). In order to make the computation of the MI feasible, we present a novel low-complexity approximation technique for the differential entropy of the multidimensional GMM. We find the proposed information theoretic approach to be successful in quantifying subjective audio quality perception, with the results being consistent across different music sequences and distortion types.