Decoding Stimulus Reconstruction-Based Auditory Attention Robustly in Unbalanced EEG Datasets

2026-05-25

AI Summary

This work addresses the issue of inflated performance estimates in auditory attention decoding (AAD) when existing stimulus reconstruction methods are evaluated on imbalanced EEG datasets. To mitigate this bias, we propose the Leave-One-Pair Envelope-Out (LOPEO) cross-validation protocol, which enforces strict consistency between the envelope distributions of training and test samples, thereby effectively reducing evaluation artifacts caused by data imbalance. Using deep neural network–based reconstruction models, we validate the efficacy of LOPEO on three public EEG-AAD datasets: KUL, DTU, and NJU cEEGrid. Experimental results demonstrate that LOPEO substantially curbs assessment bias and provides a robust, reliable framework for evaluating decoding performance on imbalanced neurophysiological data.

Abstract

In the past decade, numerous studies have applied deep neural networks (DNNs) to decode auditory attention (AAD) from Electroencephalogram (EEG) signals via stimulus reconstruction. However, the influence of dataset balance on the decoding performance of stimulus reconstruction-based AAD remains unexplored. In this study, three publicly available EEG-AAD datasets - KUL, DTU, and NJU cEEGrid - are used to construct both balanced and unbalanced experimental conditions. We hypothesize and demonstrate that stimulus reconstruction-based DNN decoders tend to produce overestimated decoding performance on unbalanced datasets. To address this issue, we propose a leave-one-paired-envelope-out (LOPEO) cross-validation protocol. Experimental results confirm that LOPEO effectively prevents inflated decoding accuracy on unbalanced datasets. While balanced datasets are generally preferred in experimental design, LOPEO provides a principled evaluation framework for unbalanced datasets that have already been published, filling an important gap in the field.

Problem

• auditory attention decoding
• stimulus reconstruction
• unbalanced EEG datasets
• decoding performance
• dataset bias

Innovation

• stimulus reconstruction
• auditory attention decoding
• unbalanced EEG datasets
• LOPEO cross-validation
• deep neural networks