Unsupervised EEG-based decoding of absolute auditory attention with canonical correlation analysis

2025-04-24

AI Summary

This study addresses absolute auditory attention decoding (aAAD)—the task of determining, without labeled data, whether a subject is actively attending to auditory stimuli from electroencephalographic (EEG) signals. We propose the first fully unsupervised, end-to-end framework for aAAD, integrating unsupervised discriminative canonical correlation analysis (CCA) for feature extraction with a Minimally Informed Linear Discriminant Analysis (MILDA) classifier—requiring neither attention-state labels nor subject-specific calibration. Unlike existing supervised approaches, our method achieves higher accuracy on non-stationary, cross-session EEG data while incurring negligible computational overhead, eliminating calibration, enabling plug-and-play deployment, and supporting cross-subject adaptability. Experimental results demonstrate statistically significant performance gains over state-of-the-art supervised models. The framework establishes a deployable, real-time paradigm for individualized auditory attention monitoring, advancing practical brain–computer interface applications in dynamic listening environments.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Unsupervised & Self-Supervised Learning
• Humans and AI: Brain-Sensing and Analysis

Application Category

• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Search and Retrieval-Augmented AI: Ad search and search for Web retail
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

We propose a fully unsupervised algorithm that detects from encephalography (EEG) recordings when a subject actively listens to sound, versus when the sound is ignored. This problem is known as absolute auditory attention decoding (aAAD). We propose an unsupervised discriminative CCA model for feature extraction and combine it with an unsupervised classifier called minimally informed linear discriminant analysis (MILDA) for aAAD classification. Remarkably, the proposed unsupervised algorithm performs significantly better than a state-of-the-art supervised model. A key reason is that the unsupervised algorithm can successfully adapt to the non-stationary test data at a low computational cost. This opens the door to the analysis of the auditory attention of a subject using EEG signals with a model that automatically tunes itself to the subject without requiring an arduous supervised training session beforehand.

Problem

• Detects active listening versus ignored sound from EEG
• Uses unsupervised CCA and MILDA for aAAD classification
• Adapts to non-stationary EEG data without supervised training

Innovation

• Unsupervised discriminative CCA for feature extraction
• Combines with MILDA for aAAD classification
• Adapts to non-stationary data cost-effectively