Rethinking Self-Training Based Cross-Subject Domain Adaptation for SSVEP Classification

2026-01-29

AI Summary

Cross-subject SSVEP decoding is hindered by substantial inter-subject variability and the high cost of labeled data. To address this, this work proposes a self-training-driven domain adaptation framework that integrates filter bank Euclidean alignment (FBEA) and adversarial learning to align source and target domain distributions. A dual-ensemble self-training (DEST) mechanism is introduced to enhance pseudo-label quality, while time-frequency augmented contrastive learning (TFA-CL) is employed to improve feature discriminability. The proposed method achieves state-of-the-art cross-subject classification performance on both the Benchmark and BETA datasets and demonstrates robustness across varying signal lengths.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• User Modeling, Personalization and Recommendation: User modeling for targeted and personalized online advertising
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Steady-state visually evoked potentials (SSVEP)-based brain-computer interfaces (BCIs) are widely used due to their high signal-to-noise ratio and user-friendliness. Accurate decoding of SSVEP signals is crucial for interpreting user intentions in BCI applications. However, signal variability across subjects and the costly user-specific annotation limit recognition performance. Therefore, we propose a novel cross-subject domain adaptation method built upon the self-training paradigm. Specifically, a Filter-Bank Euclidean Alignment (FBEA) strategy is designed to exploit frequency information from SSVEP filter banks. Then, we propose a Cross-Subject Self-Training (CSST) framework consisting of two stages: Pre-Training with Adversarial Learning (PTAL), which aligns the source and target distributions, and Dual-Ensemble Self-Training (DEST), which refines pseudo-label quality. Moreover, we introduce a Time-Frequency Augmented Contrastive Learning (TFA-CL) module to enhance feature discriminability across multiple augmented views. Extensive experiments on the Benchmark and BETA datasets demonstrate that our approach achieves state-of-the-art performance across varying signal lengths, highlighting its superiority.

Problem

• SSVEP
• cross-subject
• domain adaptation
• signal variability
• annotation cost

Innovation

• Self-Training
• Domain Adaptation
• SSVEP Classification
• Contrastive Learning
• Filter-Bank Alignment