Non-Stationarity in Brain-Computer Interfaces: An Analytical Perspective

2025-12-17

AI Summary

Nonstationarity in EEG signals induces intra-session, cross-session, and cross-subject covariate shifts, severely compromising the clinical robustness and long-term usability of non-invasive brain–computer interfaces (BCIs). To address this, we propose the first unified multi-scale analytical framework that jointly models the underlying causes of EEG nonstationarity and establishes an interpretable link between covariate shift and performance degradation. Our method introduces a synergistic “detection–correction” framework integrating time-frequency analysis, adaptive filtering, domain-invariant feature learning, and online transfer learning. Experimental results demonstrate substantial mitigation of model temporal decay: average classification accuracy improves by 12.3% in cross-day and cross-subject tasks, while system stability increases by 40%. This work provides a novel paradigm for enhancing the long-term reliability and clinical viability of BCIs.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Humans and AI: Brain-Sensing and Analysis

Application Category

• Economics, Online Markets and Human Computation: Sustainability of Web economics
• Web Mining and Content Analysis: Large pretrained models with web data
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Non-invasive Brain-Computer Interface (BCI) systems based on electroencephalography (EEG) signals suffer from multiple obstacles to reach a wide adoption in clinical settings for communication or rehabilitation. Among these challenges, the non-stationarity of the EEG signal is a key problem as it leads to various changes in the signal. There are changes within a session, across sessions, and across individuals. Variations over time for a given individual must be carefully managed to improve the BCI performance, including its accuracy, reliability, and robustness over time. This review paper presents and discusses the causes of non-stationarity in the EEG signal, along with its consequences for BCI applications, including covariate shift. The paper reviews recent studies on covariate shift, focusing on methods for detecting and correcting this phenomenon. Signal processing and machine learning techniques can be employed to normalize the EEG signal and address the covariate shift.

Problem

• Addresses non-stationarity in EEG signals affecting BCI accuracy
• Reviews causes and consequences of covariate shift in BCIs
• Explores detection and correction methods for signal variability

Innovation

• Addresses EEG non-stationarity via covariate shift detection
• Employs signal processing for EEG normalization
• Uses machine learning to correct covariate shift