iFuzzyTL: Interpretable Fuzzy Transfer Learning for SSVEP BCI System

TL;DR

iFuzzyTL introduces an adaptive framework that combines fuzzy logic principles with neural network architectures, focusing on efficient knowledge transfer and domain adaptation, and bolsters the model's precision and aligns with real-world operational demands by effectively managing the inherent variability and uncertainty of EEG data.

Abstract

The rapid evolution of Brain-Computer Interfaces (BCIs) has significantly influenced the domain of human-computer interaction, with Steady-State Visual Evoked Potentials (SSVEP) emerging as a notably robust paradigm. This study explores advanced classification techniques leveraging interpretable fuzzy transfer learning (iFuzzyTL) to enhance the adaptability and performance of SSVEP-based systems. Recent efforts have strengthened to reduce calibration requirements through innovative transfer learning approaches, which refine cross-subject generalizability and minimize calibration through strategic application of domain adaptation and few-shot learning strategies. Pioneering developments in deep learning also offer promising enhancements, facilitating robust domain adaptation and significantly improving system responsiveness and accuracy in SSVEP classification. However, these methods often require complex tuning and extensive data, limiting immediate applicability. iFuzzyTL introduces an adaptive framework that combines fuzzy logic principles with neural network architectures, focusing on efficient knowledge transfer and domain adaptation. iFuzzyTL refines input signal processing and classification in a human-interpretable format by integrating fuzzy inference systems and attention mechanisms. This approach bolsters the model's precision and aligns with real-world operational demands by effectively managing the inherent variability and uncertainty of EEG data. The model's efficacy is demonstrated across three datasets: 12JFPM (89.70% accuracy for 1s with an information transfer rate (ITR) of 149.58), Benchmark (85.81% accuracy for 1s with an ITR of 213.99), and eldBETA (76.50% accuracy for 1s with an ITR of 94.63), achieving state-of-the-art results and setting new benchmarks for SSVEP BCI performance.

Figures (9)

• Figure 1: The diagram of three classification scenarios. (A) intra-subject classification; (B) inter-subject few-shot classification; (C) inter-subject zero-shot classification.
• Figure 2: Illustration of the architecture for predicting target frequencies in an SSVEP task using the proposed iFuzzyTL model. (A) Main structure of the iFuzzyTL model. (B) Design of the spatial and temporal fuzzy filters. (C) Detection of the center using the spatial fuzzy filter. (D) Firing strength of a demonstration sample to show the learned neural pattern as identified by the spatial fuzzy filter. (E) Detection of the center using the temporal fuzzy filter. (F) Firing strength of a demonstration sample to show the learned neural pattern as identified by the temporal fuzzy filter.
• Figure 3: Detailed illustration of a real-time SSVEP experiment setup. (A) Demonstration setup for the experiment. The EEG channels are located in the occipital lobe. (B) Example of a filtered EEG signal used in the demo. (C) Min-max normalized firing strength across the rules. (D) Averaged weight distribution among the rules (without normalization). (E) Data distribution following the application of the spatial filter. (F) Fourier Transform results of the demo EEG signal.
• Figure 4: FFT features that triggered each fuzzy rule across different target frequencies, illustrating the identification of harmonic peaks.
• Figure 5: Visualization of demographic subjects from the 12JFPM dataset (2s) illustrating what iFuzzyTL learned. (A) Data distribution post-application of the spatial filter, highlighting the position (Red Star) of a sample needing explanation at 11.75Hz. (B) Filtered EEG signals and their Fourier Transform to display the data characteristics. (C) Representation of firing strength and the center, using min-max normalization across the channel dimension to accentuate differences within one rule. The center is reconstructed from the query space to the raw EEG signal space as described in proposition \ref{['proposition:2']}.