Novel Encoding and Decoding Methods for Enhancing Short-Time High-Frequency VR-SSVEP Performance.

Abstract

Virtual reality (VR) brain-computer interfaces (BCIs) based on steady-state visually evoked potentials (SSVEPs) face a critical challenge: achieving high information transfer rates (ITR) requires short time windows, while user comfort necessitates high-frequency stimulation. Decoding this high-frequency and short-time combination is highly challenging due to low signal-to-noise ratios. To address this, we applied the independent visual field stimulation (IVFS) paradigm to encode six high frequency targets (31-36 Hz) using three common encoding designs. Furthermore, we propose a novel Inter-class and Spatial Domain Collaborative Decoding (ICSD-CD) framework. By incorporating the Target-Band enhanced Spatial Decomposition Pattern (TB-eDSP) for lateralized feature extraction, ICSD CD synergistically integrates conventional inter-class difference (ICD) features with IVFS-specific spatial domain difference (SDD) features via a dynamic, confidence-based fusion strategy. Experimental results demonstrate that this framework effectively resolves the decoding bottleneck, achieving average accuracies of 89.11% (119.6 bits/min) with a 0.3 s data length and 93.93% (114 bits/min) with 0.5 s. This significantly outperforms single-domain methods, providing a viable pathway for high-performance VR BCIs.