Enhanced BCI Classification Through Dual-Paradigm Integration: A 240-Element Individually Addressable LED Array for SSVEP-P300 Fusion

Brain–computer interfaces (BCIs) offer transformative potential for individuals with severe motor impairments, yet single-paradigm systems remain constrained by classification inaccuracies and susceptibility to signal degradation. This pilot study presents a four-class hybrid SSVEP + P300 BCI system for discrete smart-environment device control, employing a custom-designed 240-element individually addressable NeoPixel LED matrix platform as a novel stimulation architecture. The platform supports simultaneous multi-frequency SSVEP stimulation and event-related P300 cueing within a unified hardware array, whilst its 24-bit colour rendering capability, supporting up to 16 million distinct colours and synchronised multi-unit animations, affords substantial extensibility beyond conventional single-paradigm stimulus systems. Hardware-level frequency verification confirmed accurate stimulus generation across all target frequencies, with a maximum deviation of 0.044% from nominal programmed values. Five healthy participants performed ON/OFF control tasks directed towards a fan and lamp across 3 s and 5 s EEG analysis windows, yielding 400 trials in total. Preliminary descriptive analysis suggests that the hybrid SSVEP + P300 paradigm maintained consistent group mean classification accuracy of 79.0% across both window durations, whilst the SSVEP-only condition exhibited a 7.0 percentage point decline from 79.0% to 72.0% at the extended window. As the two conditions were recorded in separate sessions and the P300 stage functions as a veto gate that cannot increase accuracy, this difference is interpreted cautiously. At the 5 s analysis window, the hybrid paradigm additionally demonstrated a conservative error profile characterised by predominant false negative over false positive activations, with a lower false positive rate than the SSVEP-only condition (8.0% vs. 12.0%). These exploratory findings from a proof-of-concept pilot investigation suggest potential suitability of the proposed approach for reliable and safety-conscious assistive device control, pending validation through larger-scale confirmatory studies with formal statistical analysis.