An Enhanced Brain-Computer Interface for Assisted Typing Using Random Forest and LSTM Models

Abstract

Over 75 million people globally suffer from severe motor impairments, hindering their ability to communicate independently. Brain–Computer Interfaces (BCIs) offer a promising solution, yet many existing systems are limited by low typing speeds and accuracy. This paper proposes a hybrid EEG-based BCI system that combines Random Forest classifiers and Long Short-Term Memory (LSTM) networks to interpret imagined motor actions and eye blinks. EEG signals were recorded using the low-cost NeuroSky MindWave headset, and the system was tested on 100 participants. The Random Forest model achieved an accuracy of 88.5%, while the LSTM achieved 87.9%, both enabling typing speeds of up to 12 words per minute. Beyond technical metrics, the system also demonstrated high user satisfaction and reduced cognitive load. Our main contribution lies in designing a scalable, cost-effective, and real-time BCI framework that significantly improves communication accessibility for individuals with motor disabilities, outperforming previous blink-dependent and single-model approaches.