Enhanced Letter Recognition in Intracortical BCI via Optimized Feature Extraction and Classification

This paper investigates methods to optimize letter recognition using intracortical brain-computer interfaces (BCIs). We replicate and analyze the reference method employed by Willett et al. (2021), achieving improved results through refined evaluation methodologies. Our proposed approach utilizes a combination of Principal Component Analysis (PCA) and K-Nearest Neighbor (KNN) classification. Furthermore, we explore Convolutional Neural Networks (CNNs) for classifying letter trajectories reconstructed with different parametrization techniques. Our findings demonstrate the potential for enhanced letter recognition in intracortical BCI systems through a combination of optimized feature extraction, classification algorithms, and trajectory reconstruction methods.