Real-Time Decoding of Motor Gestures from Hemodynamic Signals Using Single-Channel fNIRS

Functional near-infrared spectroscopy (fNIRS) is an emerging neuroimaging technique for real-time brain activity monitoring. This study investigates the classification of motor gestures using a single-wavelength fNIRS system, which captures total hemoglobin (HbT) variations in the motor cortex (MC). We compare a Long Short-Term Memory (LSTM) neural network with and without Independent Component Analysis (ICA) and Principal Component Analysis (PCA) for feature extraction. Experimental results demonstrate that the standard LSTM model achieves a higher Area Under the Curve (AUC) of 0.9172 compared to 0.8831 for the LSTM+ICA&PCA model, suggesting that while ICA and PCA improve noise reduction, they may also filter out critical signal components. Additionally, real-time performance evaluation confirms the feasibility of fNIRS for rapid and accurate motor gesture classification.