Letter Composition Task Classification Using NIRS and Neural Network

We describe letter composition task classification using near-infrared spectroscopy (NIRS) and our proposed neural network model called Neo-ISRM. Brain-Computer Interfaces (BCIs) are a new paradigm for communication between humans and machines and can provide a communication means for people with severe physical disabilities. Our objective is to develop a BCI system that can handle various intentions of users. A BCI should recognize various kinds of commands for controlling a machine. The number of commands corresponds to the number of categories into which a classifier can classify. In this paper, we describe the classification of NIRS data acquired while a subject performed letter composition tasks involving both category fluency and letter fluency using Neo-ISRM. For classifying tasks, two different areas were selected: the whole area (frontal and temporal lobes of the left hemisphere) having 22 data channels, and a local area having 9 data channels. The results for the local area showed that category fluency and letter fluency tasks can be classified from the NIRS signals by using our neural classifier Neo-ISRM. The results show that it is possible to increase the number of categories into which a BCI system can classify.