Near-infrared Spectroscopy Signal Classification: Towards a Brain-computer Interface

• Main Author: Tai, Kelly
• Other Authors: Chau, Tom
• Language: en_ca
• Published: 2008
• Subjects: Near-infrared spectroscopy; Brain-computer interface; Locked-in syndrome; Emotional induction; 0541
• Online Access: http://hdl.handle.net/1807/19309

A brain-computer interface (BCI) allows individuals to communicate through the modulation of regional brain activity. Clinical near-infrared spectroscopy (NIRS) is used to monitor changes in cerebral blood oxygenation due to functional activation. It was hypothesized that visually-cued emotional induction tasks can elicit detectable activity in the prefrontal cortex. Data were collected from eleven participants as they performed positively and negatively-valenced emotional induction tasks. Baseline and activation trials were classified offline with accuracies from 75.0-96.7% after applying a feature selection algorithm to determine optimal performance parameters for each participant. Feature selection identified common discriminatory features across participants and relationships between performance parameters. Additionally, classification accuracy was used to quantify NIRS hemodynamic response latency. Significant increases in classification rates were found as early as 2.5 s after initial stimulus presentation. These results suggest the potential application of emotional induction as a NIRS-BCI control paradigm.