Processing Near Infrared Spectroscopy Signals Using Canonical Correlation Analysis

Functional near infrared spectroscopy signals (NIRS) are measured by a number of LED-detector pairs. Estimation of event related responses from these signals needs multivariate analysis methods. This study concentrates on a particular multivariate method, namely the canonical correlation analysis and investigates its use for eliminating trend and noise effects from fNIRS signals and estimating event related haemodynamic response. Proposed methods are applied to real cognitive data for validation.