This thesis outlines approaches to improve the signal processing and anal- ysis of Near-infrared spectroscopy (NIRS) based brain-computer interfaces (BCI). These approaches were developed in conjunction with the implemen- tation of a new customized exible multi-channel NIRS based BCI hardware system (Soraghan, 2010). Using a comparable functional imaging modality the assumptions on which NIRS-BCI have been reassessed, with regard to cognitive task selection, active area locations and lateralized motor cortex activation separability. This dissertation will also present methods that have been implemented to allow reduced hardware requirements in future NIRS-BCI development. We will also examine the sources of homeostatic physiological interference and present new approaches for analysis and at- tenuation within a real-time NIRS-BCI paradigm.