Brain computer interface based on nonlinear characteristics identification of neuronal activities evoked optical properties

Brain computer interface (BCI) technology has been developed for decades as an alternate mode of communication for disabled, such as patients suffering from amyotrophic lateral sclerosis (ALS), brain stem stroke and spinal cord injury. Near-infrared spectroscopy has recently been investigated as a non-invasive brain imaging method for developing BCI. Previous research has shown that task related hemodynamic signal recorded by NIRS from the cortex can be distinguished. However, the hemodynamic signal is a slow response for building a BCI. Near-infrared spectroscopy (NIRS) can detect two different kinds of signals from the human brain: the hemodynamic (slow) optical response, and the neuronal (fast) optical response. In this paper, we conducted a pilot study on investigating the feasibility of using fast optical response for building a NIRS-BCI. We explore the nonlinear aspects of the tactile-stimulus-evoked neuronal optical response (fast optical response) over a NIRS time series (light intensity variation). The fast optical responses (FORs) over time series recorded in stimulus sessions are confirmed by event-related averaging. The chaos levels of NIRS time series recorded both in stimulus and in rest sessions are then identified according to the estimated largest Lyapunov exponent. The obtained results strongly suggest that the chaos level can be used to recognize the FORs in NIRS time series and, thereby, the state of the pertinent brain activity.