In this study we applied phase-space analysis on the hemodynamic signals recorded from the motor cortex of human subjects using functional near infrared spectroscopy (fNIRS) to decode the direction of intentional hand movements. Our goal is to develop a brain-computer-interface (BCI) based on optical imaging that can control a wheelchair. To establish the relationship between the hemodynamic response and movement direction, participants were asked to perform repetitive arm movements in two orthogonal directions (right-left and front-back) on a horizontal plane, while the time course of the oxy-hemoglobin (oxy-Hb) and deoxy-hemoglobin (deoxy-Hb) responses were recorded. We applied phase-space analysis on oxy-Hb and deoxy-Hb signals to characterize movement direction. Our results show that movement directions taken pairwise (left vs. right, and forward vs. backward) are mapped onto different quadrants in the oxy-Hb vs. deoxy-Hb phase plane. These findings demonstrate that phase-space analysis can be used to decode the movement direction in a BCI controlling a wheelchair. In conclusion, phase-space analysis can be used to differentiate intentional movement direction without correlating the temporal movement kinematics with the hemodynamic response.