The paper presents a method for reconstructing the impulse hemodynamic response (HR) of functional near-infrared pectroscopy (fNIRS) data measured in the motor cortex of a human brain. The empirical impulse HRs are averaged across the most active channels of nineteen healthy volunteers. The impulse HR is modeled by a state space equation in which the optimal order of this model is computed by using the Akaike information criterion. The subspace method is utilized to estimate the parameter matrices in the state space model based on the obtained fNIRS-HR to an impulse stimulus. The stability of the reconstructed HR to an impulse stimulus will be investigated. It is worthy to note that the third-order state-space equation can be utilized to describe the impulse HR
of the motor cortex in generating the predicted HR to an arbitrary stimulus sequence in an online imaging.