“…Since EEG signals are often statistically nonstationary, particularly where there is an abnormal event captured over the observation interval, such as an ERP, a TFR technique is the solution to the problem. In general, the most applied non-linear measures found in the literature are attractor dimension [7], correlation dimension [8], Lyapunov exponent [9], fractal dimension [10 -11] and sample entropy [11][12][13]. All these measures depend on the setting of estimation parameters, namely embedding dimension, time delay of phase space reconstruction, prediction time horizon, and partitioning.…”