Hurst exponent is used to evaluate the presence or absence of long-range dependence and its degree in a time series, and hence is known as the long term memory of the time series. Fractal Dimension on the other hand is a measure of data complexity. Hurst Exponent and Fractal Dimension were used as features for nonlinear classification by QDA and SVM with a polynomial kernel of order 3. Since both Hurst Exponent and Fractal Dimension has a large inter individual variability, we used these features of consecutive sessions to study the intersession variability of classification accuracy of the proposed classifiers. QDA provided better classification for the trials trained by motor execution, while SVM with the polynomial kernel differentiated better when the training was done by motor imagery data.