Epilepsy is a very common neurological disorder. Electroencephalogram (EEG) is the major diagnostic tool used for analyzing the human epileptic seizure activity and there is a strong need of an efficient automatic seizure detection using it to ease the diagnosis. This work aims at an automatic system for diagnosis of epilepsy. Here we extract some features like fractal dimensions, sample entropy, Lyapunov exponent, etc of both normal and epileptic EEG signals. These feature values are used as inputs to train classifiers like Artificial neural networks, support vector machines, probabilistic neural networks etc., After the training, we test the classifier with test EEG data.