The ability to effectively classify electroencephalograms (EEG) is the foundation for building usable Brain-Computer Interfaces. In this presented work EEG signals were used to extract the information and classify with different mental task. EEG data was collected from a source. This data contains recording of 5 subjects in different mental task conditions (Resting, math, letter composition, geometric figure rotation task). EEG Signals were pre-processed and filtered. EOG artifacts were removed by visual inspection. For classification of these mental tasks wavelet was used to extract the features. Second order Daubechies mother wavelet has been used to get the wavelet coefficients for the selected EEG epochs. Mean, maximum, minimum and standard deviations values of wavelet coefficients for the EEG epochs were selected as inputs for the training the network and to classify mental tasks. The ANN architecture trained with one step secant method used in present study shows overall very good results of classification. This architecture of ANN was also found effectively differentiating the EEG from different mental tasks conditions Resting (98%) multiplication (92%), Letter composition (92%) and rotation (96%).