In this paper, classification of electroencephalography (EEG) signals of motor imaginary tasks is studied using cepstrum analysis and linear predictive coding (LPC). The Brain-Computer Interface (BCI) competition III dataset IVa containing motor imaginary tasks for right hand and foot of five subjects are used. The data was preprocessed by applying whitening and then filtering the signal followed by feature extraction. A random forest classifier is then trained using the cepstrum and LPC features to classify the motor imaginary tasks. The resulting classification accuracy is found to be over 90%. This research shows that concatenating appropriate different types of features such as cepstrum and LPC features hold some promise for the classification of motor imaginary tasks, which can be helpful in the BCI context.