The human brain is probably the most amazing and complex system in universe. With approximately hundred billion neurons, brain transforms a variety of inputs such as thought, feeling, sound and odor into a physical or emotional response. A way to measure the response of the brain into such inputs is analyzing the electroencephalogram (EEG) signals. In this work, the EEG signals, which were recorded during smelling of cheese and rosewater odors, are analyzed and classified based on features which were extracted using autoregressive model. EEG signals were taken from 5 healthy subjects in the cases of eyes open and eyes closed at Swiss Federal Institute of Technology. By using k-nearest neighbor classifier, we obtained a mean classification accuracy rate of 96.21 % for the subjects at eyes closed case and a mean classification accuracy rate of 72.24 % for the subjects at eyes open case. The obtained results prove that the proposed method have great potential for classifying the EEG signals recorded during smelling of cheese and rosewater odors and instead of using subject-specific model it can be applied to all subjects.