Olfaction recognition by EEG analysis using wavelet transform features

Abstract

The responses of the brain into different information coming from sense organs could be analyzed by various kinds of measuring techniques. Among the existing techniques, Electroencephalography (EEG) is widely used because of its low setup costs, easy implementation and noninvasive nature. the response of the human brain to olfaction has been analyzed in recent years. Particularly, it has not been exactly proved how the human brain gives response to different odors because of the limited kind of odor usage and different kinds of proposed methods. the present study demonstrates the effect of lotus flower and cheese odors on EEG signals, which were recorded from 5 healthy subjects at the eyes open and eyes closed conditions. in order to show the effectiveness of the proposed method, we categorized the EEG trials into two classes between lotus flower and cheese odors. in order to represent the EEG trials, we extracted features by using Wavelet Transform coefficients. As wavelet function, we tested five kinds of wavelets including Morlet, Mexican, Meyer, Coiflet and Daubechies on delta, theta, alpha, beta, whole band of the EEG signal. the extracted features were classified by k-nearest neighbor algorithm. the achieved results showed that among the tested wavelet functions, Mexican wavelet has a great potential to represent the EEG signals which were recorded during smelling of lotus flower and cheese odors under the eyes open and eyes closed conditions. Moreover, we achieved with Mexican 98.29% and 94.08% average classification accuracy rates on the eyes open and closed conditions, respectively.