Gender clasification based on single channel EEG signal

This paper presents an approach for gender recognition from single channel EEG signal. For this purpose, approximately 24 hour-long EEG data, obtained during daily routine activities including sleep, was used. First, cepstrum coefficients of EEG signals were obtained in the frequency domain to construct the features SET. Second, a machine learning step was performed using these features with Support Vector Machines (SVM). Finally, gender identification was performed on the test data for which features were obtained in the same manner. Based on the initially obtained experimental results, epoc based gender classification success rate of the proposed method is 77.84% for the awake phase of the day while success rate is 89.66% for the sleep phase. Based on these results, it was determined that the biometric discriminative capability of the EEG signal varies at different times of the day.