Regresar

A method for studying how much time of EEG recording is needed to have a good user identification

Abstract:

Recently, several researchers have tried to develop reliable biometric systems based on biological signals. Brainwave signals, like Electroencephalograms (EEG), are unique to each person. Also, EEGs are harder to steal and replicate than traditional biometrics like fingerprint or face recognition. Even though, there are many related works, to our knowledge none of them has studied what is the impact of the duration of the recorded signals in user identification accuracy. In order to answer this question, this paper presents a method for the development of biometric systems based on EEG signals. The proposed method uses a Discrete Wavelet Transform (DWT) to extract relevant features, and a hyperparameter selection for adjusting the base models following a greedy strategy. In the task of user identification, using five classifiers as base models, the experiments showed that just 2 seconds of recording reach an accuracy of approximately 90% and with 20 seconds the accuracy increases to 99%.