The Single-Channel SSVEP-based Biometric Approach Using RNN-based Deep Models: A Systematic Comparison

Abstract

Biometric technology is the branch of science dealing with identification and verification of individuals, focusing on physiological and behavioral traits. These traits can be reliable, permanent, unique, and capable of distinguishing one person from others. Fingerprints, the iris, and the face are conventionally used in biometric techniques involving ID cards and passwords. However, due to security threats and fraud, these static textures and approaches can be abused. Consequently, biometric studies based on electroencephalography (EEG) have received increasing attention because each individual has a dynamic and unique pattern. However, classic EEG-based biometrics have significant deficiencies, including noise-prone signals, gel-based electrodes, and the need for multi-training/multi-channel acquisition and high mental effort. In contrast, steady-state visually evoked potential (SSVEP)-based biometrics have the important advantages of low signal-to-noise ratio and untrained usage. The SSVEP signals also embrace a number of other advantages. First, the occipital lobe is the only suitable region for collecting SSVEP signals without obstruction. Second, because elicited dynamic brain responses are a natural subconscious activity, the omnipresent SSVEP signals are the most frequently emitted signals. Anyone can continuously look into flickering lights having distinct frequencies, such as cell phone flashes, without extra physical or mental effort. Few studies involving multi-channel/multi-trial SSVEP-based biometric research are available in the current literature. Moreover, there is a lack of research comparing them to the single-channel single-trial SSVEP-based biometric approach using dry electrode-implemented recurrent neural network (RNN)-based deep learning models. To the best of our knowledge, no prior work has proposed such a biometric comparison of the RNN-based deep learning models and different sized time-series data compiled over raw and discrete wavelet transform (DWT)-based SSVEP signals. By achieving up to 100% accuracy using 11 individuals, the biometric recognition results are promising. This single-channel SSVEP-based biometric approach using RNN-based deep learning models may offer low-cost, user-friendly, and reliable individual identification authentication, leading to significant application domains.