Emerging EEG and kinect face fusion for biometrie identification

Multimodal biometrie systems attract attention as security systems due to their ability to eliminate the limitations of unimodal systems, such as noisy data, intra-elass variability, inter-class similarities etc. This article presents for the first time a multimodal biometric system that combines emerging electroencephalograph (EEG) brain signal with the KINECT face modality by utilizing match score level fusion. For the EEG modality, we have considered motor imagery EEG brain signals and proposed a novel combination of time-frequency features extraction in various bands of EEG signal by utilizing the back propagation (BP) classification algorithm. For the Kinect face modality, a novel approach based on histogram oriented gradient (HOG) feature has been proposed. The major advantage of using the HOG feature is it allows to extract the prominent information from the gradient intensity of the facial image. The K-nearest neighbors (KNN) feature matching algorithm is then used for feature classification. Finally, the match score fusion technique has been applied to the outcomes of individual bands of the brain signal identification system and face recognition systems. The method achieves an accuracy of close to 100% for the fusion of EEG and KINECT EUROCOM face datasets.