Feature level fusion in multimodal biometrie identification

In this paper, we propose the fusion of two uni-modal biométric verification systems, based on face and offline signature. The extraction of Gabor filter parameters is studied in two ways. A new paradigm is proposed in machine learning as the spiking neuron network) called Liquid State Machine, strategy at fusion feature vector is used and tested. The experiment is performed on a multimodal database consisting of 400 images of 80 subjects (i.e. five images per subject,), three images are used for training and two are used for testing. Good performance is obtained by merging: the contribution of multi-modality is confirmed. This preliminary study confirms the feasibility of a robust and reliable multimodal biométrie system.