Feature-Level Fusion of Multimodal Biometric for Individual Identification by Training a Deep Neural Network

Digital era needs obligatory requirement of multimodal biometric to spot a person to access a specific environment. To provide such a high secure platform, the proposed system used fingerprint and face modalities to check an individual’s identity. The objective of this research work is to give complete recognition accuracy without performing any pre-processing on the acquired images of face and fingerprint. To obtain this, the features are extracted using histogram of oriented gradients (HoG) and Speeded up Robust Features (SURF) algorithms. These features are fused to provide as input to train a deep network. The images are acquired from public databases, AT & T are used for face images, and FVC 2004 is used for fingerprint images. The pre-defined convolutional neural network (CNN) models, AlexNet, GoogLeNet, VGG16 and ResNet50 are also tested with the acquired images. But the proposed system well-behaved and has given highest recognition accuracy than other CNN methods.