Hybridization of Feature Level Fusion with Ant Colony Optimization in Multimodal Biometrics

In biometric system, multimodal biometrics provides stronger security as compared to unimodal biometrics. Even though multimodal biometric improves the accuracy and reliability of the system, but requires large memory storage and consumes numerous execution time due to use of high dimensionality datasets. Search is being an NP-hard problem in biometrics, which garnish an attention for research in biometric system. Due to NP-hard nature of searching in biometric, accurate solutions could not be discovered in limited time. Therefore, researchers use heuristic or random search methods such as PSO, GA, ACO and Cuckoo search etc. to obtain optimal or approximate optimal solutions for such problems. This paper proposes a hybrid approach of feature level fusion in biometric system with Ant Colony Optimization based feature sub selection method to aiming to improve performance. The median filter and morphological operations are used for pre-processing of finger vein and fingerprint images respectively. Confusion matrix plot with equal error rate and accuracy are the evaluation parameters.