MULTIMODAL BIOMETRIC IDENTIFICATION WITH THE AID OF ADVANCED TRANSFORMS AND RANDOM FOREST CLASSIFIER

Many of the existing biometric systems are unimodal, i.e. they identify using only any one of the characteristic features like a fingerprint, palmprint, voice, face etc. Hence this paper aims to develop a biometric identification system which is multimodal with the help of advanced transforms and random forest classifier. The use of multiple biometrics reduces the system error rate. The developed biometric system will use the iris, fingerprint and palmprint data to identify the individual. Noise reduction in the data is achieved using bilateral filtering for all the three, i.e. iris, fingerprint and palmprint database whereas three different segmentation techniques of cellular automata, threshold-based segmentation and canny edge detection are used for iris, fingerprint and palmprint respectively. The features are extracted using wavelet and geometric-based method for iris and GLCM method for both finger and palm prints and a random forest classifier is used in the data classification process. After the evaluation of the performance metrics in the developed system, 90% accuracy, 100% sensitivity and nil false rejection rate are obtained.