Verification by using multi-biometric methods

Systems using biometric features are more secure than encryption systems using numbers. The passwords used by people can be stolen, but there is no such possibility as they are part of the person using the biometric features. However, biometric features are not 100% safe. Malicious people can use these biometrics by forcing people with biometric characteristics. In this study, we tried to find solutions to these problems. A safer biometric system has been proposed. Today's results in singular identification studies have not achieved enough performance in terms of success rate, error and time. There are difficulties in getting biometric parameters such as iris (which has high performance) because people do not release these biometric characteristics. Besides, problems arise about safe recognition. Another disadvantage is that the in singular-biometri, information can not be read due to the conditions and results in failure. With this study, even if the problem is encountered when any one of the biometrics of the person is being obtained by using multiple biometrics, the system works successfully thanks to other biometrics. Face and thermal facial images were taken from the "USTC NVIE Spontaneous Database" database and ear images were taken from the "IIT Delhi Ear Image Database" database. For each biometry, three different feature extracting methods and four different classification algorithms (multi-layered perceptron, decision tree, support vector xiv machines and statistical neural networks) were used for identification. Two different fusion methods (matching score level fusion and feature level fusion) are used with these methods. The results (FAR, FRR, and accuracy) resulting from multiple biometry and fusion have been more successful than results from individual biometrics of individuals.