Biometric Authentication of an Individual Using Multilayer Perceptron and Support Vector Machine

The proposed method uses multi layered perceptron neural network and support vector machine to classify the normal subjects. Data used for training, testing and cross validation was recorded from normal persons (without any heart disease) within thirty six months, in the interval of 10/15 days. Ten hybrid features were extracted from the recorded signals by using wavelet transform. The classification performance is evaluated based on percent average classification accuracy and mean squared error. During analysis the optimum results were observed using SVM classifier. An accuracy of 98.86 percent is achieved and mean squared error is found to be 0.0080. Keywords— Classification, Discrete wavelet transform, ECG, Multilayer Perceptron, Support Vector Machines.