| Biometric identification is a technology using the biological characteristics as humanidentification. The biological characteristic can be divided into physical characteristic (such asfingerprints,factial,image,iris,palm,etc.)and behavioral characteristics (such asgait ,voice ,handwriting, etc). With the development of information technology, accurate and easyidentificationofindividualshavebecomeahottopicofpeople'sattention.Based on the principle of pattern recognition, firstly, the deficiency of neural network andgenetic algorithm are analyzed. The necessity and possibility of combining neural network aredemonstrated,. The method of heart sound identity verification based on. optimizing neuralnetworks topology structure and weights with genetic algorithm is presented , Back propagationANN model trained with genetic algorithm is presented. The related algorithms and programsmainly based on Matlab7.0 experiment platform are developed. The characteristics of numericaland application for the methods are illustrated byusingsimulation testing. Especially, the influenceof some factors about system performance, such as feature parameter, number of Hidden-layernerve cell, background noise ,is discussed. Then, considering the deficiency about FSR and MFCCparameterappliedinneuralnetwork.,adistanceidentificationmethodispresented.The research results show that The method of heart sound identity verification based onoptimizing neural networks topology structure and weights with genetic algorithm not only has thecapability in fast learning of neural network's connection weights and the capability in escapinglocal optima and initial weights. but also has a higher recognition rate. In addition, a distanceidentification method has also a higher recognition rate. Obviously the above two methods isappropriate .The results further show that heart sound signals as a biological recognition proved tobefeasible,hasahugeapplicationprospect. |
PDF Full Text Request