Pattern Recognition System of Optical Fiber Fusion Defect Based on Fuzzy Neural Network in EPON

Because of having many advantages, optical fiber network is applied widely in high-tech fields. But the existence of optical fiber fusion defects will debase the quality of message transmission. A set of defect recognized system is established based on the compensatory fuzzy neural network of using wavelet and with fast algorithm in this paper. The 'energy-defect' method to extract eigenvalue is used firstly, then defect classification is recognized by fuzzy neural network. The results of simulation show that the model established by making use of this algorithm has higher efficiency, and the possibility of wrap in local minimum value of the network during the training process is smaller, which can compare to approach the precision utmost steadily and classification recognize the defect precision.