Ear Recognition Based on Residual Network

Abstract

Abstract To solve the problem of ear image recognition under complex conditions, a human ear recognition method based on depth residual neural network (ResNet) is proposed. In the proposed method, the residual module with jump connection is added to the neural network. The module alleviates the problems of gradient disappearance and model degradation caused by adding deep numbers in the deep neural network, and improves the performance of the neural network in all aspects. In this experiment, the USTB3 ear image data set is used as the experimental data, and the experimental test is carried out on the depth neural network model ResNet18. Through the experimental comparison and adjustment of parameters, the final recognition accuracy rate of 98.27% is obtained. A large number of experimental results show that the ear recognition method based on ResNet18 has higher recognition degree for ear image, and the recognition effect is better than the existing common ear recognition methods and other deep learning methods.KeywordsEar recognitionMachine learningDeep learningDeep residual