A Novel Human Identification Model Based on Multi-objective Optimization of Electrocardiogram

This paper presents a novel human identification method which is based on multi-objective optimization and dynamic time wrapping of Electrocardiogram (ECG). A modified preprocessing method is proposed to suppress baseline wander and high frequency noises of real ECG. Then we establish dynamic time wrapping (DTW) between reference ECG and others. Finally, the multi-objective optimization model of human identification is developed based on the above results and Pareto entropy particle swarm optimization is used to solve the problem. Compared with RBF neural network (58.82%), and multitask learning approach (64.70%), our proposed method could achieve an accuracy of 82.35% after running for 10min in speed of 10km/h. The proposed model will make human identification based on ECG more widely applied in everyday life.