ECG identification fusion model based on siameseneural networks

Since the birth of human civilization, identity recognition has been a very important work. For a long time, human beings often encounter the need to verify their personal identity in social activities, from ancient Dudie and signature Monogram to modern fingerprint, face and pupil recognition. Also driven by this demand, biometric technology has been greatly developed. Because the common physiological features such as fingerprint and palmprint may be forged, there is an urgent need for a new biometric feature with convenience, reliability and high anti-counterfeiting. With the deepening of people's understanding of physiological signals, the research of identity recognition based on electrocardiogram (ECG) is gradually popular at home and abroad. As a unique physiological signal of human body, ECG naturally has good anti-counterfeiting properties, so it is gradually used in military security, e-commerce and other fields. With the development of portable ECG equipment, convenient and fast finger acquisition has been popularized and has high universality. ECG signal is more and more valuable in the field of identity recognition.In this papaer, an ECG recognition algorithm based on the fusion model of Siamese Neural Networks is proposed. The model combines Siamese Neural Networks and XGBoost, takes into account the deep features and manual features, and integrates the model through LR weighting. Our evaluation achieves 96.31% accuracy in the 90-person-scale database and 98.12% recognition accuracy in the 20-person-scale self-built database.