ECG-based personal recognition using a convolutional neural network

摘要

To meet increasing security and privacy requirements, ECG-based personal recognition is attracting more attention. This paper proposes a novel feature representation method to extract distinctive features from electrocardiogram (ECG) signals for personal recognition via a deep convolutional neural network. Specifically, it can extract distinctive features from an ECG segment without any reference point detection, avoiding the complicated signal fiducial characteristic points extraction process. Moreover, we use the mean and standard deviation of feature maps as global features for classification. To the best of our knowledge, this is the first attempt to apply this strategy in the field of ECG signals. Unlike most existing methods, the proposed architecture does not require any domain knowledge and is easy to train and optimize. A simple voting step is utilized to facilitate the practical applications of ECG-based personal recognition. Three public ECG datasets are used to evaluate the effectiveness of our method. The proposed method achieves an average recognition rate of 97.7% and 98.7% with the nearest neighbor classifier and support vector machine, respectively, which outperforms most of the state-of-the-art methods. The experimental results demonstrate that our method can well capture distinctive features from the ECG signal and has good generalization ability. (C) 2019 Elsevier B.V. All rights reserved.