Heart Rhythm Anomaly Recognition Based on ECG Lorenz Plot

Heart disease has become one of the important diseases threatening people’s life and health. Therefore, automatic recognition and classification of arrhythmia has gradually become a research hotspot. The traditional machine learning identification methods need to manually extract the feature information of one-dimensional ECG signal, which is time-consuming and labor-intensive and affects the accuracy. To solve this problem, this paper proposes to transform one-dimensional ECG data into two-dimensional image ECG lorenz plot, which can better use the feature information of spatial ECG lorenz plot for learning. Secondly, a two-dimensional convolution neural network model is designed to automatically learn the characteristics of graphics, and train and identify five types of abnormal heart rhythm images. Finally, the recognition accuracy of the proposed algorithm for five types of ECG signals can reach 86.5%, which can prove the feasibility of the algorithm.