| Cardiovascular disease has been the leading cause of death globally for the past20 years,according to the World Health Organization(WHO).Cardiovascular disease deaths have increased by more than 2 million since 2000,increasing to nearly 9 million in 2019.The cardiovascular disease now accounts for 16% of all causes of death.Therefore,a timely and effective diagnosis of cardiovascular disease is essential.Electrocardiogram(ECG)monitoring,processing,identification,and classification technology has become an important auxiliary tool for diagnosing cardiovascular diseases and making treatment plans.An abnormal heartbeat can be detected early for various cardiovascular diseases by monitoring and processing the patient’s ECG signal and classifying it.Traditional time series classification methods include dynamic time warping(DTW),feature-based techniques,deep learning,etc.The main disadvantage of these methods is that they use heuristic manual construction with shallow feature learning architectures,and it is not easy to find the most suitable features.And these features are the key to improving the accuracy of ECG classification.Therefore,the research on feature recognition and classification of new time series is an essential topic in clinical medicine.Aiming at the problem that traditional time series classification methods are easy to ignore the characteristics of nonlinearity and timing in ECG classification,this paper proposes a new ECG classification by mixing time-series imaging(EC-MTSI),which combines the gramian angular field(GAF)recurrence plot(RP)and flat spread realize the two-dimensional ECG timing signal by mixed conversion,which can fully retain the correlation and time dependence of ECG original time series.This paper uses different neural networks to extract features and carry out feature fusion and classification to preserve detailed information and highlight local news.Many experimental results show that time series can be classified by image.Compared with a single transformation method,the method proposed in this paper can effectively improve the accuracy of ECG classification.Multi-model feature fusion can further enhance the model performance and obtain competitive results. |
