Research And Implementation Of ECG Signal Abnormal Classification

In recent years,the death rate due to Cardiovascular Disease(CVD)has been rising.The average number of deaths per year in the world is approximately 17.7 million.According to statistics from the Beijing Emergency Medical Center,71%of the deaths of heart patients are due to the loss of the best rescue time.Therefore,identifying abnormal heart rhythms and their types in time has become an important research topic in recent years.Abnormal heart rhythm,that is irregular heartbeat,it will lead to irregular rhythms.At present,the most common method of clinical diagnosis is to rely on the experience of doctors to make judgments.However,ECG signals vary from person to person.Doctors process a large amount of ECG data every day,and they are easily misjudged based on experience.Therefore,it is very important for clinical medical diagnosis to use an algorithm to dig out the intrinsic features of ECG and automatically classify abnormal heart rhythms.significance.Based on the above analysis,this article carried out the following work:1.ECG signal preprocessing.ECG signal preprocessing includes three aspects:denoising,feature point identification.and feature extraction.In this paper,noise reduction based on complementary empirical noise based collective empirical mode decomposition is used.A real-time detection of QRS complex in ECG signals is used to timely and effectively detect waveform feature points for accurate positioning.In order to fully excavate the characteristics of ECG signal,this topic discusses the time-domain characteristics,energy index characteristics and frequency-domain characteristics of ECG from different perspectives.Finally,15 feature attributes are extracted.2.Rhythm abnormality classification algorithm.Taking into account the small and varied characteristics of the data set,this topic classifies the abnormal heart rhythm types based on the gcforest algorithm.In addition,this paper also based on the multi-task learning algorithm can use the characteristics of commonalities and differences between related tasks to solve multiple ECG recognition and classification tasks.Compared with the single-task learning algorithm,it can not only improve the task's learning efficiency,but also effectively improve the accuracy of the task's prediction.The performance of the multi-task learning and gcforest algorithm was evaluated through the data in the MIT-BIH database.The results show that the multi-task learning algorithm can better recognize the type of heart rhythm abnormalities,in which the accuracy of gcforest algorithm reaches 99.5%,and the accuracy of multi-task learning reaches 99.56%.3.Based on the Android platform,the abnormal heartbeat system is built.In order to monitor the patient's heartbeat data in real time,this article also built a heart rhythm variability analysis system on Android and wrote a corresponding application program.In summary,this article through the MIT-BIH data to test,the results show that the algorithm used in this paper can accurately identify the type of arrhythmia can be applied to clinical diagnosis.