Study On Automatic Recognition Of Atrial Fibrillation Based On Electrocardiograph

Atrial Fibrillation(AF)is an arrhythmia in which the rhythm of the heart is disturbed,it is prone to cause stroke,hemiplegia and other diseases,so the early detection and treatment of AF is crucial.The initial attack time of AF is short,and the traditional chest-stick collection method is so complicated that it is difficult to record the Electrocardiograph(ECG)in time and realize the instant recognition of AF.Portable handheld devices are simple to operate and can timely collect short-term ECG,which is of great significance for the recognition of AF.However,the quality of ECG collected by portable handheld devices is poor,which makes it difficult to identify AF using short-time ECG.Therefore,how to reduce the influence of poor ECG quality on the recognition of AF and improve the accuracy of automatic recognition of AF using short-term hand-ECG in real-world applications has become a research hotspot.Aiming at the above problems in the recognition of AF,this paper proposes a set of AF recognition method using the time domain characteristics of ECG.The method uses a Chebyshev filter and a median filtering algorithm to remove noise interference in the ECG.Aiming at the singular waveforms caused by limb movement or poor electrode contact during the hand-ECG acquisition process,a singular waveform screening mechanism based on Information Entropy is proposed to calculate the Information Entropy of the ECG segment and remove the singular waveform in the ECG sequence.In this paper,the singular waveform screening mechanism based on Information Entropy is verified on the PCinCC2017 database.The results show that the detection rate of some singular waveforms reaches 99%,the sensitivity of AF recognition increased by 3.39% using the filtered data,and the specificity increased by 6.44%.It proves that the Information Entropy screening mechanism can effectively remove the singular waveform in the ECG signal and reduce the impact of poor quality ECG on the accuracy of AF recognition.After preprocessing the ECG signal,in order to effectively extract the time domain characteristics of the hand-ECG signals collected in the real environment,based on the traditional CWT-based R-wave detection method,the paper proposes an improved R wave detection method based on CWT for the hand-ECG collected in real-world application scenarios with complex data quality,R-wave inversion and low R-wave amplitude.The method uses the extreme point and the angle to determine the R wave position.The accuracy of R wave recognition in the PCinCC2017 and MITDB databases is above 95%,which is11.7% higher than the traditional R wave recognition algorithm.The results show that the proposed method has good universality and robustness,and improves the accuracy of R-wave recognition in real-world applications.In order to verify the validity of the proposed AF recognition method on hand-ECGcollected in a large number of real scenes.In this paper,10 consecutive heartbeat cycle ECG sequences are used as experimental data to extract time domain characteristics of ECG,BP neural network and SVM classifier are trained to identify AF.The sensitivity of AF recognition in both the AFDB database and the PCinCC2017 database is above 94.66%,and the specificity is also higher than 91.76%,confirming the effectiveness of using the time domain characteristics for AF recognition.The method is performed in the PCinCC2017 database to distinguish between AF,normal and other arrhythmia,the correct recognition rate reached 82.69%,and the correct recognition rate manually marked by professional doctors was 63.88%.The experiment result proves the effectiveness of the AF recognition method based on time domain characteristics of hand ECG signals,which has certain significance for the automatic recognition algorithm of AF from theory to practical application.