This article first introduces the symptoms, incidence and harm of atrial fibrillation (AF) , and then focuses on the time-domain, frequency-domain and non-linear analysis automatic identification technique of AF based on body surface electrocardiogram. Finally, the paper reports the sensitivity, specificity, positive predictive value, and accuracy of using atrial fibrillation recognition algorithm to identify atrial fibrillation, and the advantages and disadvantages of various methods are compared. It is found that the use of multiple R-R interval correlation information for feature extraction can improve the accuracy of detecting atrial fibrillation. In addition, the algorithm based on the R-R interval only requires longer ECG to accurately identify atrial fibrillation, and the recognition accuracy of atrial activity is significantly improved. When atrial ventricular tachycardia occurs, or when the heart rhythm changes rapidly, the signal is more suitable for frequency-domain analysis. The non-linear analysis is an improvement based on the R-R interval algorithm in the time-domain, which can further improve the recognition accuracy.