Research Of Detection Algorithm Of Atrial Fibrillation Based On RR Interval And Sparse Decomposition

With increasing numbers of aging people and factors leading to atrial fibrillation(AF), AF has become the most common arrhythmia in our country, even all over the world. Therefore,early diagnosis of atrial fibrillation has important clinical and social significance in reducing patient’s morbidity and mortality as well as economic burden.At present, the existed AF detection algorithm has not already solved the problem of feature extraction appropriately. As a result, the error rate of AF detection is still high. Sparse decomposition can achieve data compression efficiently, and what is more important, the dictionary redundancy features can be used to capture the essence of characteristics of the signal. In this paper, the superiority of sparse decomposition will be applied to the detection of AF algorithm. An important feature of AF is that RR interval is absolutely irregular in ECG, which leads to some intensive research. Thus we designed the detection algorithm of atrial fibrillation based on RR interval and sparse decomposition. The main content of this paper is as follows:(1) ECG signal preprocessing. In this paper, Δ RR interval histogram will be used to preprocess normal and AF signals, so as to make the distribution characteristics of AF RR interval more apparent. Choosing 101 as the width “M” of histogram is to improve the image resolution. However, to have dimension not increased, only the efficient part of histogram needs to be remained, to the contrary, redundant part of “0” should be removed and dimension decreased to 15 in order to improve the integral flexibility. The result of experiment shows that this method can not only elevate accuracy, but also reduce the computing time.(2) The application of sparse decomposition in AF detecting algorithm. A number of samples, chosen at random from the histogram data after preprocessed, will be used in the design of atom dictionary. On this basis, two subclass dictionaries of normal and AF ECG signal are constructed. Seeking for ECG signal sparse representation on the two dictionaries respectively, which can keep more characteristics of the signal. At last, we classify AF signal by LS-SVM.The result of experiment shows that this algorithm reaches a high degree of accuracy when detecting AF. Also, it proves that sparse decomposition has feasibility and validity on AF detecting algorithm.