The Research Of Abnormal Electrocardiogram Automatic Detection Algorithm

As the enhancement of people’s living standard and the development of computer technology, the combination of computer and medical promotes the study of abnormal electrocardiogram detection. In the study of the automatic detection of abnormal electrocardiogram, electrocardiogram preprocessing methods mainly include the band-pass filter and wavelet transform method. Anomaly diagnosis methods mainly include feature extraction method, support vector machine method and the neural network and so on. The feature extraction algorithm is complicated. Support vector machine method and neural network cost a long time, so they are not suitable for real-time detection.The paper mainly studied the automatic detection of abnormal electrocardiogram, mainly divided into preprocessing, feature parameters of electrocardiogram abnormal detection and diagnosis. The paper first to extract electrocardiogram data of binary from arrhythmia database provided by Massachusetts Institute of Technology and Beth Israel Hospital(MIT-BIH database), and then to deal with the noise in it. Electrocardiogram preprocessing operations include removal of high frequency noise and baseline drift signal caused by the low frequency signal in electrocardiogram. First using wavelet transform to extract the electrocardiogram signal in the low frequency component, which filter out the baseline drift signal, and then compare and analyze the threshold functions and threshold selection method. The paper use unbiased likelihood estimation threshold and soft threshold function to smooth after wavelet transform of the high frequency component processing, so as to realize the high frequency noise filtering, which achieved good pretreatment effect. After the pretreatment for electrocardiogram signal, extreme value point is acquired using the first-order difference operation, combining with the threshold method to obtain the R wave and T wave location parameters, then detect Q wave and S wave, and finally testing P wave. In the diagnosis of abnormal, first making the initial template by normal electrocardiogram template obtained by electrocardiogram characteristics, then using multiple template matching method to build the template library. Obtaining the template which is matching with electrocardiogram to be detected by using correlation coefficient, then diagnosing the electrocardiogram.In the study of abnormal electrocardiogram detection, we use MIT-BIH database data to test result analysis, it is better to remove the baseline drift, can remove most of the high frequency noise signal, anomaly detection accuracy is better. But the high frequency noise filtering effect is still need to be improved. The efficiency of anomaly detection also has room for improvement.