Research On Ecg Waveform Detection And Arrhythmia Classification

In recent years, with the improvement of living standard, the prevalence of cardiovascular disease increases significantly, meanwhile, the number of deaths due to cardiovascular disease is rising year after year.Cardiac arrhythmia is one of the more common cardiovascular diseases, also, it is a source of sudden cardiac death. Therefore, how to detect patient’s cardiac arrhythmias timely and accurately is useful to prevent heart disease and sudden cardiac death.Cardiac arrhythmia diagnosis is affected by the doctors’ experience greatly, because it mainly depended on the doctors’ analysis of ECG waveform which could determine the specific type of arrhythmia at the early stage. Also, the wide variety of arrhythmias leads to complex arrhythmia ECG waveform. Therefore, it cannot be satisfied only by manual analysis to the requirement of patients. With the wide application of computer technology, to realize ECG signal detection and arrhythmia classification by intelligent processing technology is becoming a hot spot of research in recent years. However, due to noise interference and individual differences, there still have many problems to realize the fully automatic analysis. This paper will study on the ECG signal waveform detection and arrhythmia classification, main contents and innovations are as follows:(1) A new adaptive threshold estimation method was proposed for removing the baseline drift, electromyogrphy interference, power frequency interference and the other noise in the ECG signal. This threshold has different value in different layers of wavelet decomposition, thus it can denoise self-adaptively. Denoised by wavelet transform based on soft threshold, and experiment results show that the adaptive threshold estimation method is better than other methods, which can better remove all kinds of noise and maximize retention of the signal’s original features. FPGA was used to implement denoising algorithm by wavelet transform based on adaptive soft threshold. Simulation results using modelsim show that the denoising effects based on hardware and software are quite the same.(2) A kind of QRS complex detection algorithm was proposed based on combination threshold. At first difference operation is used to determine the local extreme value points of the signal, then using amplitude threshold to determine the candidate points of R peak, finally wavelet threshold is used to determine the final R peak points. Simulation experimental results show that the algorithm is significantly reduce the amount of calculation and has a good accuracy, instantaneity and robustness.(3) An elite genetic algorithm which global search first, local search after, screening step by step for ECG signal feature selection was proposed. In the algorithm, three genetic operators was proposed, respectively was: “select superior, eliminate inferior and enlarge space” selection operator, splicing operator and cutting operator. Combined with the naive bayesian classifier, the simulation experiment results show that the algorithm can find the optimal feature subset of classification, and achieve arrhythmia classification for ECG signal.(4) A sort of extreme learning machine based on genetic algorithm optimization(GAELM) was put forward for arrhythmia classification. The genetic algorithm was used to optimize learning parameters of hidden layer in the single hidden layer feed forward neural network. At the same time a combined method based on wavelet transform, kernel principal component analysis and GAELM was proposed for ECG signal waveform detection and arrhythmias classification. The experimental results show that the proposed method has faster running time and higher accuracy compared with other algorithms.