Research On Preprocessing Algorithms Of ECG Signals

ECG signal is one of important electrical biological signals of human. Working condition and reliable intrinsic characteristic information of heart is reflected from different levels. It has very significant clinical reference value for diagnosis and treatment of heart disease. ECG signal is so weak that it's often affected by varieties of noises in acquisition process, including powerline interference, baseline wander, Electromyography (EMG) interference and random noise, which can make accruate? diagnosis and analysis of cardiovascular disease more difficult. The ECG signal denoising plays the key role in ECG detection, analysis and diagnosis, which directly determines the effect of diagnosis and treatment for clinical cardiovascular disease.In the paper, the ECG signal preprocessing algorithms based on Empirical Mode Decomposition were mainly studied. After analyzing ECG signal noise characteristic, several corresponding filter algorithms?based on EMD were designed to remove powerline interference, baseline wander, EMG interference and other noise from ECG signal by using clinical ECG signals. Simulation results indicated the algorithms proposed in this paper could remove three kinds of noise effectively.The main contribution of this thesis is as follows:The novel powerline denoising method was proposed on the basis of EMD and adaptive filter. ECG signal was decomposed into a series of IMFs by EMD. Then, the IMF which contained powerline interference was sent to 50Hz Notch Filter. Powerline interference was removed by adaptive filter with estimated powerline interference as reference input. Firstly, EMD theory was described. Secondly, the EMD-adaptive filter algorithm?was studied, including design of 50Hz Notch Filter and variable-step LMS algorithm.Finally, algorithm performance was validated with real clinical ECG signals. Results showed that 50Hz powerline interference was removed effectively.The novel baseline wander denoising method was proposed on the basis of EMD and adaptive filter. ECG signal was decomposed into a series of IMFs by EMD. High-order low-frequency IMFs which contained baseline wander was processed by the Moving Average Filter. Baseline wander was removed by adaptive filter with estimated baseline wander as reference input. The denoising performance of traditional Moving Average Filter, Kalman Filter and Band-pass Filter was compared systematically with the novel method. The results indicated that baseline wander was removed effectively. The novel denoising method was proposed on the basis of EMD and threshold denoising. ECG signal was decomposed into a series of IMFs by EMD, ECG signal was reconstructed by IMFs which were shrinkaged with threshold function.Fristly, the generalized threshold function was proposed. Adaptive threshold based on decomposition scale was deduced.The best EMD threshold scheme was obtained after processing four kinds of analogical signals with different characteristic. Comprehensive noises of the clinical ECG signal were removed effectively by the novel methods, while the characteristic wave of ECG signal were remained.The denoising methods of ECG signal proposed based on EMD in the paper remove noises effectively and remain characteristic waves. The methods provide the foundation for the detection, analysis and diagnosis of clinical ECG signals.