Research On ECG Signal Processing Method Based On Time-Frequency Analysis

As a kind of bioelectrical signal containing a lot of pathological information,electrocardiogram(ECG)signal is widely used in the clinical diagnosis of heart disease,and it is an important material to promote the application of telemedicine.As a weak,nonlinear and unstable physiological signal,ECG signal has the characteristics of low amplitude,low signal-to-noise ratio and random noise.High quality ECG signal can help doctors identify physiological characteristics and diagnose pathological phenomena.Aiming at the characteristics of ECG,this thesis studies the preprocessing,feature extraction and medical application of ECG,and proposes a series of analysis methods for multi-type ECG records.The main research contents are as follows1.An automatic ECG signal preprocessing algorithm based on complementary ensemble empirical mode decomposition and signal screening is proposed.By fusing the complementary set empirical mode decomposition algorithm with the signal similarity algorithm,the signal with noise in the decomposition result is screened out By studying the characteristics of the baseline drift signal,the signal with baseline drift is screened out by using the zero crossing rate characteristics of the decomposition result.The automatic screening of the decomposition result and the automatic de-noising of the noisy signal are realized,and the empirical mode decomposition problem is solved effectively The problem of mode aliasing in decomposition algorithm and the problem of noise residue and large amount of calculation in ensemble empirical mode decomposition algorithm are discussed.The comparison experiments on MIT-BIH database ECG data set show that the proposed method is better than the traditional filter denoising method and wavelet based denoising method for both synthetic and real noisy signals.2.An algorithm based on fast multi-dimensional empirical mode decomposition(mdemd)and continuous wavelet transform(CWT)is proposed to extract fetal heart rate(FHR)from multi-dimensional abdominal ECG signals of mothers.Focusing on extracting fetal R-peak features from maternal abdominal ECG signals,using the extracted R-peak position to form fetal RR time series is helpful to analyze fetal heart rate variability in clinical intervention,especially in the autonomic nervous system.By fusing fmemd algorithm and wavelet features,this method can effectively eliminate the pollution of external noise and internal mother’s QRS complex,extract fetal heart rate features from multi-dimensional ECG signals of mother’s abdomen,realize fetal health monitoring during pregnancy,and implement fetal health monitoring in abnormal and direct fetal ECG Database data sets verify that the method can obtain higher accuracy detection results,and it is a faster detection algorithm.