Research On Ecg Denoising,Fetal ECG Extraction,Compressed Sampling And Reconstruction

Electrocardiogram(ECG)denoising is an important research direction in the field of ECG research.Electromyogram(EMG)noise is the main noise contained in ECG signals,and their spectrum overlap with adult and fetal ECG signals.In the time domain,it shows sudden and intermittent changes.Empirical mode decomposition(EMD)is applied to ECG signal denoising,which is helpful to solve the problem of spectral overlap which cannot be solved by traditional filtering.However,there are still some problems that the fixed threshold cannot adapt to the intermittent variation of EMG noise,and the soft and hard threshold denoising strategy cannot identify and retain the useful data less than the threshold.Fetal ECG extraction from mixed ECG signals is an important means to realize fetal health diagnosis,pregnancy monitoring and telemedicine.Blind source separation(BSS)has the advantage of separating independent source signals from mixed signals with low signal-to-noise ratio(SNR).However,due to the weak correlation between maternal and fetal ECG signals,the effect of fetal ECG extraction is not good.How to eliminate the correlation between maternal and fetal ECG signals to improve the extraction success rate is the key problem of BSS in fetal ECG extraction.With the rapid development of Internet and telemedicine,the method of compressed sampling and reconstruction of mixed ECG signals has become one of the key technologies to be solved in the field of fetal ECG remote monitoring.Due to the non-sparsity of mixed ECG signals and the complexity of both maternal and fetal ECG signals,no more references have been found except the study of compressed sensing for mixed ECG signals by using the block sparse Bayesian learning method.Integrated empirical mode decomposition(IEMD)is proposed to provide a decomposition tool for ECG signal denoising and fetal ECG signal extraction.Based on the completeness feature of signal decomposition of CEEMDAN,the construction method of intermediate auxiliary noise is improved to solve the pseudo-modal problem.Complementary noise is introduced to reduce the residual noise in intrinsic mode functions(IMFs),thus reducing the total amount of noise and improving the decomposition efficiency.The smoothing filtering method is adopted for the initialization of IMF cycle screening to improve the screening speed and avoid the distortion of spline interpolation caused by the QRS extreme envelope.Through the decomposition of pure and noisy simulated ECG signals,it is verified that IEMD method can effectively improve the decomposition quality of ECG signals.It can avoid mode aliasing and pseudo-mode phenomenon,and keep the characteristics of binary filter banks and the completeness of ECG decomposition.Compared with EEMD,IEMD reduces the total noise quantity to less than 50 pairs.By changing SNR to decompose the noisy ECG signals,the results show that,for EMD,there is still mode aliasing phenomenon when the SNR is higher.IEMD still has an obvious advantage in the decomposition of noisy ECG signals.Aiming at the removal of EMG noise and white noise,a denosing method for ECG signal based on IEMD time-varying threshold and peak loop eliminating strategy(IEMD-TTh PE)is proposed.Based on IEMD decomposition of ECG signal,combined with Pauta criterion and windowing-adding method,the time-varying threshold is proposed to address the amplitude variability and time intermittency of EMG noise.In order to ensure the identification and retention of useful data less than the threshold on QRS in the IMFs of ECG signals,the peak loop eliminating strategy is studied.In order to eliminate the isolated noise larger than the threshold,an interval threshold assisted denoising method is established.By adding EMG noise and white noise to the adult ECG signals and the mixed ECG signals respectively,the denoising is carried out to verify that the proposed method can effectively suppress the EMG noise and white noise in the ECG signals.Compared with the existing EMD threshold denoising methods,IEMD-TTh PE is superior to other methods in terms of SNR improvement and correlation coefficient.Aiming at the weak correlation between maternal and fetal ECG signals in mixed ECG signals,which leads to the low success rate of extracting fetal ECG sign als by quasi-periodic component extraction(QPCE)method,a fetal ECG signal extraction method based on IEMD(IEMD-QPCE)is proposed to eliminate maternal ECG interference.According to the binary filter bank characteristics of IEMD and the different corresponding frequencies of maternal and fetal QRS energy peaks in mixed ECG signals,the main energies of the two are decomposed into different IMFs,so as to eliminate the related interference of maternal QRS.Maternal QRS interference in the IMFs of multi-channel mixed ECG signal is eliminated and then FECG is reconstructed by adding the IMFs of FECG extracted by QPCE,which effectively improves the success rate of extracting FECG signals by QPCE.By extracting fetal ECG signals from the abdominal mixed ECG signals before and after denoised,it is verified that the extraction success rate of IEMD-QPCE is significantly higher than that of PCA,QPCE and ICA.The results show that the fetal ECG signals extracted successfully by IEMD-QPCE have good quality.In view of the non-sparsity of mixed ECG signals,a compressed sampling and reconstruction method for mixed ECG signals based on the the shift of spectrum slice and the baseband sampling is proposed.The mixed ECG signals are decomposed into multiple single-band slices by slicing the effective frequency range.After the frequency band of each slice signal is shifted to the baseband by sine and cosine function modulation,Nyquist rate sampling is carried out according to the baseband to realize the compression of the mixed ECG signals.The mixed ECG signals are reconstructed from the compressed sampled slice signals by the process of zero filling,low pass filtering interpolation,restoring to the original sampling rate sequence and reversed shifting the frequencies before adding them together.The results show that the Nyquist rate sampling based on the baseband can ensure the integrity of the information in the slice frequency bands,and the reconstructed signal does not lose the useful information in the original signal,which effectively improves the quality of reconstruction.Different from the classic compressed sensing method,the proposed method does not require sparse transformation and redundant channels larger than the sparsity to ensure a unique sparse solution,resulting in good stability and little loss of useful frequency information.Through the compression and reconstruction of mixed ECG signals in ADFECGDB and NIFECGDB databases,the results verify that the proposed method can effectively impro ve the compression rate and reconstruction quality of useful frequency components.The comparison between the extracted fetal ECG signals from the original and reconstructed data in the Da ISy database shows that the compressed and reconstructed mixed ECG s ignals in the proposed method do not affect the extraction of fetal ECG.