Research On Preprocessing And R Peak Detection For ECG Signals

Electrocardiogram (ECG) signal is the heart activity’s comprehensive reflection in the body. Itcontains amout of cardiovascular disease’s information which is used to clinical diagnosis. Theanalysis of ECG signal has the importance in diagnosing heart disease. ECG itself is a weakelectrical signal with low frequency and small amplitude. Usually its frequency is between 0.05 Hzand 100Hz, and the amplitude is not more than 4mv. ECG also tends to be polluted by power lineinterference, baseline wander, and muscle artifact. The existence of the noise reduces the accuracyof dignosing, so how to efficitively eliminate the noise which exists in the ECG signal has become afocus that researchers pay attention to. The R peak detection is a significant component in an ECGautomatic analysis system. Accurate and reliable R peak detection can directly influence theaccuracy in ECG analysis. R peak detection is quite important to obtain the heart rate and hear ratevariability.This paper mainly studies the preprocessing algorithms and R peak detection algorithm forECG signals. In the aspect of preprocessing, we propose ECG denoising algorithms that are basedon translation-invariant of sparse coding threshold and based on EEMD and noise statisticaldistribution. These algorithms have well improved the Signal to Noise Rate (SNR) of ECG signal.In the aspect of R peak detection, we propose a R peak detection algorithm based on the bandpassfilter. It employs a Trous algorithm in spline wavelet transforming in order to ensure that thewavelet coefficients are time-invariant. Evaluate this method by employing the clinical ECG signaland the result shows that the proposed algorithm effectively eliminates the noise that exists in theECG signal and realizes the accurate R peak detection.The main work of this paper is as follows:1. The ECG denoising algorithm based on translation-invariant of sparse coding threshold isstudied. Firstly, the basic theory of traditional wavelet threshold denoising is introduced.Translation-invariance is studied. Then a denoising algorithm of an improved threshold functioncalled sparse coding threshold function is set up. Finally, this method is proved by using the analogsignal and ECG signals from the MIT-BIH database.2. The ECG denoising algorithm based on EEMD and noise statistical distribution is studied.Firstly, the basic theories and implementation processes of Empirical Mode Decomposition (EMD)and Ensemble Empirical Mode Decomposition (EEMD) are introduced. Then EEMD and noisestatistical distribution is studied. Complete the confidence interval distribution of noise to separatethe signal component and the noisy component from the Intrinic Mode Function (IMF). Program soft threshold to the noisy component, reconstruct it with the signal component to obtain thedenoised signal. Finally, this method is proved by using the analog signal and ECG signals from theMIT-BIH database.3. An approach by design of bandpass filter and spline wavelet transform to detect R peaks inECG signals is studied. Firstly, a bandpass filter is designed to filter the ECG’s baseline wander.And then the filtered ECG is wavelet transformed by a Trous algorithm in order to ensure that thewavelet coefficients are time-invariant. The performance has been done on four MIT-BIHDatabases. The result shows that the proposed algorithm has good sensitivity and predictivity.The algorithms studied in this paper retain the waveforms of ECG signal, eliminate the noise inECG signal and accurately detect R peaks. The work of this paper lays an important foundation forthe ECG signal’s anlysis and dignosis.