A wavelet method for the noise reduction in electrocardiographic signals

The electrocardiogram (ECG) is the routinely used biomedical signal for diagnosis of cardiovascular diseases, and the removal of noise in ambulatory ECG recordings is essential in a number of clinical applications. In this paper, we present a Daubechies wavelet analysis method with a decomposition tree of level 5 (Wdb5)for analysis of noisy ECG signals. The implementation includes the procedures of signal decomposition and reconstruction with hard-thresholding. The experiments were tested with seven ambulatory ECG records from the benchmark MIT-BIH Ar- rythmia Database, and our results demonstrate the effectiveness of the Wdb5 analysis method for the noise reduction in ECG signals. Furthermore, the quantitative study of result evaluation indicates that the Wdb5 filtering method is superior to the popular least-mean-square (LMS) filter by achieving significantly higher signal-to-noise ratio and better filtered-noise entropy values.