Electrocardiogram compression using modulus maxima of wavelet transform

This paper presents a new ECG compression method for applications in both ECG transmission and ECG data storing. The local modulus maxims of the wavelet transform of an ECG signal indicate the locations of sharp variation points in the signal that carry the most important information. Although a signal cannot be exactly reconstructed from the modulus maxima of its wavelet transform, we can obtain a close approximation of the original signal from these modulus maxima. By ignoring small modulus maxima and encoding only the locations and amplitudes of the remaining modulus maxima and the coarse scale signal, we can compress ECG signal significantly. Considering different energy levels at different scales, suitable thresholds are determined to remove unnecessary modulus maxima, such as those generated by noise and artifacts. With a high compression ratio (CR) and a low percentage root-mean-square (PRD), our approach to ECG compression compares favorably with other methods tested.