A hybrid Fourier/wavelet technique for improved ECG signal approximation

With the growing ubiquity of digital record keeping, succinct representation of ECG signals subject to a reconstruction accuracy level assumes significance. In this regard, inadequacy of Fourier analysis is well known, whereas relative efficacy of wavelet representation has recently been demonstrated. In this backdrop, we propose a novel hybrid technique that improves upon wavelet methods. We observe that the larger share of ECG signal energy is contained in a low-pass band, which admits an efficient Fourier representation, whereas the remainder of the signal comprises of temporally localized features, compactly represented by wavelets. Accordingly, we propose a hybrid Fourier/wavelet technique that efficiently exploits both the low-pass and the high-pass components. Finally, using signals derived from ANSI/AAMI EC13 test waveforms (standard for testing medical equipments), we demonstrate lower volume of encoded data required by the proposed technique to achieve a target approximation accuracy in comparison to both pure Fourier and pure wavelet strategies.