Compression of one-dimensional ECG signals using two-dimensional representations

A new method for compression of one-dimensional signals using two-dimensional representations is investigated. In particular, this work focuses on compression of electrocardiogram (ECG) signals. Electrocardiograms measure the electrical activity of the heart. The pulse beat of the heart is reflected in the ECG by various waveforms. The morphology of the waveforms is useful in diagnosing heart abnormalities, therefore, it is important that the compression algorithm preserves these characteristics. Electrocardiograms exhibit periodic characteristics when observed in the time domain. A two-dimensional representation can be used to distinguish and exploit this periodicity which is not readily apparent in typical one-dimensional representations. By utilizing inter-period correlations of electrocardiogram signals, improved compression rates may be achieved. In this study, ECG compression is accomplished by creating a two-dimensional representation of the ECG. The approach involves dividing the ECG signal into short segments, each representing one pulse beat. The pulse beats are then stacked to form a two-dimensional matrix. The matrix is compressed with an Embedded Zerotree Wavelet (EZW) image encoder. Reconstructed signal quality has been evaluated in two ways: by the Percent Root-Mean-Squared Difference percentage (PRD) and by comparison with results produced by a one-dimensional wavelet encoder. Results indicate that the two-dimensional technique produces high quality results, particularly for very low bit-rates.