Adaptive identification of nonlinear models for data storage and compression
T. Schimming, Herve Dedieu, MACIEJ J. OGORZAŁEK · 1998
We propose an adaptive identification scheme for a low-dimensional nonlinear model of the human heart's ECG dynamics. We show that this scheme is suitable for data compression and possibly the detection of diagnostically significant features. Tests on real clinically measured ECG signals confirm a very good performance of the model in terms of modelling error and compression ratio. I. INTRODUCTION The identification of nonlinear models is of great interest in various fields including biology, medicine and economics. Here we consider a model for the Electrocardiogram (ECG), a recording (measurement) of the electrical activity generated by the heart carried out using sensors positioned on the body surface, which will be identified. The analysis of the ECG signal provides the most common non-invasive method to diagnose cardiac disfunctions. As in most data storage and transmission applications, a well performing compression scheme is essential for achieving a good storage and transmissi...