Cascaded Neural Network For Classification of Artficially Modeled ECG Beats Using Error Signal

摘要

This paper proposes a potential cascaded neural network for ECG Holter system beat classification. The system works in real time and is capable of recognizing up to 20 artificial QRS templates. The parallelism of neural network increases the efficiency of computations. An error signal derived from differences between predictor and testing signal is used in the classification. A neural network is used to generate linear predictions for signals. Another neural network generates error signals measured between predictions taken from, first neural network and testing signal. A third neural network does the classifications utilizing the error signals instead of complex raw signal. Three Her-mite functions are used in generating testing signals with, and without noise. Proper thresholding for the error signals are essential for classifier immunity. The result is a compact, online, efficient, and hardware realizable signals classifier that uses minimal compressed error signal.