Scalogram-EMD distance for mobile ECGs

Abstract

Now there are devices that are capable of recording ECGs. The distance between signals can be helpful in classification problems for finding ECGs like the given one to know the expected disease scenario. This paper proposes a new distance based on the wavelet decomposition of the signal and earth mover's distance with a new base distance function. It is shown that the introduced distance is a metric over theconsidered signal equivalence classes. In addition, a method for creating new signals based on the developed distance has beenproposed, which can be used to augment data when training deep neural networks. Finally, an experimental study has demonstrated that the generated signals can improve classification quality