Time-Scale Analysis of Signals Without Basis Functions: Application to Sudden Cardiac Arrest Prediction
Sami Torbey, Selim G. Akl and Damian P. Redfearn
Most approaches to the frequency decomposition of signals require the selection of a basis function. This can cause a lack of repeatability in the analysis, particularly when the target signal is not man-made and has no known underlying basis functions (as in many biomedical signals). We propose a novel algorithm that measures frequency content in a signal without any basis function assumption. We then successfully validate it on a set of known test signals generated using a variety of functions and combinations thereof, as well as two collections of electrocardiograms, respectively evaluating the repeatability of the algorithm’s measurements and their ability to predict sudden cardiac arrest.
Keywords: Time-scale analysis, basis functions, wavelet transform, electrocardiogram, sudden cardiac arrest