Fig. 2
(1) Feature evaluation flow diagram. (2) Features were extracted for the time domain, frequency domain, and time–frequency domain. (3) Individual features were then evaluated across classifiers and the top five features were highlighted. (4) Feature sets within individual domains were evaluated via the two wrapper methods (4a) followed by the filter method (4b) to select the optimal feature set (4c). (5) The optimal feature sets produced from the individual domains were then combined and (4ab) were repeated to produce the optimal combined feature set (5a). (6) Recommendations for a generalized congenital feature set were then made.
