PCG Classification Using Multidomain Features and SVM Classifier

Table 8

Summary of the proposed features.

Index	Domain	Num. of features	Motivation
1	Time interval	20	The time interval of each state has physiological meaning based on heart physiology.
2	Frequency spectrum of state	308	To reflect the frequency spectrum within state.
3	State amplitude	12	The amplitude is related to the heart hemodynamics.
4	Energy	47	To reflect energy distribution with respect to frequency band
5	Frequency spectrum of records	27	To reflect frequency spectrum within records
6	Cepstrum	65	To reflect the acoustic properties.
7	Cyclostationary	4	To reflect the degree of signal repetition.
8	High-order statistics	16	To reflect the skewness and kurtosis of each signal state.
9	Entropy	16	To reflect the PCG signal inherent complexity.
Total	- -	515	- -