Diagnostic application of a universal kernel to time-frequency energy distribution of ECG signals

Abstract

Time-frequency representation (TFR) is a way of describing the time-variant energy of a signal. Cohen's class of distributions is one of the most commonly used methods for generating TFRs. A universal kernel (UK) function that is capable of filtering out any cross terms occurring in the Wigner-Ville distribution (WVD) is proposed for time-variant energy distribution of ECG signals. The proposed technique promises detection of low level potentials occurring in an abnormal ECG pattern. Also, TFR images of filtered WVD of ECGs show potential in diagnostic applications.