Empirical Mode Decomposition Method Based on Cardinal Spline and its Application on Electroencephalogram Decomposition

This paper presents an improved empirical mode decomposition method called cardinal-spline empirical mode decomposition (CS-EMD). Unlike the classical empirical mode decomposition (EMD), the proposed method uses cardinal splines instead of cubic splines for signal envelope estimation. The decomposition performance of the CS-EMD method on synthetic signals is compared to the classical EMD method using performance evaluation indices. The orthogonal indices OIavg and OImax for an intermittent signal using CS-EMD are 0.0024 and 0.0105, respectively, compared to those of the classical EMD of 0.4438 and 1.9537 (closer to 0 is desirable). The energy conservation index (ECI) for the intermittent signal using CS-EMD is 0.9198 compared to 13.4496 using the classical EMD (closer to 1 is desirable). For a synthetic signal with components of close frequencies, the performance evaluation indices are OIavg=0.0019, OImax=0.0095, and ECI=0.8800 for CS-EMD and OIavg=0.0719, OImax=1.7821, and ECI=9.6610 for the classical EMD. Both EMD methods were also applied to an electroencephalogram (EEG), and the amount of mixed modes were observed and compared. The results show that the signal decomposition properties using CS-EMD are more desirable than those of the classical EMD, providing an improved EMD method for biosignal processing applications.