We present a new technique MD-EMD based on the wellknown EMD method combined with a modulation and a demodulation of the PCG signals. The modulation step increases the ratio of the component frequencies which helps the EMD to correctly separate components having close frequencies. The results based on simulation signals and real PCG data show the MD-EMD outperforms the classical EMD. When the separation of the first heart sound components is achieved using the MD-EMD, the time split between T1 and M1 is estimated. This time split estimation can be used as an indicator of some cardiac pathologies.