Detection of Doppler Embolic Signals with Hilbert-Huang Transformn

In order to identify the Doppler ultrasound embolic signals for stroke prediction, it is important to confirm the embolic appearing time and frequency of movement. The paper proposes a new identification method based Hilbert-Huang Transform. The embolic signal is decomposed into a number of intrinsic mode functions which are "mono-component". Then Hilbert transform is applied to them to calculate the instantaneous frequencies and Hilbert amplitudes. The embolic appearing time can be confirmed according to the maximum amplitude which should be larger than the preset threshold. The corresponding instantaneous frequency can also be decided with the time-frequency distribution. The method is applied to 24 Doppler embolic signals. The time error for more than 95% Doppler signals is less than 70 ms. The result shows the powerful potential of HHT will have a promising future in embolic signals identification and stroke prediction in the clinical application.