| Summary: | 碩士 === 國立交通大學 === 生醫工程研究所 === 103 === Empirical mode decomposition (EMD) had been an adaptive signal processing method in recent year. It was usually applied in non-linear and non-stationary signals in real world. In EMD, the intrinsic mode function (IMF) were extracted through the iterative de-trending process and supplied for analyzed after. However, there was a limitation of real time implementation which was did not done perfectly yet. This limitation was caused by the large computing cost of interpolation and the difficult of parallelization of sifting process, which was the core of EMD. In this study, we proposed fractal empirical mode decomposition (FEMD) which combined the EMD and the fractals analysis. We introduced the box process in fractal analysis into conventional EMD. The raw data would be decimated into several modified data and decomposed by sifting process. All results would be averaged for computing a semi-IMF. Finally, the searching strategy determined proper box number for target components. FEMD make true the leaping searching and decomposing which could implement in parallel. In order of the verification of FEMD, a simulated data, a desired blood pressure (BP) data and a real BP data were used. The results show that the decomposition ability of FEMD was more powerful than conventional EMD. And the ability of parallelization enhanced the real-time implementation. In the future, the FEMD could be implemented in chips of several kinds of biomedical systems or wearable device, which supplied users and doctors the real-time biomedical information.
|
|---|
