A dynamical model for generating synthetic ballistocardiogram signals based on Extended Kalman Filter (EKF)
Date of Issue2016-08-22
School of Electrical and Electronic Engineering
In recent years, the ballistocardiogray technology receives many interests due to the development in both measurement methods and signal processing techniques. As a non-intrusive method for obtaining representation of the cardiovascular performance, it can be used as an effective as well as economical tool for long-term home monitoring of cardiovascular diseases. In our study, the ballistocardiogram (BCG) data were obtained from fiber optic sensors which put in the seat mat of a chair, which has the characteristics of lighter weight and higher accuracy compared with popular-used methods, such as force plate and static charge-sensitive sensor.In this dissertation we proposed a BCG dynamical model in combination with Extended Kalman Filter (EKF) for BCG signals which is able to generate synthetic BCG signals as well as reduce nonlinear noises in real BCG signal. EKF algorithm is used in nonlinear denoising process to obtain BCG waveform which can approach standard BCG morphology. The feature points were extracted after the desired waveforms were obtained. Then the time and phase information, like Beat-to-Beat, IJ and JK time intervals within a typical BCG waveform were calculated. The proposed BCG dynamical model is composed of three coupled ordinary differential equations i.e. Gaussian kernel functions. The synthetic BCG signal which can illustrate beat-to-beat trajectory variation in BCG morphology was then outputted from the dynamical model. This model was evaluated by adding white and Gaussian noises artificially on several normal and real BCGs and synthetic BCGs, and the output SNR and morphology changes of the filter outputs was investigated. Results show that the dynamical model can be effectively used as a valid application in synthetic bio-signal generation and nonlinear system processing. In the near future, this novel model can be utilized to evaluate bio-signal processing algorithms and help extract clinical views from the real BCG signal.
DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Signal processing