Tadesse Ghirmai- Publications

UNIVERSITY of WASHINGTON | BOTHELL

Electrical Engineering | Science & Technology

Home Teaching Research Publications Other

Projects

Extraction of fetal ECG from a maternal ECG

The objective of this project is to develop a robust signal processing algorithm for extracting a fetal Electrocardiogram (fECG) from a single recording signal obtained from the abdomen of a pregnant woman. This project is part of a broader effort that aims at building devices for monitoring the cardiac activity of a fetus and a mother during pregnancy and labor. Non-invasive and continuous monitoring of the fetus’ heart functionality during pregnancy would provide not only information about the well-being of the fetus, but also gives insights into the investigation and prevention of congenital heart diseases. Unlike the common clinical practice of investigating the recordings of fetal heartrate (fHR) and heartrate (HR) variability, the fetus heart functionality is better monitored when the full fECG waveform is extracted. The signal recorded from the abdomen area of the mother includes the fECG signal with very low SNR buried by the high-noise signals of mECG, electromyogram (EMG) and motion artifacts. This project, therefore, focusses on the challenge of extracting the morphology of the fECG, with minimal distortion, from a single-channel signal recording without any other reference maternal ECG signal.

Interference Cancellation in Heterogeneous Networks

The explosion of new wireless services has already triggered for the investigation of the deployment of the 5G communication network which is expected to be deployed in 2020. The new 5G communication network is envisioned to support multimedia at a higher data rate with high quality of service, reduced latency and less energy consumption. One of the key enabling technologies for achieving the promised function of 5G is the densification of existing cellular networks and enabling peer-to-peer (P2P) communication among nodes. Such capability is realized through having a multi-tiered heterogeneous network architecture that contains several microcells, relays and other low power nodes within a macrocell. Due to reduced cell sizes and high spectrum reuse, heterogeneous network along P2P communication undelaying cellular architecture can significantly enhance capacity, increase spectral efficiency and reduce energy consumption of the communication network. However, heterogeneous networks are plagued with inter-cell and intra-cell interferences. The promised benefits of heterogeneous networks can only be achieved if these interferences are managed. This research proposal explores signal processing as well as interference management schemes, such as power control strategies, to reduce or cancel interferences.

Estimation of parameters in high-order nonlinear dynamic systems

The problem of the estimation of the state parameters of a dynamic system can be formulated in a state-space form. If the formulation of the dynamic system is linear and Gaussian, the state estimation problem has an optimal analytic solution obtained using the celebrated filter known as Kalman filter. However, many practical problem of interest have a representation which is nonlinear and/or non-Gaussian. In such cases, particle filters are the most popular techniques for estimating time-varying parameters in nonlinear and/or non-Gaussian dynamic systems. While a particle filter method displays excellent performance for many applications, it is plagued with two notable problems – complexity and the curse of dimensionality. The complexity of a particle filtering algorithm increases proportionally with the number of particles used, and the number of particles for a typical application is rather high. More importantly, for high-order problems, problems with many unknown parameters, particle filtering based algorithms suffer from degeneracy. This research work explores other frameworks that address the curse of dimensionality of particle filtering for estimating parameters in high-order problems.

UNIVERSITY of WASHINGTON \| BOTHELL Electrical Engineering \| Science & Technology


Home Teaching Research Publications Other
Past Projects	Projects Extraction of fetal ECG from a maternal ECG The objective of this project is to develop a robust signal processing algorithm for extracting a fetal Electrocardiogram (fECG) from a single recording signal obtained from the abdomen of a pregnant woman. This project is part of a broader effort that aims at building devices for monitoring the cardiac activity of a fetus and a mother during pregnancy and labor. Non-invasive and continuous monitoring of the fetus’ heart functionality during pregnancy would provide not only information about the well-being of the fetus, but also gives insights into the investigation and prevention of congenital heart diseases. Unlike the common clinical practice of investigating the recordings of fetal heartrate (fHR) and heartrate (HR) variability, the fetus heart functionality is better monitored when the full fECG waveform is extracted. The signal recorded from the abdomen area of the mother includes the fECG signal with very low SNR buried by the high-noise signals of mECG, electromyogram (EMG) and motion artifacts. This project, therefore, focusses on the challenge of extracting the morphology of the fECG, with minimal distortion, from a single-channel signal recording without any other reference maternal ECG signal. Interference Cancellation in Heterogeneous Networks The explosion of new wireless services has already triggered for the investigation of the deployment of the 5G communication network which is expected to be deployed in 2020. The new 5G communication network is envisioned to support multimedia at a higher data rate with high quality of service, reduced latency and less energy consumption. One of the key enabling technologies for achieving the promised function of 5G is the densification of existing cellular networks and enabling peer-to-peer (P2P) communication among nodes. Such capability is realized through having a multi-tiered heterogeneous network architecture that contains several microcells, relays and other low power nodes within a macrocell. Due to reduced cell sizes and high spectrum reuse, heterogeneous network along P2P communication undelaying cellular architecture can significantly enhance capacity, increase spectral efficiency and reduce energy consumption of the communication network. However, heterogeneous networks are plagued with inter-cell and intra-cell interferences. The promised benefits of heterogeneous networks can only be achieved if these interferences are managed. This research proposal explores signal processing as well as interference management schemes, such as power control strategies, to reduce or cancel interferences. Estimation of parameters in high-order nonlinear dynamic systems The problem of the estimation of the state parameters of a dynamic system can be formulated in a state-space form. If the formulation of the dynamic system is linear and Gaussian, the state estimation problem has an optimal analytic solution obtained using the celebrated filter known as Kalman filter. However, many practical problem of interest have a representation which is nonlinear and/or non-Gaussian. In such cases, particle filters are the most popular techniques for estimating time-varying parameters in nonlinear and/or non-Gaussian dynamic systems. While a particle filter method displays excellent performance for many applications, it is plagued with two notable problems – complexity and the curse of dimensionality. The complexity of a particle filtering algorithm increases proportionally with the number of particles used, and the number of particles for a typical application is rather high. More importantly, for high-order problems, problems with many unknown parameters, particle filtering based algorithms suffer from degeneracy. This research work explores other frameworks that address the curse of dimensionality of particle filtering for estimating parameters in high-order problems.