Gå til hovedinnhold
0
Jump to main content

Synthetic Micro-Doppler Signatures of Non-Stationary Channels for the Design of Human Activity Recognition Systems

Ahmed Abdelmonem Abdelgawwad of the Faculty of Engineering and Science at the University of Agder has submitted his thesis entitled «Synthetic Micro-Doppler Signatures of Non-Stationary Channels for the Design of Human Activity Recognition Systems» and will defend the thesis for the PhD-degree Monday 16 August 2021. (Photo: Private)

The outcomes of this dissertation pave the way towards simulation-based human activity recognition systems.

Ahmed Abdelmonem Abdelgawwad

PhD Candidate

Ahmed Abdelmonem Abdelgawwad of the Faculty of Engineering and Science at the University of Agder has submitted his thesis entitled «Synthetic Micro-Doppler Signatures of Non-Stationary Channels for the Design of Human Activity Recognition Systems» and will defend the thesis for the PhD-degree Monday 16 August 2021.

He has followed the PhD-programme at the Faculty of Engineering and Science, with spesialisation in ICT.

Summary of the thesis by Ahmed Abdelmonem Abdelgawwad:

Synthetic Micro-Doppler Signatures of Non-Stationary Channels for the Design of Human Activity Recognition Systems

The research interest in human activity recognition (HAR) to provide ambient assisted living (AAL) for elders living alone has increased.

As a result, several HAR systems have been proposed. The categories of HAR systems fall into two types: wearable and non-wearable systems.

Radio-frequency (RF)-based HAR systems are considered among the non-wearable types. They do not violate the users’ privacy. Moreover, they do not require to be worn by the user. Furthermore, they are capable of sensing human activities through walls of buildings. That is why they catch more interest nowadays.

Captures non-stationary behavior

RF-based HAR systems capture the complex non-stationary behaviors of indoor channels influenced by human activities.

Micro-Doppler signatures and time variant mean Doppler shifts (TV-MDSs) are considered among the extracted features used to train the human activity classifiers (HACs).

The current approach is to train HACs by using measured RF-sensing data. The collection of such measured data is expensive and time-consuming. Furthermore, the collected data is not reproducible. An alternative approach is to generate the micro-Doppler signatures and the TV-MDSs of non-stationary channel simulation models for training the HACs.

The main aim of this dissertation is to generate synthetic micro-Doppler signatures and TV-MDSs to train the HACs. This is achieved by developing non-stationary fixed-to-fixed (F2F) indoor channel models. Such models provide an in-depth understanding of the channel parameters that influence the micro-Doppler signatures and TV-MDSs. Hence, the proposed non-stationary channel models help to generate the micro-Doppler signatures and the TV-MDSs, which fit those of the collected measurement data.

From 2D to 3D F2F channel models

First, we start with a simple two-dimensional (2D) non-stationary F2F channel model with fixed and moving scatterers. Such a model assumes that the moving scatterers are moving in 2D geometry with simple time variant (TV) trajectories and they have the same height as the transmitter and the receiver antennas. The model of the Doppler shifts caused by the moving scatterers in 2D space is provided. The micro-Doppler signature of this model is explored by employing the spectrogram of which a closed-form expression is derived. Moreover, we demonstrate how the TV-MDSs can be computed from the spectrograms.

The aforementioned model is extended to provide two three-dimensional (3D) nonstationary F2F channel models. Such models allow simulating the micro-Doppler signatures influenced by the 3D trajectories of human activities, such as walking and falling. Moreover, expressions of the trajectories of these human activities are also given. Approximate solutions of the spectrograms of these channels are provided by approximating the Doppler shifts caused by the human activities into linear piecewise functions of time. The impact of these activities on the micro-Doppler signatures and the TV-MDSs of the simulated channel models is explored.

Measured channels included

The work done in this dissertation is not limited to analyzing micro-Doppler signatures and the TV-MDSs of the simulated channel models, but also includes those of the measured channels.

The channel-state-information (CSI) software tool installed on commercial-off-the-shelf (COTS) devices is utilized to capture complex channel transfer function (CTF) data under the influence of human activities. To mitigate the TV phase distortions caused by the clock asynchronization between the transmitter and receiver stations, a back-to-back (B2B) connection is employed. Models of the measured CTF and its true phases are also shown.

The true micro-Doppler signatures and TV-MDSs of the measured CTF are analyzed. The results showed that the CSI tool is reliable to validate the proposed channel models. This allows the micro-Doppler signatures and the TV-MDSs extracted from the data collected with this tool to be used to train the HACs.

Two non-stationary CSI models

Inertial measurement units (IMUs) can be used to capture the trajectories of moving objects or human activities.

In this dissertation, two IMU-driven non-stationary CSI models are presented. Such models can be fed with trajectories collected by the IMUs to simulate realistic micro-Doppler signatures and the TV-MDSs.

The aim of the first (second) model is to capture the influence of the trajectory of a rigid body (moving human) on the micro-Doppler signatures and the TV-MDSs of CSI channels. Frameworks are proposed for processing the IMU data to compute the trajectories and feed them to the CSI model to simulate the micro-Doppler signatures and the TV-MDSs. Both of IMU data and CSI data are recorded simultaneously to validate the proposed channel models. Then, the micro-Doppler signatures and the TV-MDSs of the IMU-driven models and CSI data are evaluated.

The results show that there are strong agreements between the micro-Doppler signatures and the TV-MDSs of the IMU-driven model and the measured CSI.

The outcomes of this dissertation pave the way towards simulation-based HAR systems.

Disputation facts:

The trial lecture and the public defence will take place online, via the Zoom conferencing app (registration link below)

Dean Michael Rygaard Hansen, Faculty of Engineering and Science, University of Agder, will chair the disputation.

The trial lecture at 10:15 hours

Public defence at 12:15 hours

Given topic for trial lecture«Contact-Free Human RF-Sensing: Basic Principles, Challenges, and Solutions»

Thesis Title«Synthetic Micro-Doppler Signatures of Non-Stationary Channels for the Design of Human Activity Recognition Systems»

Search for the thesis in AURA - Agder University Research Archive, a digital archive of scientific papers, theses and dissertations from the academic staff and students at the University of Agder.

The thesis is available here:

https://hdl.handle.net/11250/2767221

 

The Candidate: Ahmed Abdelmonem Abdelgawwad (1991, Minya, Egypt) Bachelor’s and master’s degrees (Hons.) in information engineering and technology from the German University in Cairo (GUC), in 2013 and 2015, respectively. Master’s degree was related to the complexity reduction of 3D ray-tracing for indoor coverage solutions. His research interests include network modeling and simulation, channel modeling for fall detection systems, non-stationary channel models, and time-frequency analysis for non-stationary channel models. Present position: Software Engineer in STRYDE in Asker, Norge - see linkedIn.

Opponents:

First opponent: Professor Klaus David, Universität Kassel, Tyskland 

Second opponent: Associate Professor Stephan Sigg, Aalto University, Finland 

Professor Christian Omlin, Department of Information and Communication Technology, University of Agder, is appointed as the administrator for the assessment commitee.

Supervisors in the doctoral work were Professor Matthias Pätzold, UiA (main supervisor) and Associate Professor Bjørn Olav Hogstad, NTNU (co-supervisor)

Opponent ex auditorio:

The chair invites members of the public to pose questions ex auditorio in the introduction to the public defense, with deadlines. It is a prerequisite that the opponent has read the thesis. Questions can be submitted to the chair Michael Rygaard Hansen on e-mail michael.r.hansen@uia.no