Signal Processing Seminar

Sensor and Machine Learning at The Arizona State University

Andreas Spanias
Arizona State University - SenSIP

This seminar provides a description of the ASU Sensor Signal and Information Processing (SenSIP) center and its application-driven research projects. The center research activities include algorithm development for extracting information from sensors and IoT systems. More specifically center activities are focused on developing signal processing and machine learning methods for various applications including AI-enabled sensing for automotive, IoT solar energy system monitoring, surveillance systems, health monitoring, and sound systems. The center has several industry members that define and monitor research projects typically for Ph.D. student work. SenSIP has also affiliated faculty working on sensor circuits, flexible sensors, radar, smart cameras, motion estimation, secure sensor networks and other systems.


Andreas Spanias is Professor in the School of Electrical, Computer, and Energy Engineering at Arizona State University (ASU). He is also the director of the Sensor Signal and Information Processing (SenSIP) center and the founder of the SenSIP industry consortium (now an NSF I/UCRC site). Member companies of the NSF SenSIP center and industry consortium on sensor information processing include: Intel, National Instruments, LG, NXP, Raytheon, Sprint and several SBIR type companies. He is an IEEE Fellow and he recently received the IEEE Phoenix Section Award for Patents and Innovation. He also received the IEEE Region 6 section award (across 12 states) for education and research in signal processing. He is author of more than 300 papers,15 patents, two text books and several lecture monographs. He served as Distinguished lecturer for the IEEE Signal processing society in 2004.

Additional information ...

PhD Thesis Defence

Graph-Time Signal Processing Filtering and Sampling Strategies

Elvin Isufi

The necessity to process signals living in non-Euclidean domains, such as signals defined on the top of a graph, has led to the extension of signal processing techniques to the graph setting. Among different approaches, graph signal processing distinguishes itself by providing a Fourier analysis of these signals. Analogously to the Fourier transform for time and image signals, the graph Fourier transform decomposes the graph signals decomposes in terms of the harmonics provided by the underlying topology. For instance, a graph signal characterized by a slow variation between adjacent nodes has a low frequency content.

Along with the graph Fourier transform, graph filters are the key tool to alter the graph frequency content of a graph signal. This thesis focuses on graph filters that are performed distributively in the node domain–that is, each node needs to exchange information only within its neighbor to perform a given filtering operation. Similarly to the classical filters, we propose ways to design and implement distributed finite impulse response and infinite impulse response graph filters.

One of the key contributions of this thesis is to bring the temporal dimension to graph signal processing and build upon a graph-time signal processing framework. This is done in different ways. First, we analyze the effects that the temporal variations on the graph signal and graph topology have on the filtering output. Second, we introduce the notion of joint graph-time filtering. Third, we presentpr a statistical analysis of the distributed graph filtering when the graph signal and the graph topology change randomly in time. Finally, we extend the sampling framework from the reconstruction of graph signals to the observation and tracking of time-varying graph processes.

We characterize the behavior of the distributed autoregressivemoving average (ARMA) graph filters when the graph signal and the graph topology are time-varying. The latter analysis is exploited in two ways: i ) to quantify the limitations of graph filters in a dynamic environment, such as a moving sensors processing a time-varying signal in a sensor network; and i i ) to provide ways for filtering with low computation and communication complexity time-varying graph signals.

We develop the notion of distributed graph-time filtering, which is an operation that jointly processes the graph frequencies of a time-varying graph signal on one hand and its temporal frequencies on the other hand. We propose distributed finite impulse response and infinite impulse response recursions to implement a two-dimensional graphtime filtering operation. Finally, we propose design strategies to find the filter coefficients that approximate a desired two-dimensional frequency response.

We extend the analysis of graph filters to a stochastic environment, i.e., when the graph topology and the graph signal change randomly over time. By characterizing the first and second order moments of the filter output, we quantify the impact of the graph signal and the graph topology randomness into the distributed filtering operation. The latter allows us to develop the notion of graph filtering in the mean, which is also used to ease the computational burden of classical graph filters.

Finally, we propose a sampling framework for time-varying graph signals. Particularly, when the graph signal changes over time following a state-space model, we extend the graph signal sampling theory to the tasks of observing and tracking the time-varying graph signal froma few relevant nodes. The latter theory considers the graph signal sampling as a particular case and shows that tools from sparse sensing and sensor selection can be used for sampling.

Additional information ...

PhD Thesis Defence

Pan Liu

PhD Thesis Defence

Efficient computational methods in Magnetic Resonance Imaging

Jeroen van Gemert

This dissertation describes how to design dielectric pads that can be used to increase image quality in Magnetic Resonance Imaging, and how to accelerate image reconstruction times using a preconditioner.

Image quality is limited by the signal to noise ratio of a scan. This ratio is increased for higher static magnetic field strengths and therefore there is great interest in high-field MRI. The wavelength of the transmitted magnetic RF field decreases for higher field strengths, and it becomes comparable to the dimensions of the human body. Consequently, RF interference patterns are encountered which can severely degrade image quality because of a low transmit efficiency or because of inhomogeneities in the field distribution. Dielectric pads can be used to improve this distribution as the pads tailor the field by inducing a secondary magnetic field due to its high permittivity. Typically, the pads are placed tangential to the body and in the vicinity of the region of interest. The exact location, dimensions, and constitution of the pad need to be carefully determined, however, and depend on the application and the MR configuration. Normally, parametric design studies are carried out using electromagnetic field solvers to find a suitable pad, but this is a very time consuming process which can last hours to days. In contrast with these design studies, we present methods to efficiently model and design the dielectric pads using reduced order modeling and optimization techniques. Subsequently, we have created a design tool to bridge the gap between the advanced design methods and the practical application by the MR community. Now, pads can be designed for any 7T neuroimaging and 3T body imaging application within minutes.

In the second part of the thesis a preconditioner is designed for parallel imaging (PI) and compressed sensing (CS) reconstructions. MRI acquisition times can be strongly reduced by using PI and CS techniques by acquiring less data than prescribed by the Nyquist criterion to fully reconstruct the anatomic image; this is beneficial for patient's comfort and for minimizing the risk of patient's movement. Although acquisition times are reduced, the reconstruction times are increased significantly. The reconstruction times can be reduced when a preconditioner is used. In this thesis, we construct such a preconditioner for the frequently used iterative Split Bregman framework. We have tested the performance in a conjugate gradient framework, and show that for different coil configurations, undersampling patterns, and anatomies, a five-fold acceleration can be obtained for solving the linear system part of Split Bregman.

Additional information ...

MSc SS Thesis Presentation

Improving Ultrafast Doppler Imaging using Subspace Tracking

Bastian Generowicz

Ultrafast Doppler imaging provides a new way to image blood motion at thousands of frames per second. It has gained popularity due to its high spatio-temporal resolution, which is required to distinguish blood motion from clutter signals caused by slow moving tissue. By conducting functional UltraSound (fUS) experiments on the brain using this method, we are able to better understand the underlying processes during brain activity through neurovascular coupling. fUS relies on optimized signal processing techniques to acquire and process high frame-rate images in real-time.

For my thesis I have set up the backbone to allow for fUS experiments as well as created the analysis framework required to analyse and interpret the incoming data. Furthermore, I have developed a more computationally efficient method of obtaining vascular images, based on the Projection Approximation Subspace Tracking (PAST) method. The PAST algorithm is able to display accurate representations of the blood subspace, while maintaining a lower computational complexity than the state-of-the-art method, making it suitable for Doppler imaging. When applied to functional ultrasound, the exponentially weighted PASTd method achieves similar Pearson Correlation coefficients compared to the current state-of-the-art method, over multiple functional experiments. These findings highlight the potential of applying PAST to Ultrafast Doppler imaging.

Additional information ...

PhD Thesis Defence

Aleksandar Jovic

Additional information ...