Seminars and Defenses

All are welcome and encouraged to attend the seminars and defenses.

Aug 26, 11AM, ENG460
Robust Discriminative Analysis Framework For Gaze And Head-Pose Estimation
Salahaldeen Rabba • PHD FINAL ORAL EXAM
Head movements, combined with gaze, play a fundamental role in predicting a person's action and intention. In non-constrained head movement settings, the process is complex, and performance can degrade significantly in the presence of variation in head-pose, gaze position, occlusion and ambient illumination. In this thesis, a framework is therefore proposed to fuse and combine head-pose and gaze information to obtain more robust and accurate gaze estimation.
Specific contributions include: the development of a newly developed graph-based model for pupil localization and accurate estimation of the pupil center; the proposal of a novel iris region descriptor feature using quadtree decomposition, that works together with pupil localization for gaze estimation; the proposal of kernel-based extensions and enhancements to a fusion mechanism known as Discriminative Multiple Canonical Correlation Analysis (DMCCA) for fusing features (proposed and traditional) together, to generate a refined, high quality feature set for classification; and the newly developed methodology of head-pose features based on quadtree decompositions and geometrical moments, to better integrate roll, yaw, pitch and jawline into the overall estimation framework.
The experimental results of the proposed framework demonstrate robustness against variations in illumination, occlusion, head-pose and is calibration free. The proposed framework is validated by achieving an accurate gaze and head-pose estimation of 4.5° using MPII, 4.4° using Cave, 4.8° using EYEDIAP, 5.0° using ACS, 4.1° using OSLO and 4.5° using UULM datasets respectively.
Aug 12, 1PM, ENG460
Safely caching HOG pyramid feature levels, to speed up facial landmark detection
Gareth Adam Higgins • MASC oral exam
This thesis presents an algorithm for improving the execution time of existing Histogram of Oriented Gradients (HOG) pyramid analysis based facial landmark detection. It extends the work of [1] to video data. A Bayesian Network (Bayes Net) is used as a policy network to determine when previously calculated features can be safely reused. This avoids the problem of recalculating expensive features every frame. The algorithm leverages a set of lightweight features to minimize additional overhead. Additionally, it takes advantage of the wide spread adoption of H.264 encoding in consumer grade recording devices, to acquire cheap motions vectors. Experimental results on a difficult real world data set show that policy network is effective, and that the error introduced to the system remains relatively low. A large performance benet is realized due to the use of the cached features.
July 23, 1PM, ENG460
Energy Saving Schemes for Scalable Mobile Computing Networks
Ali Alnoman • PHD internal oral exam
With the ever-increasing numbers of connected devices, the provision of radio and computing resources with high quality of service (QoS) is becoming more and more challenging. In addition, the dense deployment of radio and computing nodes requires extra amounts of energy that can be largely wasted during idle times. Therefore, it is necessary to develop new energy saving strategies and improve network scalability to meet the future needs of mobile networks. In our research, we consider both heterogeneous cloud radio access networks (H-CRANs) and cloud-edge computing to introduce new energy saving techniques that consider the actual utilization of network elements such as base stations and virtual machines, and implement on/off mechanisms taking into account the QoS requirements. Moreover, we propose a non-orthogonal multiple access (NOMA)-based resource allocation scheme in the context of Internet of Things aiming to improve network scalability and reduce the energy consumption of mobile devices. First, the energy saving mechanism is formulated as a 0-1 knapsack problem and solved using dynamic programming. Afterwards, the energy saving mechanism is applied on edge computing to reduce the numbers of under-utilized virtual machines in edge devices. Herein, the square-root staffing rule and the Halfin-Whitt function are used to determine the minimum number of virtual machines required to maintain the queueing probability below a threshold value. On the user level, reducing energy consumption can be achieved by maximizing the data rate provision that reduces the task completion time. To this end, we implemented the sparse code multiple access (SCMA) scheme that allows subcarriers to be shared by multiple users. Not only does SCMA help to provide higher data rates but also to increase the number of accommodated users. Herein, a power optimization and codebook allocation problems are formulated and solved using the water-filling approach and a heuristic algorithm, respectively. Results showed the significance of the proposed schemes on the network performance with respect to both energy and scalability while satisfying the QoS requirements.
July 18, 10AM, ENG460
Yashodhan Athavale • PHD internal oral exam
The Internet of Things (IoT) framework is a trending model in the wake of recent advancements in wirelesscommunications, cloud services, ubiquitous sensors, and smart devices. Today, the IoT model is rapidly being deployed in communications, infrastructure, transportation and healthcare services. The Internet of Medical Things (IoMT) is a subset of the telehealth framework and provides a layered architecture for connecting individuals with mobile devices and wearables, such that their vital physiological data can be captured and analysed non-invasively using smart sensors embedded within these devices. Currently available wearables have embedded sensing modules for measuring movement, direction, light and pressure. Actigraphs are one such type of wearables which exclusively employ the use of accelerometers for capturing human movement-based vibration data. The objective of our research work is the analysis of unstructured, non-stationary actigraphy signals. We intend to develop an IoMT-based and device-independent actigraphy analysis system for identifying types of daily activity, markers for neuromuscular diseases, physical disabilities and joint disorders. The proposed system encodes, adaptively segments and analyses regions of peak activity at the data acquisition source. In order to test the efficiency of the proposed system, in our study, we have used four different actigraphy datasets from wake and sleep states. >From our experiments we found that in comparison to conventional signal filtering and analysis methods based on manufacturer specifications, employing a lower level of signal quantization coupled with a novel, activity-specific, adaptive segmentation technique, ensures a 20-90% increase in SNR (signal-to-noise ratio), 50-80% reduction in bit rate, 50-90% data compression, and an increase in activity recognition accuracy by over 10%. In addition to this, we have also validated our research with ground truth information and machine learning approaches. Results from our systematic investigation indicate that data analysis right at the acquisition source, optimizes signal denoising, memory and power savings, and activity recognition, thereby promoting an edge computing approach to physiological signal analysis using wearables in a low resource environment.
July 8, 12PM, ENG460
Multimodal Information fusion for Human Action Recognition
Nour Eldin Elmadany • PHD internal oral exam
This thesis presents three frameworks of human action recognition to facilitate better recognition performance. The first framework aims to fuse hand crafted features from four different modalities including RGB, Depth, Skeleton, and Accelerometer data. Moreover, a new descriptor for skeleton is proposed which provides a discriminative representation for the poses of the action. In the first framework, we propose two fusion techniques, Bimodal Hybrid Centroid Canonical Correlation Analysis (BHCCCA) for two sets of features or modalities and Multimodal Hybrid Centroid Canonical Correlation Analysis (MHCCCA) for three or more sets of features or modalities, aiming to find a more discriminative subspace. The second framework fuses hand crafted and deep learning features from three modalities including RGB, Depth, and Skeleton. In this framework, a new depth representation is introduced which extracts the final representation using Deep ConvNet. The backbone of the framework is the proposed fusion techniques: Biset Globality Locality Preserving Canonical Correlation Analysis (BGLPCCA) for two sets of features or modalities and Multiset Globality Locality Preserving Canonical Correlation Analysis (MGLPCCA) for three or more sets of features or modalities. BGLPCCA/MGLPCCA aims to preserve the local and global structure of data while maximizing the correlation among different modalities or sets. The third framework uses the deep learning techniques to improve the long term temporal modelling through two proposed Temporal Relational Network (TRN) and Temporal Second Order Pooling Based Network (T-SOPN). Also, Global-Local Network (GLN) and Fuse-Inception-Nets (FIN) are proposed to encourage the network to learn the complementary information about the action and the scene itself. Qualitative and quantitative experiments are conducted on ten different datasets and the experimental results demonstrate the effectiveness of the proposed frameworks over frameworks over state-of-the-art methods.
July 3, 12PM, ENG460
Modeling Multi-site Computation Offloading in Unreliable Cloud Environments
Marzieh Ranjbar Pirbasti • MASC oral exam
Offloading heavy computations from a mobile device to cloud servers can reduce the power consumption of the mobile device and improve the response time of mobile applications. However, the gains of offloading can be significantly affected by failures of cloud servers and network links. In this thesis, we propose a fault-aware, multi-site computation offloading model capable of finding near-optimal allocations of tasks to resources. Our model reduces both response time and energy consumption by incorporating the effect of failures and recovery mechanisms for various offloading allocations. In addition, we create a fault-injection framework to evaluate an allocation under various failure rates and recovery mechanisms. The experiments carried out with our fault-injection framework demonstrate that our fault-aware model can determine an allocation--based on the type of failure (server failure or network link failure), failure rates, and the employed recovery mechanisms--that reduces both response time and energy consumption compared to a model without failures.
June 13, 11AM, ENG460
Omid Karimpour • Meng project oral exam
Over the last decade, navigation and Simultaneous Localization and Mapping (SLAM) have become key players in developing robust mobile robots. Several SLAM approaches utilizing camera, laser scan, sonar and fusion of sensors were developed and improved by a number of researchers. In this thesis, comparisons of these methods were evaluated, especially those offering low cost benefits, and low computation and memory consumption. The aim of this thesis was to select the most reliable and cost-efficient approach for indoor autonomous robotic applications. Currently, there are numerous studies that have optimized these SLAM methods; however, they still suffer from various complications such as scale drifting and excessive computation. This study performed different experiments to observe these challenges in real-world environments. A modified Pioneer robot was used to implement the selected SLAM system and furthermore, perform obstacle avoidance and path planning in indoor office environments. The results and tests show the reliable performance of Gmapping after tuning its parameter and set right configurations.
May. 14, 10AM, ENG460
Towards building a clinically-inspired ultrasound hub: Design, Development and Clinical Validation of Ultrasound probes for Imaging, Therapeutics, Sensing and other applications
Dr. Amir Manbachi • Research Seminar
Ultrasound is a relatively established modality with a number of exciting, yet not fully explored applications, ranging from imaging and image-guided navigation, to tumor ablation, neuro-modulation, piezoelectric surgery, and drug delivery. In this talk, Dr. Manbachi will be discussing some of his ongoing projects aiming to address low-frequency bone sonography, minimally invasive ablation of neuro-oncology and implantable sensors for spinal cord blood flow measurements.

Dr. Manbachi is an assistant research professor at Johns Hopkins School of Medicine. He received his Ph.D., M.A.Sc., B.A.Sc. degrees all from University of Toronto in 2015, 2010, and 2008 respectively. He was a Postdoctoral Scholar at Brigham and Womens Hospital in Harvard-MIT Division of Health Sciences and Technology in 2016.
His research interests include engineering design, development and clinical validation of acoustic probes (ultrasound transducers) and minimally invasive medical devices for clinically inspired diagnoses and treatments.
Dr. Manbachi is an author on 25 peer-reviewed journal articles, 30+ conference proceedings, 10 invention disclosures / patent applications and a book. His interdisciplinary research has been recognized by a number of awards, including University of Toronto's 2015 Inventor of year award, and Ontario Brain Institute 2013 fellowship.
May. 6, 2PM, ENG460
Modeling Hybrid Metaheuristic Optimizer for Convergence Prediction
Noel Jose Thengappurackal Laiju • MENG PROJECT ORAL EXAM
The project aims at the design and development of six hybrid nature inspired algorithms based on Grey Wolf Optimizer with Artificial Bee Colony Optimizer (GWOABC), Moth Flame Optimization Algorithm with Ant Lion Optimizer (MFOALO), Cuckoo Search Optimizer with Fire Fly Optimization Algorithm(CSFFA), Multi-Verse Optimizer with Particle Swarm Optimization Algorithm (MVOPSO), Grey Wolf Optimizer with Whale Optimization Algorithm (GWOWOA), and Binary Bat Optimization Algorithm with Particle Swarm Optimization Algorithm(BATPSO). Hybrid optimizations assume the implementation of two or more algorithms for the same optimization problem. "Hybrid algorithm" does not refer to simply combining multiple algorithms to solve a different problem but rather many algorithms can be considered as combinations of simpler pieces.The hybrid approach combines algorithms that solve the same problem but differs in other characteristics notably performance. A hybrid optimization uses a heuristic to choose the best of these algorithms to apply in a given situation. The proposed hybrid algorithms are benchmarked using a set of 23 classical benchmark functions employed to test different characteristics of hybrid optimizers. The results of the fitness functions prove that the proposed hybrid algorithms are able to produce better or more competitive output with respect to improved exploration, local optima avoidance, exploitation, and convergence. All these hybrid algorithms find superior optimal designs for quintessential engineering problems engaged, showcasing that these algorithms are capable of solving constrained complex problems with diverse search spaces. Optimization results demonstrate that all hybrid algorithms are very competitive compared to the state-of-the-art optimization methods and validated by fitness function. The hybrid algorithms are applied for optimal efficiency determination in various design challenges based on cantilever beam problem.
May 6, 10AM, ENG460
Non-Orthogonal Multiple Access Techniques: Opportunities and Challenges
Dr. Mojtaba Vaezi • Research Seminar
Future radio access technologies are expected to support a massive number of users with a diverse set of requirements in terms of delay, throughput, etc. In view of emerging applications such as the Internet of Things (IoT), and to fulfill the need for massive numbers of connections with diverse requirements in terms of latency and throughput, 5th generation (5G) and beyond cellular networks are experiencing a paradigm shift in design philosophy: shifting from orthogonal to non-orthogonal techniques in waveform, multiple access, and random-access design. A common feature of newly designed multiple access (MA) schemes is the use of non-orthogonal multiple access (NOMA) schemes in lieu of the conventional orthogonal schemes. By allowing multiple users to share the same time/frequency/code/space, NOMA scales up the number of served users, increases the spectral efficiency, and improves user-fairness compared to existing orthogonal MA techniques.

This talk investigates the potentials of the emerging NOMA techniques for 5G and beyond mobile networks. We will first briefly review the MA methods of 1G to 4G cellular networks. After motivating the need for NOMA in 5G and beyond networks, the theoretical basis of NOMA will be discussed, and potential MA methods will be outlined. Details of power allocation and user pairing will be described, and the possibility of combining NOMA with multiple-input multiple-output (MIMO) technologies both in the single-cell and multi-cell setting will be explored. Practical implementation issues will also be discussed. Lastly, challenges and future research directions, including combining NOMA with massive MIMO and millimeter wave communications, security concern, and potential benefit of NOMA from machine learning and deep learning algorithms will be explained.

Mojtaba Vaezi received his Ph.D. in Electrical Engineering from McGill University in 2014 and held several research positions at Princeton University from 2015 to 2018. He is currently an Assistant Professor of ECE at Villanova University, PA, USA. His research interests include the broad areas of wireless communications, signal processing, and information theory, with an emphasis on fifth generation (5G) radio access technologies, physical layer security, Internet of things (IoT), and machine learning for communications. Among his publications in these areas is the book Multiple Access Techniques for 5G Wireless Networks and Beyond (Springer, 2019). Dr. Vaezi has served as the president of McGill IEEE Student Branch and the head of Mobile Radio Network Design and Optimization Group at Ericsson. He is/was an Editor of IEEE Communications Letters and IEEE Communications Magazine and the lead co-organizer of five NOMA workshops at IEEE VTC 2017-Spring, Globecom’17, ICC’18, Globecom’18, and ICC’19. Dr. Vaezi is a recipient of several academic, leadership, and research awards, including the McGill Engineering Doctoral Award, the IEEE Larry K. Wilson Regional Student Activities Award in 2013, the NSERC Postdoctoral Fellowship in 2014, and the Ministry of Science and ICT of Korea’s best paper award in 2017.
May 3, 10AM, ENG460
Electrical Micro and Nano Devices and Sensors
Dr. Arezoo Emadi • Research Seminar
The development of highly sensitive diagnostic and monitoring tools that extends well beyond human senses and perception is an essential strategy for more cost-effective and practical health and environmental maintenance. Today, micromachined sensor systems and wearable electronics are being successfully adapted and adopted to bring leading-edge technologies that transfer significant benefits of micromachining and integration to the fields of medicine and environmental monitoring. The acceleration in micromachined sensors’ implementation is primarily due to their potential for integration, device miniaturization, low power consumption, better performance, lower cost and higher reliability. This seminar introduces ongoing research efforts to introduce emerging smart sensor systems, a new generation of BioMEMS and a highperformance ultrasonic imaging system as practical, non-invasive and sensitive diagnostic tools for medical application. The potential of these technologies for routine treatment efficiency monitoring are further explained with the aim to facilitate cost-effective and more accessible secondary preventive strategies.

Dr. Arezoo Emadi is an Assistant Professor in the Department of Electrical and Computer Engineering, University of Windsor. She joined the University of Windsor in July 2017. Dr. Emadi received her Ph.D. degree from the Department of Electrical and Computer Engineering at the University of Manitoba. Her ongoing research efforts are in the area of Micro Electro Mechanical Systems (MEMS), medical MEMS sensors and transducers, bio and chemical sensors, advanced diagnosis sensor technologies, micro and nano electronic devices and fabrication and medical imaging systems. She has led academic and industry crossfunctional projects to introduce and implement next generation micromachined smart sensor systems in a wide range of fields that make abundant use of sensors and transducers and to deliver the benefits of these technologies to a wider segment of the world’s population. She is the principal author/co-author of over 50 journal and conference papers as well as patents and book chapter in this field. Dr. Emadi also holds an Adjunct Professor position at the Department of Electrical and Computer Engineering, University of Manitoba. She is a Senior Member of IEEE and a Professional Engineer, PEng, and an active member of the Windsor Cancer Research Group, WCRG.
May. 2, 3PM, ENG471
Performance Evaluation Of A Big Data Application On Apache Spark
Apache Spark enables a big data application—one that takes massive data as input and may produce massive data along its execution—to run in parallel on multiple nodes. Hence, for a big data application, performance is a vital issue. This project analyzes a WordCount application using Apache Spark, where the impact on the execution time and average utilization is assessed. To facilitate this assessment, the number of executor cores and the size of executor memory are varied across different sizes of data that the application has to process, and the different number of nodes in the cluster that the application runs on. It is concluded that all the four factors— number of executor cores, size of executor memory, number of nodes in the cluster, and the size of input data—impact the performance of the application.
May 2, 10AM, ENG460
Adaptive Electromagnetic Structures for Wireless Communications and IoT Applications
Dr. Marco A. Antoniades • Research Seminar
The recent emergence of a new class of engineered materials that have electromagnetic properties not typically found in nature, known as metamaterials, has led to the creation of novel electromagnetic structures that exhibit new phenomena and demonstrate superior qualities compared to their conventional counterparts. Notable among these are metasurfaces that allow the arbitrary manipulation of fields passing through them. In this presentation, an overview of recent work conducted in the area of electromagnetic metamaterial structures will be presented. The basic operation of transmission-line based metamaterials as these relate to guided-wave and antenna applications will be presented. This will be followed by various metamaterial-based antennas and RF/microwave devices for use in IoT and biomedical applications such as implantable/wearable devices and microwave imaging. Subsequently, compact, multi-band and highly efficient designs will be presented for wireless communication applications. Finally, ongoing and future work on adaptable metasurfaces for wireless communication applications will be discussed.

Marco A. Antoniades (IEEE S’99, M’09, SM’17) received the B.A.Sc. degree in electrical engineering from the University of Waterloo, ON, Canada, in 2001, and the M.A.Sc. and Ph.D. degrees in electrical engineering from the University of Toronto, ON, Canada, in 2003 and 2009, respectively. He is the Director of the Microwaves and Antennas Laboratory, a Founding Member of the EMPHASIS Research Centre, and an Assistant Professor in the Department of Electrical and Computer Engineering at the University of Cyprus. He is also an Adjunct Fellow in the School of Information Technology and Electrical Engineering at The University of Queensland, Australia. His research interests include passive and active antenna design, RF/microwave circuits and metamaterials/metasurfaces for use in broadband wireless communications, biomedical devices and imaging, space/satellite applications, radio-frequency identification, and wireless power transfer. He is an Associate Editor of IEEE Antennas and Wireless Propagation Letters (AWPL), and a member of the IEEE Antennas and Propagation Society (AP-S) Education Committee.