Aboveground Storage Tank (AST) tightness testing using statistical approach

The industry partner, Cantest is establishing a new leak detection procedure for analyzing data sources in aboveground storage tanks and statistical learning models to monitor AST shell dynamics and product activity over time. This is an important problem as identifying leak detection is usually associated with various environmental data and records collected from sensitive sensors attached to the ASTs. Current testing procedure for leak detection uses simple statistical rules and thresholds to detect anomalies. These methods are failing for preventing AST related environmental incidents.

Developing statistical methods to discover genetic variants underlying longitudinal decline in lung function

COPD is a common inflammatory lung condition that is characterized by airflow limitation and symptoms of cough and shortness of breath. Globally, it affects 384 million people and is responsible for ~4-7% of all deaths. Longitudinal genome-wide association studies (GWAS) are needed to unravel the molecular determinants of dynamic quantitative traits underlying COPD, such as decline in lung function over time.
Analysis of longitudinal GWAS to find biomarker of lung function decline was unsuccessful in the past. None of the discovered biomarkers were replicable.

Prediction models for pain volatility and engagement patterns of mobile pain app users

Pain is among the top 3 most common reasons for seeking medical help. ManagingLife has developed a mobile-based app, called Manage My Pain, to help chronic-pain patients by providing a simplistic, customizable and comprehensive interface to track pain symptoms and pain experience at a frequency chosen by the users. ManagingLife is interested in understanding the benefits of the app use on its 27,000 and constantly growing user base by identifying user cohorts that ultimately experience improvement in pain experience given self-disclosure tracking behaviours.

The fixating effect of LCA during ideation in eco-design: a case study

Although life cycle assessment (LCA) is a robust eco-design tool, its capacity to inspire creative ideas among designers is unclear. LCA is often used as a portrait of a product’s environmental hotspots, which are then addressed as design compromises. This highly technical design method usually leads to incremental improvements of products, and appears to do little to foster radical innovation.

Evaluation and analysis of citizen science golden eagle migration data collected by the Rocky Mountain Eagle Research Foundation, Mount Lorette, Alberta 1993-2016

The goal of this project is to evaluate the data quality of the citizen science Golden Eagle count data collected by the Rocky Mountain Eagle Research Foundation and to determine the population trend over 25 years. Since 1993 they have collected data, consisting of daily raptor migration counts in the Fall and Spring seasons. It is vital to determine the usefulness of the citizen science data, check and evaluate its quality and determine whether available datasets are suitable for further research or not.

Design of a Multi-modal Electronic Stethoscope for the Digital Acquisition and Automatic Diagnosis of Auscultation Signals

The proposed research project aims to develop a multi-modal stethoscope, containing superior digitized heart and lung sounds, telemedicine capabilities and assistive diagnostics. This is achieved by leveraging new advancements in piezo, microphone, wireless and machine learning technologies. The project will investigate these technologies and integrate them into custom made electronics and mechanical designs to achieve an optimal digitized sound that provides superior auscultation capabilities to medical professionals for lung and heart sound diagnosis.

Development of signal processing techniques for animal movement data

In the past decade, the development of sophisticated sensors attached to animals (tags) have researchers to infer of horizontal and vertical movement of marine animals across time and space. The amount of data collected from these tags along with the analytical challenges surrounding the extraction of behavioural patterns has presented a significant barrier for researchers to adopt this technology.

Detection of Fights in Crowd Video

Detection of fights and anomalous behavior of individuals in a crowd is a common problem in computer vision. Some tools that currently exist rely on optical flow of tracked features is a sequence of video frames. These motion algorithms are sensitive to independently moving objects in the frame. What constitutes an “anomaly” is context (eg. location) specific, thereby adding to the complexity.

Feature selection for Deep Learning applied to the identification of impaired drivers

DriveABLE Inc uses a set of simple video tasks to identify the impaired drivers. Video tasks come in the form of simple games and measure cognitive ability. The test results are analysed by AI powered algorithm that predicts the impairment level of the driver. Our project’s main objective is to redesign the AI in such a way that it can cover more use cases with fewer tasks. In particular we will redesign the algorithm so that it will accept incomplete tests. We will also identify redundant games in order to make overall test shorter.

Advanced pricing methods for property and casualty isurance

Pricing risks is of pivotal importance for the insurer’s well-being. Indeed, inappropriately determined prices, whether too high or too low, may result in insolvency of insurance policies, failure of business lines, and even bankruptcy of entire insurance enterprises. This project will help Wawanesa Insurance to develop sophisticated pricing techniques that will take into account (a) exogenous pricing factors, and (b) interdependencies among risks. Wawanesa Insurance will therefore benefit from the resulting competitive advantage.