Among the 143 papers submitted to the competition from across the world, the best five papers were selected as finalists through a rigorous review process by the AESS Radar Systems Panel Student Paper Competition Committee. In the final round, each participant presented their paper in a live on-line session to the jury.
In Aylin Tastan’s award-winning paper “An Unsupervised Approach for Graph-based Robust Clustering of Human Gait Signatures” she designed a parameter-free robust clustering algorithm to cluster highly contaminated human gait radar data. She extracted a new set of features from the data and deployed a graph-based outlier detection algorithm, using typical degree information of a sparse graph. This algorithm emphasizes the importance of degree information in a sparse graph, which can be valuable also for other purposes, such as weighting functions. In addition, it provides potential a priori information for designing robust sparse graphical models.