The University of Southampton

Pump-priming fund supports new strategic research activities

Published: 22 April 2021

ECS' cross-cutting strategic research centres (C-IoT, CMI and CHT) recently announced a lightweight call for small pump-priming activities in the areas of IoT, Machine Intelligence, and/or Health Technologies. Submitted proposals were evaluated by a panel comprising of the Heads of the three Cross-Cutting Centres, and assessed on their ability to maximise the impact of the limited funding. Proposals were particularly encouraged which incorporated new collaborations, and activities ultimately leading to the submission of larger research proposals.

The following three proposals were funded:

Contactless Fit-Mask Testing System

Healthcare workers face the risk of airborne diseases. Respiratory protective equipment (RPE) such as facemasks are vital in preventing nosocomial transmission but depend on a good seal to protect the user. Dr Hansung Kim is researching a simple and contactless 3D face modelling system to match a respirator most likely to fit using two cameras to simultaneously capture and build a full 3D model of the user's face. He is currently working with colleagues in the Faculty of Medicine the Faculty of Health Sciences to perform a shadow pilot trial to assess the effectiveness of this system. The pump-priming fund will be used to purchase a suitable capture system for this trial, consisting of two cameras with solid tripods and a camera slider.

Acoustic Scene Rendering for Machine Listening & Robot Audition

Acoustic signals encapsulate a wealth of semantic information about the environment as well as the listener. The ability to make sense of sound is therefore a fundamental prerequisite for Artificial Intelligence (AI). However, in practice, acoustic signals are affected by reverberation, noise, and interference from competing sources. Perhaps due to these challenges, machine listening has, to date, received only little attention and, as such, datasets that are suitable for training machine learning algorithms are not readily available. The pump-priming fund is providing Dr Christine Evers with underpinning equipment that is necessary to render and control realistic acoustic scenes.

Evaluation of dataset discovery algorithms in a wild scenario

With the proliferation of datasets available for analysis, their effective and efficient discovery becomes a relevant research problem. Recent research (Koutras et al., ICDE 2021) has presented a comprehensive benchmark of schema matching techniques for dataset discovery, but it has been noted that their evaluation was done on near-optimal conditions, hypothesising that application in the wild would perform significantly worse. Dr Luis-Daniel Ibáñez aims to advance towards the testing of this hypothesis by evaluating the suitability, cost and performance of state-of-the-art matching algorithms for finding matches in the datasets indexed by the European Data Portal (EDP), and the pump-priming fund will provide both researcher and Microsoft Azure time for this.

Articles that may also interest you

Share this article FacebookTwitterWeibo