Pervasive Sensing & Embedded Systems

Pervasive Sensing addresses advanced communication, collection, storage and analysis of big data from pervasive sensors such as wearable, implantable and ambient sensors to facilitate continuous monitoring of a person.

Embedded System Communications takes into consideration wearable and implantable wireless sensor network solutions and real-time cooperation of embedded and wearable sensor networks. Across the healthcare sector the capability to utilise such embedded system communications enables monitoring of patient’s health status at any time. Across the transport sector embedded systems communications opens the research gates in areas such as embedded traffic monitoring systems or intelligent vehicles autonomous capabilities, self-regulatory, and self-repairing systems to improve safety, driver comfort, and efficient use of infrastructures. 

Key areas of expertise:

• Natural Language Processing (NLP): This research strand focuses on the application of NLP for use in remote patient care and a hospital environment. Here we leverage NLP to manage user interactions with platforms through vocal responses to the NLP requests where machine learning monitors speech changes which can be assessed. By leveraging open-source audio analysis technologies to train federated models learning from a user’s particular vocal traits, and classifying changes in different features e.g. pitch, loudness, speaking rate, articulation, negative trends in their health can be identified. Such monitoring of changes in speech/emotion over time provides Personalised Emotion Inference.
• Explainable AI: As the reach of AI grows and we see it transforming industries such as medicine, transport and defence, we find ourselves entrusting our health, safety and security to intelligent machines. Many worry that these machines are “black boxes” i.e. closed systems that receive an input, produce an output, and offer no clue why. Here, we approach AI Explainability from several angles, considering model reproducibility as well as model transparency.
• Disease Modelling: The capability to remotely analyse depression markers using federated learning to predict negative trends in the mental health of cancer patients that have undergone therapy, giving healthcare providers advanced warnings to allow for timely intervention. A particular focus is to analyse speech for disease biomarkers. This research strand will encompass several areas including:
• Analysis of the voice samples and visualisation of the constituent features
• Deployment of privacy-preserving algorithms e.g., Federated Learning
• Overlap and integration with current NLP approaches
• Leveraging modern web tools e.g., React and TensorFlow.js

• NetApps and Core Networks: This research strand focuses on the application of network architecture knowledge for the design and generation of optimised core network elements in the realisation of a dynamic and robust design allowing for the ease of integration of third party applications into the core 5G telecoms network. Creation of a 5G application testbed within the Walton data centre provides a playground for experimentation in the design and implementation of VNFs and their associated vertical applications.
• Quality of Service & Quality of Experience in applications: This strand investigates the application of metrics and KPIs interpretation for the generation of an overall score that can be representative of the end users Quality of Experience (QoE). The outputs allow for the appropriate re-configuration of both the core network components and the strategy employed by the applications themselves for the content generation and consumption.
• Transport Dynamic Routing (Traffic Management): Here we are creating a framework of the ingestion of traffic data (routes, density, direction, infrastructure, etc.) where an analysis of that data can be carried out with the result being a dynamic traffic management system that would route and move traffic based on time, direction, density, priority. Such dynamic routing outcomes are compared against the actual travel times for the vehicles to demonstrate just how efficient the dynamic routing system is while also generating metrics such as time saved, CO2 emissions reductions, etc.