Electrical and Electronic Engineering - Masters by Research Theses
Permanent URI for this collection
Browse
Recent Submissions
Item Implementation of an AI-assisted sonification algorithm on an edge device(University College Cork, 2023) O'Sullivan, Feargal; Popovici, Emanuel; Temko, Andriy; QualcommOxygen deprivation at birth leads to brain injury, which can have serious consequences. It is the dominant cause of seizures. Quickly and accurately detecting seizures is a challenging problem for neonates. A severe shortage of medical professionals with the necessary expertise for Electroencephalogram (EEG) analysis leads to significant delays in decision-making and hence treatment. These problems are made worse in disadvantaged communities. Artificial intelligence (AI) techniques have been proposed to automate the process and compensate for the lack of available expertise. However, these models are ’black boxes', and their lack of explainability dampens the wide adoption by medical professionals. AI-assisted sonification adds explainability to any such automated methodology, empowering medical professionals to make accurate decisions regardless of their level of expertise in EEG analysis. The feasibility of an implementation of an AI-assisted sonification algorithm on an edge device is presented and analyzed. A lightweight derived algorithm for resource-constrained implementation scenarios is also evaluated and presented, suggesting suitability for further ultra-low power, mobile and wearables implementations. Furthermore, a neural network is analysed for the potential of low-precision implementation, enabling inference on specialised hardware.Item Smart marine sensing systems for integrated multi-trophic aquaculture (IMTA)(University College Cork, 2021) Peres da Silva, Caroline; O'Flynn, Brendan; Belcastro, Marco; Horizon 2020; Science Foundation Ireland; European Regional Development FundAquaculture farming faces challenges to increase production whilst maintaining sustainability by reducing environmental impact and ensuring efficient resource usage. One solution is to use an Integrated Multi-Trophic Aquaculture (IMTA) approach, where a variety of different species are grown in the same site, taking advantage of by-products (such as waste and uneaten food) from one species as inputs (fertilizer, food, and energy) for the growth of other species. However, the remote monitoring of environmental and biological conditions is crucial to understand how the species interact with each other and with the environment, and to optimise the IMTA production and management system. Environmental monitoring of aquatic environments is already well supplied by commercial off-the-shelf sensors, but these sensors often measure only one parameter, which increases the power consumption and cost when monitoring multiple environmental variables with a fine-scale resolution. Current monitoring solutions for seaweed and kelp also include satellite and aerial sensing, which cover large areas effectively. However, these methods do not offer high-resolution, specific local data for growing sites, and are usually limited by turbidity and weather conditions. Another limitation of available commercial systems is data recovery. Most of them require that the sensor be retrieved to download data directly, increasing cost of maintenance. Radio Frequency Identification (RFID) systems that transmit in the near field (Near Field Communication – NFC) are less attenuated by the seawater environment than higher-frequency communications, and thus potentially provide a more viable alternative for underwater data transmission. In this work, we present a novel miniature low-power multi-sensor modality NFC-enabled data acquisition system to monitor a variety of farmed aquaculture species. This sensor system monitors temperature, light intensity, depth, and motion, logging the data collected internally. The sensor device can communicate with NFC-enabled readers (such as smartphones) to configure the sensors with custom sampling frequencies, communicate status, and to download data. It also has an internal machine learning enabled microcontroller, which can be used to perform data analysis internally. The device is designed to be attachable to seaweed and kelp blades or stipes. The system designed was tested in lab to characterise its sensors and to determine its battery lifetime. The sensor device was then deployed in an IMTA farm in Bertraghboy Bay, Connemara, Ireland, with the help of the Marine Institute. The data collected from the device was then correlated with environmental sensors placed in the site. Future work involves incorporating data analytics and machine learning algorithms to process data internally, allowing for lower transmission requirements.