Abstract:
"
An organism’s ability to move around in its environment depends on the capacity of its brain to
process visual stimuli. To understand this important process better, neuroscientists use Visually
Evoked Potentials, which are the responses a brain creates based on visual information. However,
researchers face a major bottleneck since there are millions of data points extracted from their
experiments, and only a relatively low number of those data points are Visually Evoked Potentials.
Identifying these specific potentials manually is a time-consuming process which requires an expert
in the domain. This project aims to partially automate the process so that research is performed
more efficiently by analysing and implementing algorithms which can be used to smooth distorted,
noisy data along with algorithms that do basic pattern matching. Several signal processing algo rithms were reviewed and novel signal processing algorithms have been created to solve the problems
faced. The key argument from this project showed that there was little possibility of identifying
each VEP response accurately, and the best option providable was to point out all the possible VEP
responses. The system has a minimum accuracy of 76.5% with the accuracy increasing depending
on the number of VEPs, with a maximum accuracy of 88.2%. The final result is a system that takes
in a directory of raw Electroencephalogram data and visually represents all possible VEPs to the
researchers by analysing multiple data streams asynchronously and comparing them.
"
