Fast Parallel Processing for Neurological Signals

This project will develop methods to analyse complex multivariate time-delay relationships between neurological signals recorded from human brains. The analysis will extend Empirical Mode Decomposition (EMD) which, although effective, suffers from limitations rendering it impractical for the online analysis of the interactions of brain signals. These drawbacks will be addressed by building a mathematical framework of EMD which will enable to understand the behaviour of this method better and will also lead to more robust methods of extracting relationships from multivariate brain signals. The developed framework will enable reformulation of the algorithm in a way that lends itself to fast and efficient implementation on affordable parallel computing architectures such as Graphics Processor Units (GPUs) leading to a fast parallel EMD analysis opening up this signal processing technology to new applications such as Brain Computer Interfaces.