Dr Paul O'Grady

Currently, I am a postdoctoral researcher working with Dr Scott Rickard and his Sparse Signal Processing Group, which is a part of the Complex and Adaptive Systems Laboratory at University College Dublin.

Previously, I was a research assistant for Prof. Barak Pearlmutter at the Brain and computation lab, which is part of the Hamilton Institute at the National University of Ireland, Maynooth .

My main research interests are Compressive Sampling , Non-negative Matrix Factorisation (NMF) and Blind Source Seperation (BSS), which reside in the broader topics of Sparse Signal Recovery and Machine Learning. I have developed a number of algorithms with sparseness constraints (a constraint which enforces succinct/compressed solutions) including Sparse Convolutive NMF, which I applied to speech phoneme discovery and source separation, and Non-negative Under-determined iteratively Reweight Least Squares, which I applied to the recovery of compresively sampled non-negative data—data which is sampled below the nyquist rate. I have applied NMF to automatic guitar transcription and developed a BSS algorithm called the LOST (Line Orientation Separation Technique) algorithm, which treats speech separation as a radial line clustering problem. I am also interested in Sensory Substitution.