The following papers form part of my PhD student Violet Johnson's dissertation.
V. Johnson and
I. Parberry,
"Music upscaling using convolutional neural networks",
In Preparation, 2020.
Abstract
Audio upscaling with generative neural networks has been studied in the fields of super-resolution and speech bandwidth expansion. Previous approaches have worked well for speech, but not for music. We propose a DNN approach with a novel dilated and residual architecture for this domain and an additional refinement method which outperforms other methods when upscaling music according to the log RMS spectral distance error metric.
V. Johnson and
I. Parberry,
"Towards steganography detection in time-series data",
In Preparation, 2020.
Abstract
We examine a recently developed system for censorship and surveillance resistant communication within a censored region which uses online games as a cover. This system works by embedding information directly into the game network traffic, and demonstrates resistance to current detection methods. We propose a novel approach to detect the use of such a system by comparing attributes of the prediction accuracy of a sequence-predictive convolutional neural network trained on altered and unaltered signals. Our method is able to to reliably detect the presence of alteration of data within a reasonably small window of a multi-featured signal from a parametric signal generator.