Voice Morphing

Nonlinear Speech

Low-order Models

Continuous/Discrete Dynamics

Baroclinic Flows

Voice morphing

Voice morphing technology with speaker recognition enables a user to transform one person's speech pattern into another person's pattern with distinct characteristics, giving it a new identity, while preserving the original content. This is an area of numerous applications and for this reason it has been the subject of considerable research effort.Some examples of ongoing projects that will benefit from successful voice conversion techniques are:

Nonlinear dynamics of speech

This project is concerned centrally with the problem of voice modelling for technological applications. It investigates the use of state-of-the-art, nonlinear biophysical models of voice production for media technology where the computational capacity of the hardware is severely limited.

Links

Low-order models of the Martian atmosphere

The aim of this project is to generate reduced-order models suitable to study the long-term behaviour of the Martian atmospheric dynamics. At the moment we are investigating the possibilities of Proper Orthogonal Decomposition and related techniques.
The Martian atmosphere is an interesting case since despite having similar structure to the Earth (thin atmosphere surrounding a solid surface) the lack of oceans, the presence of dust and the large influence that the huge topographic features exert on the atmosphere make it different enough to encourage the search of new modelling methods.

Links

Comparing continuous and discrete nonlinear dynamics

This project deals with the nonlinear dynamics of electronic circuits focusing on the Moore-Spiegel equations. The study has both an experimental and analytical component. Huge data sets are generated at certain parameter values and then subjected to analysis by different tools.

Links

Baroclinic flows

This project builds on the results of two existing DPhil studies into the use of empirically determined basis functions, calculated from time series of the velocity fields to generate low order nonlinear models of more complicated baroclinic flows. These include numerical models of the Martian atmosphere in which topographic features are included, as well as laboratory models of more complicated baroclinic flows.

Predicting nonlinear systems with uncertainties: The data assimiation method

Data assimilation encompasses the techniques which utilize the information content from both the observations of the system states and the prior information (also called the background) to estimate the system states. This is an area with important applications to meteorology, geophysics, oceanography and so on. This project aims to study advanced data assimilation techniques like four dimensional variational data assimilation (4D-Var), ensemble Kalman filters (EnKFs), and investigate the factors that may affect their performances.