Previous Projects

Implementing Olfaction in Silicon [Completed March 2005]
With Giacomo Indiveri, I implemented portions of the Brody / Hopfield model of the mammalian olfactory bulb in hardware on an analog VLSI neuron array. Our adaptation includes STDP processes on the neuron array, which increase sensitivity and selectivity to odours under network damage.

This work resulted in a paper presented at ISCAS 2005, available under Publications.

Cheap Vision Project [Completed January 2004]
Supervised by Joaquin Sitte and Shlomo Geva, I designed an FPGA-based vision processing system. The processor implements massively-parallel feature extraction algorithms, as well as traditional sequential algorithms. The key design goal is computationally inexpensive vision. The processor can output a user-definable stream of extracted features, encoded in a flexible format. This is in contrast to traditional massively-parallel architectures such as silicon retinas, which perform processing but do not significantly reduce the required communications bandwidth. This system also eliminates the need for off-board DSP hardware as well as image stream encoding.

The Project Abstract is available from Joaquin Sitte's web site.

An unfinalised paper accepted for AMiRE 2003 is available.

Combined Visual Attention [Completed December 2003]
I assisted Jim Hogan in developing a model of visual attention combining bottom-up and top-down effects. Our bottom-up saliency wass taken from the Itti model. For top-down effects, we used fixation information obtained from a semantically annotated indoor scene database. Some semantic relationship information came from WordNet.

Processor node for the Agora architecture [Completed November 2001]
Clive Hobson and I designed a processor board to replace the transputer processor board on the QUTy mini-robot. The implementation replaced the transputer with a commercial processor (a Hitachi SH4) and a serial-link emulator housed in an FPGA.

The Project Abstract is available, and the Final Report is also available (minus the appendices).

Analyser / Playtime [Completed January 2001]
Joaquin Sitte and I developed a platform for learning computer architecture by experimenting on PDAs.

From the Analyser / Playtime Users' Guide
Analyser and PlayTime together comprise a toolkit for learning about programs and data structures in memory. The two tools run under Windows CE, which provides a portable and accessible environment on which to experiment. Students can inspect and modify all programs on the machine, including portions of the operating system. Because the Windows CE OS is stored in Read-Only Memory, it cannot be corrupted. The system is therefore quite robust, and places no limitations on tinkering.

Language Processing Group [Completed October 2001]
The LPG was a long-running project under the old Machine Learning Research Centre, led by Joachim Diederich. Among other avenues in which the whole group was involved, Michael Towsey and I investigated training Elman Simple Recurrent Networks on transcribed spoken-language data. We performed soft clustering of the context layer during training, to facilitate the extraction of Deterministic Finite State Automata with fewer states and lower prediction error rates.

We investigated using standard spherical clustering techniques, while also developing a hyper-ellipsoidal correlation measure and an algorithm to employ it. The goal was to better identify the elongated clusters we saw forming in state-space. Unfortunately the Machine Learning Research Centre ceased activities before this new technique could be fully realised.

A Technical Report was written to record the research we performed, and is available under Publications. The software I developed for this project is also available under Software.