Melbourne School of Engineering
Nonlinear Signal Processing Laboratory


Our research at NSP Lab focuses on the development of computational algorithms for control, signal and information processing in real-world applications. Concrete examples include algorithms for biomedical devices and for the detection and measurement of radiation. To provide efficient and effective solutions to these problems, a fundamental, systematic methodology is preferred, aimed at obtaining energy efficient algorithms with guaranteed performance.

Our strategy for developing better algorithms is twofold. We use the language of mathematics to formulate problems in novel and more natural ways, allowing the power of fundamental mathematics--such as algebraic topology, differential geometry, information geometry and stochastic calculus--to be brought to bear.

The second component of our strategy is to engage in synergistic research in the fields of systems neuroscience and systems biology. These fields provide an abundant supply of exciting problems amenable to our core systems theory skills. Conversely, by understanding better how nature processes information, we hope to develop new paradigms and architectures in the engineering community for control, signal and information processing.

The reasons we adopted the above strategy are now summarised. Linear algebra (including linear systems theory) is the traditional workhorse of signal processing. Systematic theories for nonlinear problems are scarse in signal processing. Mathematicians though, have developed vast theories for studying nonlinear problems. The knowledge gained by immersing ourselves in the mathematics literature leverages our ability to tackle nonlinear signal processing problems.

While mathematics is one source of inspiration, biology is another. The human brain is the most efficient signal processing device
known! Using only 20 watts of power, it outperforms engineered devices at myriad signal processing tasks such as source separation,
feature extraction, and speech and image recognition. The brain is inherently nonlinear. Its study is therefore synergistic with our