Peripety Scientific has expertise in a range of signal and image processing domains. In particular we are interested in improving deconvolution algorithms for sparsely sampled data, improving feature detection in the low signal-to-noise regime and combining these processes into effective automated knowledge generation systems.
We have worked on several feature detection problems, primarily associated with detecting features in the low signal-to-noise regime. Working with collaborators at Victoria University of Wellington, we have successfully improved low signal-to-noise feature detection in a way previously only achievable via humans. Below is video of our CEO discussing some of recent work in the context of the Square Kilometre Array. We are currently exploring the use of this technology in a variety of commercially relevant settings.
Peripety Scientific CEO, Melanie Johnston-Hollitt, talking about 'Big Data' challenges associated with the SKA project and in particular the need for automated knowledge extraction processes, particularly those that work in the low signal-to-noise regime.
This presentation is from the Singularity University Summit in Christchurch, NZ and was published in February 2017.
Removing instrumental responses in sparely sampled data is a critical task in numerous fields including radio astronomy. The most recent work we have undertaken is looking at improving deconvolution algorithms for complex data products in the radio astronomy domain. We recently showed that the process by which such products were being treated by radio astronomers had been incorrect for the last 40 years (Pratley & Johnston-Hollitt 2016).
Pipelines for Knowledge Extraction
Combining our expertise in deconvolution algorithms and feature extraction, we are also interested in the development of knowledge extraction pipelines. Most recently we have considered the challenges of automated knowledge extraction in the big data regime of the Square Kilometre Array (SKA) project.