Institute of Bio-Sensing Technology

IBST is an initiative of UWE, Bristol

Latest News

» All News

Signal processing - data analysis

Signal_processing_example High dimensional data analysis methods are used to interpret data generated by innovative techniques developed by other researchers. Data in its raw form produced by bio-sensing applications is often too complex for traditional statistical analysis techniques and so customised data processing techniques are developed.

This has the effect of turning the data generated into useful information. In particular these techniques can be used to classify particular samples for diagnosis purposes. The techniques have been applied to data from the biomedical research community and from food research and have the potential to be applied to high dimensional, highly correlated data sets from other sources.

Examples of projects

These methods have been used to classify food samples into pure / adulterated using Near Infra-Red Spectroscopy; are being used in conjunction with the OdoReader (in development with Norman Ratcliffe) to diagnose the cause of diarrhoea in stool samples using the vapours emitted and are also being used with spectroscopic data to identify tumour cells with a team based in T.U. Dresden

Image above is an example of a classification challenge separating Near Infra-Red (NIR) Spectra of pure extra virgin olive oils (in red) from samples adulterated with sunflower oil (in black) - 100% correct classification (out of sample) was achieved on this data.

Find out more

Lead researcher
Dr Deirdre Toher, University of the West of England

For more information about signal processing-data analysis, please contact Deirdre Toher.