Medical signals and instrumentation

Signal and Information Processing for Sensing Systems

Dr. Marco, Santiago
Group Leader


Torre I - Planta 9 | Baldiri Reixac, 4 | 08028 | Barcelona
Email : smarcoibecbarcelona.eu

Research Topics

Biologically inspired signal and data processing / Development of smart chemical instruments / Statistical signal processing and chemometrics / Image processing / Chemical sensor arrays / Miniaturised infrared spectrometers / Ion mobility spectrometry / Odour perception


Artificial olfaction (AO) systems are intelligent chemical instruments for the detection of volatile compounds and smells. These systems usually combine an array of nonspecific chemical sensors with a pattern recognition system. The emphasis is not on the identification and quantification of the individual components – as is the case with analytical instruments – but rather on the overall evaluation of the odour. Moreover, AO systems tend to favour miniaturised devices capable of analyzing an odour in seconds. The focus of our research in this field is the development of signal and data processing systems inspired by the neuronal processing of the biological olfactory pathway.

 

 

Neuromorphic Odor robot including chemical sensor arrays and embedded computational models of the insect olfactory system developed in the NEUROCHEM project (joint development with UPC and UPF)

 

 

 

 

 

Our research in 2012 included the following:

Analysis of olfactory bulb activity maps:

  • We have done a cluster analysis of the activity of the olfactory bulb in response to a large set of odorants. Clustering results show only a minor number of stable clusters that subdivide hierarchically in finer cluster with lower stability.
  • From glomerular activity maps in the olfactory bulb, we have studied the distribution of receptive ranges. The olfactory bulb displays a large diversity of receptive ranges, from very selective to broadly tuned receptors. The study using information theory tools shows that sets of broadly selective sensors with low correlation values are the optimum setup for chemical coding.

Ion Mobility Spectrometry:

  • We have proposed a blind source separation technique (NMF) for preprocessing non-linear ion mobility spectra and after building quantitative multivariate models.
  • We have explored the detection of TCA in wine (cork taint) and biogenic amines with ion mobility spectrometers and we have determined the limits of detection that can be achieved without any preconcentration.

Genomic signals:

  • We have proposed a detector of transcription factor binding sites in genomic sequences based on numerical coding of DNA followed by multivariate statistics. The method improves on PSSM methods and equals methods that consider interdependences with a much lower computational cost.

 

Ultraviolet-Ion Mobility Spectrometer from GAS, Germany