PUMA
Istituto di Scienza e Tecnologie dell'Informazione     
Tonazzini A., Coltelli P., Gualtieri P. Statistical analysis of microspectroscopy signals for algae classification and phylogenetic comparison. Petra Perner and Ovidio Salvetti (eds.). (Lecture Notes in Computer Science, vol. 4826). Berlin / Heidelberg: Springer, 2007.
 
 
Abstract
(English)
We performed microspectroscopic evaluation of the pigment composition of the photosynthetic compartments of algae belonging to different taxonomic divisions and higher plants. In cite{Bar07}, a supervised Gaussian bands decompositions was performed for the pigment spectra, the algae spectrum was modelled as the linear mixture, with unknown coefficients, of the pigment spectra, and a user-guided fitting algorithm was employed. The method provided a reliable discrimination among chlorophylls $a$, $b$ and $c$, phycobiliproteins and carotenoids. Comparative analysis of absorption spectra highlighted the evolutionary grouping of the algae into three main lineages in accordance with the most recent endosymbiotic theories. In this paper, we adopt an unsupervised statistical estimation approach to automatically perform both Gaussian bands decomposition of the pigments and algae fitting. In a fully Bayesian setting, we propose estimating both the algae mixture coefficients and the parameters of the pigment spectra decomposition, on the basis of the alga spectrum alone. As a priori information to stabilize this highly underdetermined problem, templates for the pigment spectra are assumed to be available, though, due to their measurements outside the protein moiety, they differ in shape from the real spectra of the pigments present in nature by unknown, slight displacements and contraction/dilatation factors. We propose a classification system subdivided into two phases. In the first, the learning phase, the parameters of the Gaussians decomposition and the shape factors are estimated. In the second phase, the classification phase, the now known real spectra of the pigments are used as a base set to fit any other spectrum of algae. The unsupervised method provided results comparable to those of the previous, supervised method.
URL: http://www.springerlink.com/content/u5x8t18266441k17/fulltext.pdf
DOI: 10.1007/978-3-540-76300-0
Subject Blind Source Separation
Bayesian estimation
Microspectroscopy signal classification
Parameter learning
Algae classification
I.2.6 Learning. Parameter learning
G.3 Probability and Statistics. Probabilistic algorithms (including MonteCarlo)
J.3 Life and Medical Sciences. Biology and Genetics


Icona documento 1) Download Document PDF


Icona documento Open access Icona documento Restricted Icona documento Private

 


Per ulteriori informazioni, contattare: Librarian http://puma.isti.cnr.it

Valid HTML 4.0 Transitional