Pattern recognition in noisy signal processing
Noisy signal processing methods

The purpose of this paragraph is not to present an introduction to pattern recognition — to which, incidentally, sections of this treatise ( [16] ) are devoted — but to highlight the links between the general methodology of statistical decision making (as presented in the preceding paragraphs) and the theoretical foundations of algorithms commonly used for the recognition or classification of individuals.

It will be shown that, as soon as the observations made on individuals can be considered as the realization of a random vector (or processor), pattern recognition appears as a special case of Bayesian classification. This Bayesian framework will have the advantage, through consideration of Bayes risk, of enabling a coherent approach to the difficult problem of primitive extraction, i.e. reducing the...