Statistics for Pattern Recognition
Applications of image and speech processing have to deal with uncertainty and noise effects. These factors can be partially suppressed by suitable preprocessing operations. You can, for example, normalize the intensity of light, which is ordinarily different under varying illumination conditions, or the energy of speech signals. Probability theory and statistics provide a mathematical framework to handle these phenomena. As outlined in Sect. 1.4, pattern analysis deals with the mathematical part of perception. It is, therefore, natural to use all kinds of mathematical tools for solving pattern recognition problems. In addition to preprocessing, typical applications of probability theory in pattern recognition are statistical learning and pattern classification (Sect. 6.1). Lots of examples can be found in [Dud73, Nie83, Fuk90, Vap96].
Unable to display preview. Download preview PDF.