This is an interesting introduction to pattern recognition, emphasizing primarily the human cognition capabilities ranging from concrete object perception to abstract theory building. The title originally considered for this work was: “Form, Figure, Feature,” implying that the general concept (Form) is presented in terms of class characteristics (Feature) derived from the intuitive image (Figure) of a class, which is typical of the way the material is presented in the book. The text also includes, besides the mathematical modeling and analysis, non-mathematical, historical descriptions (Chapter 1-4). Chapter 6 (pattern recognition as entropy minimization), is especially important since the overall presentation of known pattern recognition algorithms can, according to the manuscript, be derived from the heuristic principle of minimum entropy. Chapters 7-10 deal with pattern recognition as covariance diagonalization, statistical decision making, discrimination process, and structure analysis, respectively.
Among the challenging aspects of the text are its emphasis on the philosophical, psychological, and neurophysiological bearings of pattern recognition and a repudiation of the deductive, logical, statistical, and linguistic tendencies among researchers in the field and its establishment as an inductive empirical science. Watanabe criticizes the widespread aftereffect of the logical empiricism and introduces the concept of “paradigmatic symbol” as the tool to establish a parallel between pattern analysis and new findings in brain neurophysiology.
With a few exceptions, like sample 18 of the so-called “mon,” or family crests (Figure 10.2.4, p. 401) (which I was surprised to find in a book written by Watanabe and published with the inspiration of Bea Shube of the Wiley-Interscience Division), I really liked the book.
This book is the first interesting attempt to provide a unified presentation of pattern recognition, which should interest the non-mathematical audience, as well as those specialized students of science who have been exposed to the many mathematical theories on pattern recognition.