Page 184 - Introduction to Statistical Pattern Recognition

166 Introduction to Statistical Pattern Recognition

When a quadratic classifier is designed, Properties (4) and (7) are particularly
useful.

Example 6: For Data RADAR, n = 33 and m = 2. Both R,. and R,.
matrices of (4.134) and (4.135) are formed based on the assumption that the
toeplitz form of (4.126) holds. Then, the correlation matrix for mi is

Rj = [ RCi pT)Rcl

pT)RCj R,.; and RCi = . (4.143)

The inverse matrix and determinant of R, are

(4.144)

(4.145)

and

(4.146)

In order to verify the effectiveness of the above approximation, the fol-
lowing two experiments were conducted.