Approximated invariant method for pattern detection

Abstract

A system and a method for classifying input vectors into one of two classes, a target class and a non-target class, utilize iterative rejection stages to first label the input vectors that belong in the non-target class in order to identify the remaining non-labeled input vectors as belonging in the target class. The system and method may be used in a number applications, such as face detection, where the members of the two classes can be represented in a vector form. The operation of the system can be divided into an off-line (training) procedure and an on-line (actual classification) procedure. During the off-line procedure, projection vectors and their corresponding threshold values that will be used during the on-line procedure are computed using a training set of sample non-targets and sample targets. Each projection vector facilitates identification of a large portion of the sample non-targets as belonging in the non-target class for a given set of sample targets and sample non-targets. During the on-line procedure, an input vector is successively projected onto each computed projection vector and compared with a pair of corresponding threshold values to determine whether the input vector is a non-target. If the input vector is not determined to be a non-target during the successive projection and thresholding, the input vector is classified as a target.