This document discusses sign recognition using principal component analysis (PCA). It begins by describing the image acquisition process using a web camera. It then explains that PCA is used to project sign images into a feature space defined by the variation among known sign images. The eigenvectors resulting from the PCA calculation represent the eigenvectors of the set of sign boards. It outlines the steps in sign recognition which include initializing the training set, calculating eigenvalues to define the eigenspace, classifying new signs based on their weights, and incorporating repeatedly recognized unknown signs into the known class.