High Dimensional versus Low Dimensional Chaos in MPEG-7 Feature Binding for Object Classification

Abstract

We proposed [1] a feature binding method to generate a MPEG-7 compliant feature vector, defined as C-MP7. Here, we study the excellence of C-MP7 as a feature vector, using either low-or highdimensional chaos. With high-dimensional chaos-based C-MP7, we find, 1) the accuracy in SVM classifier improves 10% to 20%, for all classes of video objects over MPEG-7, 2) larger binary class separation among video objects in different classes, 3) vehicle objects are clustered well, which leads to above 99% accuracy for only vehicles against other objects in SVM, and 4) drifts in chaotic attractors allow the C-MP7 to include subtle variations in coefficients for video objects.