Nontraditional signal processing techniques employing linear transforms

Abstract

Nontraditional transform domain signal processing techniques, applied to two important signal processing areas are given. In the first technique, audio and video signals are represented using the superposition of nonorthogonal basis functions. It is shown that this nontraditional representation yields efficient compression algorithms, i.e, for the same bit rate, higher representation accuracy is achieved, or for the same representation accuracy, lower bit rate is required, relative to previously reported methods. In the second technique, signal autocovariance formulations in the transform domains are presented. It is shown that these formulations yield practical solutions to important signal processing classification and recognition problems. This is because these nontraditional autocovariance formulations result in considerable reduction in the storage requirements and the computational complexity, without sacrificing the signal representation accuracy associated with the traditional autocovariance formulation in the time and/or in the spatial domains.