ABSTRACT
In the signal-processing field, many discrete random processes are correlated. In other words, the processes are associated with a certain degree or they are somewhat similar. The Karhunen–Loeve transform (KLT) The Karhunen-Loeve transform is a well-established unitary optimal transform in terms of energy compaction and decorrelation. The shortcoming of the KLT is that it is signal dependent. Hence, there is no fast algorithm to implement it. Nevertheless, the KLT has been used as a benchmark to evaluate the performance of other transforms such as the discrete cosine transform. The KLT has also provided an incentive to the researchers to develop other signal independent transforms that approach the performance of KLT and have fast algorithms. The transform considers local KLTs in turn at the various terminals of the network, leading to a locally convergent algorithm; depending on the overall correlation structure this iterative algorithm can reach the global maximum.
