Self-Orthogonalizing Adaptive Filter and its Application to Echo Cancellation

In the recent decades developers of telecommunication systems have adopted NLMS (Normalized Least/Mean Square), a simple but efficient adaptive algortihm, as a preferred method for noise cancellation. This article discusses an algorithm named SOBAF (Self Orthogonalizing Block Adaptive Filter), which has an advantage over the conventional NLMS in that it first "whitens" input speech sample before applying them to the basic mean-square estimator. By doing so, the adaptive process should exhibit an improved convergence rate and a lower level of residual error. In order to decrease the computational complexity of the "whitening" operation a DCT transform is utilized which, in turn, is calculated recursively.