Complexity-constrained LS estimation for sparse systems


The Sparse RLS (SRLS) algorithm, suitable for large-order sparse channels, chooses to update a subset of taps which will minimize the computational complexity of the algorithm given the constraint of maximum allowable increases in Mean Squared Error (MSE). The ability to predict the increase in MSE due to decision to neglect some of the taps is essential for its implementation. To illustrate the principles behind the algorithm, an example is presented by way of an experimental transmission of known data through an unknown shallow-water acoustic channel.