Multiple plant identifier via adaptive LMS convex combination

The least mean square (LMS) algorithm has become a very popular algorithm for adaptive filtering due to its robustness and simplicity. A difficulty concerning LMS filters is their inherent compromise between tracking capabilities and precision, that is imposed by the selection of a fixed value for the adaption step. An adaptive convex combination of one fast LMS filter (high adaption step) and one slow LMS filter (low adaption step) was proposed as a way to break this balance. We propose to generalize this idea, combining multiple LMS filters with different adaption steps. Additional speeding up procedures are necessary to improve the performance of the basic scheme. Some simulation work has been carried out to show the appropriateness of this approach when identifying plants that vary at different rates.