Comparison of the Least Absolute Deviation Algorithms and Its Application in CPA

The least mean square (LMS) algorithm has been widely used to system identification, but it cannot deal with non-Gaussian noise well. In this paper, we utilize the least absolute deviation (LAD) algorithm to parameter estimation instead of the LMS algorithm, for systems with non-Gaussian interference. Three algorithms are introduced to solve the LAD criterion. It is proved that the LAD algorithm outperforms the LMS algorithm for process with non-Gaussian noise. However, LAD is not calculated as fast as least squares, which needs further improvement in the future.