Outlier Resistant Filtering and Smoothing.
Interim rept. 1 Jul 85-30 Jun 86,
CONNECTICUT UNIV STORRS DEPT OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE
Pagination or Media Count:
Consider a stationary Gaussian information process transmitted through an additive noise channel. Assume that the noise and information processes are mutually independent, and model the noise process as nominally Gaussian with additive independent outliers. For the above system model, one first develop a theory for outlier resistant filtering and smoothing operations. Then design specific such nonlinear operations, and study their performance. The performance criteria are asymptotic mean squared error at the Gaussian model, the breakdown point, and the influence function. The operations combine excellent at the nominal model performance, with strong resistance to outliers. Keywords Filtering Smoothing Qualitative robustness Outlier Resistance.
- Statistics and Probability