Signal plus noise models in shape classification

摘要

In this study we show how a signal plus noise formulation can be used to improve the performance of auto-regressive shape classifiers in environments with time varying noise characteristics. Variations of this type may be due to changes in atmospheric conditions or the relative effect of discretization errors. This approach is compared to existing methodologies using a number of noisy templates in a cluttered environment. Typical applications are to aircraft identification and automatic signature validation.