Probabilistic pursuit (Q2722580)

The aim of the present paper is to propose a probabilistic extension for the paradigm of greedy pursuit algorithms used in signal decomposition analysis. For the new pursuit algorithms, the search characteristics incorporate prior knowledge of the signal model. The concept of waiting time for finding a matching dictionary element is introduced, and the idea of minimum time decomposition is developed for the classification of speech sounds. A classification scheme based on this novel approach to pursuit algorithms may be used for the design of an efficient and highly parallelizable classifier with a large spectrum of applications.NEWLINENEWLINEFor the entire collection see [Zbl 0961.00036].