INVESTIGATION OF FILTER FUNCTIONS.
Abstract
Results are presented of the first phase of a basic research program in adaptive filtering. The construction of optimal filters in the Wiener-Kalman-Bucy sense requires knowledge of the statistics (second-order properties) of the underlying random processes; the filter is called adaptive when such prior knowledge is not assumed but is obtained in real time by analysis of the observed random signals. The theoretical considerations of this work are based on the following Variational Principle of Adaptation. One examines simultaneously a large number of filters each one of which is optimal for certain types of random signals. At each moment of time the output is taken from that filter which gives the smallest real-time error up to time t. In this way the unknown random process that generates the observed signal is identified indirectly. Experimental (digital computer) results are presented which confirm the logical correctness of the Principle. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1964
- Accession Number
- AD0430058
Entities
People
- R. E. Kalman
Organizations
- Glenn L. Martin Company