Research on the Classification of Statistically and Graphically Defined Patterns.
Abstract
The first part of the report discusses extracting and classifying simply binary patterns in binary data streams which have been derived from analog signals. It is shown that, although the A/D and subsequent coding operations are inherently nonlinear, certain characteristics such as periodicities present in the original analog data are retained in the binary sequences. Superposition is not maintained through the conversion, for example. Clustering of binary representations of analog data do not usually cluster in the classification space so as to permit easy separation. The effect of quantizing noise and the inability to reconstruct the original analog data from the once quantized data steams presents additional problems. The second part of the report is concerned with automata theory and extends and generalizes the results of last year's work in this area. This year, a generalization of the binary pattern classes was considered. This generalization is based upon relaxation of the restrictions imposed upon the nature of the pattern sets. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 1972
- Accession Number
- AD0737729
Entities
People
- Kenneth W. Drake
- Lester A. Gerhardt
Organizations
- Bell Aircraft Corporation