A Computer Decision-Making Process for the Elimination of Noise from Data.
Abstract
Data acquired in many scientific and engineering activities are contaminated by noise or extraneous readings that are superimposed on base-line values. Automated data-analysis routines normally resort to some form of numerical averaging to suppress noise with the assumption that the smoothed values will closely approximate the base data. There are, however, circumstances where averaging may not produce acceptable results such as in situations of serve noise that are biased in magnitude and polarity. The Air Force Geophysics Laboratory was faced with this problem in the analysis of snow weight/rate data because of wind acceptable error boundaries. It was then noted that the base values could be very closely replicated by a hand-drawn, best guess line on the plotted, noisy-raw data. This revelation prompted the development of a computer process that, by mimicking the logic of human reasoning, can eliminate extraneous readings and reconstruct the base-line data to a vert close approximation. As such, this computer decision-making procedure may be classified as a form of artificial intelligence that may be applicable to other analytical routines. This report discusses the problems associated with extraneous or noisy data and describes the technique that was developed to eliminate superfluous readings from snow weight/rate measurements.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 27, 1987
- Accession Number
- ADA185746
Entities
People
- Robert O. Berthel
Organizations
- Air Force Research Laboratory