Document Details
Title:
Memo dated 07/09/1962 from Robert Crile to Hugh Dubberly
Author(s):
Document Type:
REPORT
Publication Date:
1962 Jul 09
Declassification Date:
1998 Aug 04
Declassification Status:
Declassified
Document Pages:
3 pages
Originating Research Org.:
Los Alamos National Laboratory, Los Alamos, NM
OpenNet Entry Date:
1999 Mar 23
OpenNet Modified Date:
1999 Mar 23
Description/Abstract:
For several reasons, Bayesian parameter estimation is superior to other methods for extracting features of a weak signal horn noise. Since it exploits prior knowledge, the analysis begins from a more advantageous starting point than other methods. Also, since "nuisance parameters" can be dropped out of the Bayesian analysis, the description of the model need not be as complete as is necessary for such methods as matched filtering. In the limit for perfectly random noise and a perfect description of the model, the signal-to-noise ratio improves as the square root of the number of samples in the data. Even with the imperfections of real-world dam Bayesian approaches this ideal limit of performance more closely than other methods. A major unsolved problem in kind mine detection is the fusion of data from multiple sensor types. Bayesian data fusion is only beginning to be explored as a solution to the problem. In single sensor processes Bayesian analysis can sense multiple parameters horn the data stream of the one sensor. It does so by computing a joint probability density function of a set of parameter values from the sensor output. However, there is no inherent requirement that the information must come from a single sensor. If multiple sensors are applied to a single process, where several different parameters are implicit in each sensor output data the joint probability density function of all the parameters of interest can be computed in exactly the same manner as the single sensor case. Thus, it is just as practical to base decisions on multiple sensor outputs as it is for single sensors. This should provide a practical way to combine the outputs of dissimilar sensors, such as ground penetrating radar and electromagnetic induction devices, producing a better detection decision than could be provided by either sensor alone.