PROBABILITY DENSITY FUNCTION ESTIMATION (WITH APPLICATIONS TO RECEIVER DESIGN FOR RECEPTION IN NON-GAUSSIAN NOISE)
Abstract
This note considers the problem of estimation of an unknown probability density function, p(x), and the derivative of the logarithm of that density function, f(x), given N samples from an ensemble whose probability density is p(x). Our principle objective is to estimate f(x) since it has been shown that the optimal receiver for known threshold signals in additive white (but possibly non-Gaussian) noise consists of a filter matched to the signal preceded by a no-memory device. Although our approach is primarily motivated toward obtaining a good estimate of f(x), the method and results would appear also to be applicable to finding 'good' estimates of p(x). The 'goodness' of the estimation procedure is investigated theoretically and experimentally.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 29, 1969
- Accession Number
- AD0694552
Entities
People
- James E Evans
Organizations
- Massachusetts Institute of Technology