Classification of Overlapping Waveforms with Pattern Recognition Techniques.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OHIO SCHOOL OF ENGINEERING
Pagination or Media Count:
An investigation of the utility of pattern recognition techniques in the classification of overlapping waveforms is made using Sporadic-Poisson signals as a model. The signals consist of repetitions of three-bit binary components which occur and also overlap in a random manner. The sporadic occurrence rate and overlap features of the signal model approximate to some extent the nature of overlapping radar returns from closely spaced targets. Elements of pattern recognition models applied include the Fast Fourier Transform, filtering in the discrete frequency domain, and the Euclidean distance metric. Classification tests are made on four types of Sporadic-Poisson signals using various filtering combinations including variance filters and with a nearest prototype classification rule. Author