Identification of Periodically Amplitude Modulated Targets.
Abstract
An algorithm that may be used for the classification of periodically amplitude modulated (PAM) targets is presented. The data base used to test the algorithm is derived from radar returns from vehicles moving at various velocities and aspect angles, but the techniques are applicable, as well, to other active wave devices such as sonar and laser. The received radar signal is considered to be a time series that is a function of target type, range, velocity, orientation, and noise. Classification is implemented in the frequency domain; short time spectra are computed using the Fast Fourier Transform (FFT). Features are extracted from the information-bearing sidebands of the resulting spectra. The radar signatures are classified using both linear discriminant and nearest neighbor classifiers, and performance is presented for two, three, five and six class cases using single and sequential looks. Probabilities of error of less than ten percent are achieved for five or fewer classes. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1978
- Accession Number
- ADA056516
Entities
People
- Clayton Verne Stewart
Organizations
- Air Force Institute of Technology