A pattern recognition expert system for the detection and classification of narrowband and broadband signals

Abstract

A pattern recognition expert system for detecting narrowband and broadband signals in the presence of noise is presented. The system detects and classifies signals based upon the shape of the power spectrum. From the extracted shape description, decisions are made on the presence or absence of narrowband and broadband signals. The system can either detect signals from user-defined parameters or from a time sequence sample of the desired sequence. The system was tested on simulated (Gaussian shape) and real scale model data. Results from simulated data showed the successful detection of narrowband signals in environments dominated by broadband and narrowband noise. The expert system successfully detected narrowband signals down to -28 dB SNR (broadband noise dominated) and -23 dB SNR (narrowband noise dominated). Test results for real data were equally supportive. >