Adaptive Neuro-Fuzzy Inference Systems for Wideband Signal Recovery in a Noise-Limited Environment

Abstract

A neuro-fuzzy detector in the continuous wavelet transform (CWT) domain is developed to enhance the performance of wideband acoustic signal detection in a noise-limited environment. The aim of the detector is to determine the motion parameters (radial range and velocity) of moving targets in active wideband sonar echolocation system at very low signal-to-noise ratio (SNR). The detection is based oh time-scale and time-delay of the received echo. The fuzzy detector is composed of two parts: noise reduction based on the adaptive noise cancelling (ANC) concept, and motion parameters estimation based on the correlation process. Using learning intelligent systems named adaptive neuro-fuzzy inference systems (ANFIS), noise embedded in the return signal is minimized which improves the output SNR. The resultant signal is then proceeded by a similarity measurement technique known as the wideband cross correlation process equivalent to the CWT operation for determining the motion parameters. Simulation results demonstrate that the neuro-fuzzy detector is effective in accurately predicting the motion parameters with less than 0.2% false target detection rate.