An On-Line Unsupervised Neural Network to Adaptive Feature Extraction of Lower SNR DS-SS signals

Abstract

This paper introduces an on-line unsupervised LEArning neural network (NN) for adaptive feature extraction via Principal component analysis (LEAP) of lower signal to noise ratios (SNR) direct sequence spread spectrum (DS-SS) signals. The proposed method is based on eigen-analysis of DS-SS signals. The PN sequence and the strength of the signal can be extracted by the first and second principal eigenvectors and their associated eigenvalues of autocorrelation matrix of DS-SS signals blindly. However the eigen-analysis method is belongs to a batch method, it is difficult to real-time implementation. In this case, we can use the LEAP NN method to realize on-line adaptive principal feature extraction of the lower SNR received DS-SS signals effectively. Unlike other feature extraction methods, the estimate of the PN sequence and signal strength improves steadily with the number of code repeats. The method is applicable to arbitrary PN spreading sequence and message sequences and can theoretically operate in environments containing arbitrary levels of white background noise, and for signals with arbitrary unknown timing phase. The method requires the DS-SS signal to have a constant-modulus spreading sequence and unrelated message and code-repeat rates. This paper introduces the basic technique, and demonstrates the algorithm via computer simulation for a single DS-SS signal received in the presence of white Gaussian noise.