Use PCA Neural Network to Extract the PN Sequence in Lower SNR DS/SS Signals

Abstract

In this paper, we firstly propose an approach of discrete Karhunen-Loeve transformation to blind estimation of the PN (Pseudo Noise) sequence in lower SNR DS/SS signals. As the K-L approach is based on the decomposition of autocorrelation matrix, it has computational defects when the signal vectors became longer. In order to overcome the defects of K-L approach, we choose the PCA (Principal Components Analysis) neural networks to extract the PN sequence. Theoretical analysis and experimental results are provided to show that the approach can work well on lower SNR input DS/SS signals. The proposed method can be extended to the case of DS/CDMA (Direct Sequence Code Division Multiple Access) too.