A Neural Network Method to Adaptive Feature Extraction of Weak DS-CDMA Signals

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 code-division multiple-access (DS-CDMA) signals. The proposed method is based on eigen-analysis of DS-CDMA signals, and exploits the cyclo-stationarity of communication signals adequately. The received signal is firstly sampled and divided into non-overlapping signal vectors according to a temporal window, which duration is a period of pseudo noise (PN) sequence (signature waveform). Then an autocorrelation matrix is computed and accumulated by these signal vectors one by one. Since the synchronous point between the symbol waveform and observation window is a randomize determination point, therefore, each vector must contain all information of a whole period of PN sequence. In the end, the PN sequence and its strength can be extracted by the principal eigenvectors and their associated eigenvalues of autocorrelation matrix blindly. But, the eigen-analysis method is belongs to a batch method, it is difficult to real-time implementation. We can use the LEAP NN method to realize on-line adaptive principal feature extraction and tracking of the weak input DS-CDMA signals effectively.