A new neural network based multiuser detector in impulse noise

In most practical direct-sequence spread spectrum multiple access (DS/SSMA) communications, we are confronted with the demodulation of signals corrupted by both multiple-access interference and impulsive noise. A new symbol by symbol multiuser detector, in the form of recurrent correlation neural network, to jointly suppress the multiple-access interference and the impulsive noise for a synchronous code-division multiple-access (CDMA) system is presented. In this detector, a sgn() function is embedded in the conventional steepest descent method to work against the impulse noise. Computer simulations illustrate that the detector has good performance against multiple-access interference and impulse noise. Comparison with other detectors is also given.