Abstract
Artificial neural networks can be used effectively to filter out noise from frequency shift keying (FSK) and phase shift keying (PSK) modulation signals corrupted with random noise. In the present paper, the time and frequency domain filtering schemes are investigated. The number of data points are optimized using a method described as selective truncation. In order to evaluate the performance of both the time and frequency domain filters, a series of tests is conducted using test signals. The training network parameters are optimized in order to speed up convergence.<>