Line tracking using multi-layer neural estimator

A line tracker based on a combination of an autoregressive model and a multilayer neural network is presented. It can be considered as an adaptive linear filter where the coefficients are updated by using the gradient backpropagation (GBP) technique. Computational examples show that the GBP method gives better results than the least mean square (LMS) algorithm for frequency resolution or line tracking. This is hardly surprising, as the LMS is only a particular case of the GBP. The results are applicable to deep space telemetry systems where line tracking in the presence of strong Doppler shifts and carrier recovery in a very low SNR environment are of primary importance