Semi-blind Data-Selective and Multiple Threshold Volterra Adaptive Filtering

This work proposes the use of data-selective semi-blind schemes in order to decrease the amount of data used to train the adaptive filters that employ Volterra series, while reducing its computational complexity. It is also proposed a data-selective technique that exploits the structure of Volterra series, employing a different filter for each of its kernels. The parameter vector of these filters grows as the order of the kernel increases. Therefore, by assigning larger error thresholds to higher-order filters, it is possible to decrease their update rates, thus reducing the overall computational complexity. Results in an equalization setup indicate that both proposals are capable of achieving promising results in terms of mean square error and bit error rate at low computational complexity, and in the case of semi-blind algorithms, using much less training data