Performance Comparison of Data-Aided Interference Suppression Methods in Fractional Fourier Transform Domains

Abstract

Interference is a big problem with the proliferation of communications systems worldwide. Conventional interference suppression is based on time domain methods that give the minimum mean-square error (MMSE) between a true signal and its estimate, or frequency domain techniques using fast Fourier Transforms (FFTs). When the signal-of-interest (SOI) overlaps the interferer, separation is difficult. The Fractional Fourier Transform (FrFT) overcomes these limitations by using the entire time-frequency plane and finding the best rotational parameter ‘a’ in which the SOI and interferer are most separable. We choose ‘a’ using a short training sequence, and then filter out the interference along the new time axis ‘ $t_{a}$ ’. In this paper, we describe and compare both conventional time and frequency domain algorithms and new algorithms based upon the FrFT. We show that the repeated reduced rank FrFT algorithm outperforms all of the other methods by at least one or two orders of magnitude reduction in mean-square error (MSE) and bit error rate (BER), typically with just $L\leq 3$ repetitions. Reduced rank filtering also overcomes limitations of MMSE techniques, which require many samples in non-stationary environments and are computationally expensive.