Source Separation of Asynchronous OFDM Signals using Superimposed Training

Abstract

Abstract-In multi-user orthogonal frequency division multiplexing (OFDM) systems, an asynchronous interferer is defined as a signal whose cyclic prefix (CP) does not align with the desired user CP. Such ad-hoc multi-user OFDM systems are considered and an algorithm to perform blind space-time source separation is presented. In this paper we consider a scheme where a periodic sequence is added to the transmitted OFDM signal to be able to identify it. The periodic sequence has a cyclostationary property and can be used to estimate the channel and to separate the interfering users from the desired user signals.

Joint source separation and offset estimation for asynchronous ofdm systems using subspace fitting

“…We start by following the approach described earlier by us in [3]. The objective is to estimate the desired user data using a space-time beamformer (mNr k−τ .…”

“…Similar to [2][3] we make the following assumptions about the transmitted signals from desired and interfering users: A1: The information sequence of the desired user s…”

“…This requires P > mNr. To estimate τ we exploit the shift invariance property [3,5] that a delay in time domain corresponds to phase progression in frequency domain.…”

“…In [2] the desired user is assumed to be synchronized for a multi-user channel with memory and a beamformer is designed using the correlation between the receive antenna array and superimposed sequence. In [3] the problem is extended for OFDM systems and a multiple signal classification (MU-SIC) type search gives offset of the desired user. However this scheme performs relatively poor for high SNR's and longer channel lengths.…”

“…Superimposed training algorithms (STA) [2][3] consider a non random (not necessarily periodic) sequence being arithmetically added to the transmitted sequence. These sequences introduce a time-varying mean in the received symbols at the antenna array.…”