Blind sequential symbol estimation of co-channel finite alphabet signals

A novel recursive equaliser is presented which is capable of removing multiuser interference and intersymbol interference due to several co-channel finite alphabet signals. The algorithm exploits shift structure present in the data model and subspace relations between the observed array data and the symbol sequences. We show how the source sequences separate naturally so that they may be estimated independently of one another and, for many symbol alphabets common in digital communications, the symbol decisions can be made simply by computing a test statistic and comparing it to possibly multiple thresholds.