Blind source separation and signal classification

We address the problem of classifying a digitally modulated signal received after propagation through an unknown frequency-selective channel. Channel dispersion induces intersymbol interference or fading, which must be handled before the modulation format can be classified. Spatial diversity is exploited through the use of (possibly uncalibrated) arrays, which also provides the ability to separate cochannel signals. We use a CMA-initialized alphabet-matched equalization algorithm in conjunction with a low-complexity adaptive classifier based on fourth-order cumulants. Temporal diversity and other available side information may be exploited as well. The performance of the proposed algorithms is illustrated both via theoretical analysis, as well as via extensive simulations.