Exploiting multipath for blind source separation with sensor arrays

The problem of blind source separation by spatial processing is considered for convolutive mixtures of narrowband sources received by a sensor array. Multipath propagation is due to time-varying doubly-spread finite impulse response channels where the signal modes are Doppler shifted and arise from diffuse scattering rather than specular reflection. This paper introduces a new blind spatial filtering technique, referred to as the Generalized Estimation of Multipath Signals (GEMS) algorithm, which exploits multipath as the basis for signal separation under relatively mild assumptions regarding the signal spatial signatures and source waveform properties. The performance of GEMS is compared with a benchmark signal-copy procedure using real data acquired by an experimental high frequency (HF) antenna array system.1