An adaptive filtering algorithm for blind waveform estimation in diffuse multipath channels

The use of spatial filtering is considered to isolate a single propagation mode of an unknown narrowband signal received as a superposition of multipath components by a sensor array for the purpose of source waveform estimation. In the problem addressed, multipath propagation from source to receiver occur due to diffuse scattering from a number of distinct but extended regions of an irregular medium rather than ideal specular reflection. The combined presence of diffuse scattering and array calibration uncertainties may cause the spatial signatures of the incident distributed signal modes to deviate significantly from the plane-wave array manifold. Furthermore, the temporal signature of the waveform to be recovered may also be quite arbitrary, with no known deterministic or statistical properties that can be utilised for multipath separation. The lack of knowledge regarding the propagation channel, array manifold and source waveform poses a major challenge for the task at hand. This study proposes a new blind spatial filtering technique, referred to as the generalised estimation of multipath signals (GEMS) algorithm, which can effectively separate distributed signal modes for accurate source waveform estimation. This algorithm actually exploits the presence of multipath propagation and wavefront distortions as the basis for signal separation. The performance of GEMS is compared with two benchmark signal-copy procedures using real data from an experimental finite impulse response single-input multiple-output high frequency (HF) system.