An associative memory approach to blind signal recovery for SIMO/MIMO systems

SIMO (MIMO) stands for single-input-multiple-output (multiple-input-multiple-output) systems, where multiple observed output signals are generated by a single or multiple source signal(s). Our approach is based on a dynamic diversity combiner to effectively combine FIR filtered subchannel signals to recover the original signal(s). The approach is structurally resembling to that of Deterministic Maximum Likelihood, the difference being that it adapts on the combiner parameters, as opposed to subchannel parameters. While our approach implicitly adapts on the FIR recovery filters, the actual implementation is manifested in terms of associative memory models (AMMs): FASIMO/FAMIMO for SIMO/MIMO signal recovery. This work is based on three theoretical foundations: (1) finite-alphabet "exclusiveness" (FAE), (2) FIR signal recovery based on Bezout Identity, and (3) associated memory model (AMM). A "Generalized Bezout Identity" serves as the mathematical foundation for SIMO/MIMO FIR signal recoverability. The approach naturally exploits the (polynomial algebra) property of the subchannels and the "exclusiveness" property of finite-alphabet (FA) inherent in digital communication systems. Theoretical analysis on convergence property, number of attractors, and (optimal) system delays for FASIMO/FAMIMO recovery is provided. This approach exhibits several advantages over the traditional Cross Relation(CR) approaches (based on Bezout null space). For examples, the same AMM model can handle MIMO blind recovery and it significantly alleviates the burden of having to first estimate the channel lengths exactly, as required by the CR. Simulations confirming the theoretical results are provided