Stochastic Encoding based Distributed Blind Estimation for Deterministic Vector Signal

In large-scale wireless sensor networks (WSN), a large number of spatially dispersed sensors and distributed signal estimation schemes provide ubiquitous sensing. However, low-cost sensors are insufficient to support conventional distributed estimation schemes, since the channel training process causes an enormous resource consumption in the large-scale WSN. This paper proposes a distributed blind estimation scheme that consists of two components: stochastic coding and statistical inference. The stochastic coding turns the desired vector signal into statistical parameters to govern the quantized symbols. At the fusion center (FC), statistical inference based on unsupervised clustering algorithms is utilized to recover the vector signal. The channel information is not required in the proposed distributed estimation. Besides, we investigate the asymptotic properties of the estimation error. Simulation results demonstrate the effectiveness of the proposed blind estimation scheme.