Optimal Prediction Likelihood Tree Based Source- Channel ML Decoder for Wireless Sensor Networks

In this work, we develop a framework for optimal joint Source-Channel Maximum Likelihood (SCML) decoding in Wireless Sensor Networks (WSNs). The proposed scheme employs a novel Generalized Likelihood Ratio Test based Prediction Likelihood Tree (PLT) approach to exploit the spatio-temporal narrowband properties of the sensor data for sequence detection in wireless sensor networks. Further, analytical bounds are derived to characterize the performance of the low complexity decision feedback and optimal Viterbi based Maximum Likelihood Sequence Detection (MLSD) for joint decoding over fading wireless channels, where only ad hoc schemes exist in current literature. The PLT based SCML scheme, which has a low complexity, is ideally suited for implementation in practical wireless sensor networks with limited computational power and achieves a performance close to the optimal MLSD bound. Simulation results are presented to validate the performance of the SCML algorithm and the proposed analytical bounds for sensor data reception in WSNs.