Distributed estimation in sparse heterogeneous networks

Advances in wireless communication and adaptive signal processing has resulted in increased interest in multidimensional sensor systems. The most popular form of multidimensional sensor system is the wireless sensor network (WSN), which is essentially a group of distributed sensors fully or partially connected wirelessly. The distributed nature of the sensor network has opened up the area of distributed estimation, which is a special case of signal or parameter estimation. In order to achieve e ective parameter estimation in scenarios where the parameter to be estimated is sparse in nature, a new sparse distributed estimation scheme has been designed in this thesis. The convergence behaviour of the proposed estima- tion approach has been further enhanced using a set of algorithms. A dynamic heterogeneous network estimation scheme has been also designed to reduce the computational overhead. An attempt has been made to design and implement a distributed speech estimation scheme in real time and promising results have been obtained.