Source Extraction in a Bandwidth Constrained Wireless Sensor Network
Hongbin Chen, Chi K. Tse and Jiuchao Feng
International Symposium on Performance Evaluation of Computer and Telecommunication Systems (SPECTS 2007)
San Diego, California (USA), July 16-18, 2007
SPECTS_Summary
In this paper, we propose a blind source extraction scheme in a bandwidth constrained wireless sensor network. We consider the extraction of multiple source signals in a sensing field. Each sensor observes a linear mixture of the source signals and their observations are corrupted by additive white Gaussian noise (AWGN). First, a sensor in the network acts as a cluster head and performs local extraction of the source signals using a fast fixed-point algorithm based on its own observation and the received noisy quantized data from other sensors. Then, the extracted signals are quantized using a pulse code modulation (PCM) method and the quantized data are sent to the sink in a multi-hop way through a fading channel. Finally, the sink performs global extraction of the source signals using a constant modulus algorithm (CMA) and PCM decoding. In order to improve the accuracy of source extraction, a multi-cluster structure is also proposed. Simulation results confirm effective extraction of the source signals by comparing with the case of ideal source extraction