Decision fusion rules based on multi-bit knowledge of local sensors in wireless sensor networks

For Wireless Sensor Networks (WSNs) with a small quantity of sensors and very low SNR, distributed detection and decision fusion rules based on multi-bit knowledge of local sensors are proposed. At local sensors, observations are quantized to multi-bit local decisions. Three quantification algorithms are investigated, which are based on weight, statistics and redundancy, respectively. Corresponding suboptimal fusion rules at the fusion center are also discussed by approximating the optimal likelihood ratio test. System level detection performance measures, namely probabilities of detection and false alarm, are derived analytically by employing probability theory. Finally, Monte Carlo methods are employed to study the performance of proposed decision fusion rules with parameters such as Rayleigh fading channel and Gaussian noise. Numerical results show that, under non-ideal channel, commonly used schemes based on weight cannot improve the system performance even with a large number and high SNR. Fortunately, schemes based on statistics and redundancy can enhance the system capability when the node is deficient and SNR is low. Furthermore, schemes based on statistics have the best stability among the three schemes, and schemes based on redundancy have the best performance among the three when quantization degree is high.