Low-bit Quantization Algorithm Over Wirelesssensor Network

The rapid development of electronic information,wireless communication,microprocessor and other technologies,has promoted the popularity and application of wireless sensing networks.And its related theory and application research has received widespread attention.A wireless sensor network comprises multiple sensing nodes with certain communication,storage,and data processing capabilities.These nodes and the central processor or other nodes collect and process data in a certain collaborative manner to get extraction,perception,and analysis of the environmental information surrounding the sensor network.Parameter estimation is a very important research issue in wireless sensing networks.It uses the data collected by the sensors to achieve the estimation of the relevant unknown physical quantity and has a wide range of applications in environmental monitoring,indoor positioning,target tracking,natural disaster prediction,and other fields.In general,data transmission needs to consume more energy and communication bandwidth.But in sensor networks,each node is small in size and limited in energy storage,so it is impossible to carry out high complex and high energy consumption data processing and transmission.Although some of the existing parameter estimation algorithms have good estimation performance,they do not take into account these constraints of the existence of nodes.To some extent,it increases the node load and results in a shorter working period of nodes.Therefore,how to make better use of limited resources and prolong the life cycle of wireless nodes is a challenge in the practical application of sensor networks.Compressing the redundancy of the node sampling signal and reducing the bits transmitted by the node is one of the methods to overcome this problem effectively.According to this idea,many 1-bit quantization algorithms are proposed,such as 1-bit linear regression,1-bit system identification.Most of the algorithms can reduce the node load well,but some of them still have some shortcomings in performance,such as with slow estimation speed and reduced estimation accuracy.Therefore,this paper studies the use of low bit quantization data from wireless sensor network nodes to reduce node load,prolong node working period,and ensure that the parameter estimation algorithm has better estimation performance.Based on distributed parameter estimation,this paper first studies a 1-bit quantization algorithm,which quantifies the measurement data to one bit data and considers the maximum likelihood estimation problem.Since the likelihood estimation contains the implicit variable of measurement data,the algorithm combines the EM framework to solve this problem and uses two adaptive algorithms,LMS and RLS,to update the parameters.The 1-bit quantization algorithm achieves high estimation accuracy in simulation analysis and can also reduce the node load,but has slow convergence rate.To this end,we consider a low-bit quantization improved algorithm.The algorithm increases the number of quantized bits,quantifies the measurements to different bits according to specific quantization rules,and considers the maximum likelihood estimation problem.It also uses the EM framework and correlation lemma to solve the likelihood estimation containing implicit variables,and recovers the approximate measurements from the quantization values.Finally,we obtain the low-bit quantization algorithms using LMS and RLS respectively.In this paper,a large number of simulation experiments are carried out on the low-bit algorithm.These experimental and analytical results verify the effectiveness of the algorithm and show that the algorithm can better balance the relationship between estimation performance and node loads.