Link Quality Evaluation Method For Wireless Sensor Networks Based On D-FNN

Wireless sensor network is a new computing model of integrated micro electronics, wireless communication network and distributed processing of information. Is widely used in intelligent transportation, military, medical, ecological environment monitoring and other fields. Because the radio signal will be interfered by various external factors in the process of information communication, the communication link quality. Link quality assessments of wireless sensor network routing strategy, resource management and reliable deployment is particularly important, link can increase the probability of successful transmission of a message, reduce information caused by retransmission of energy consumption, reduce the number of route reconstruction. The link quality is disturbed by many factors, the prediction of the accuracy of the link.To solve these problems, this paper carried out the following research work: put forward an evaluation method of wireless sensor network link quality based on dynamic fuzzy neural network. The Mica wireless sensor nodes access the notebook computer, using the TinyOS operating system, nesC programming language to simulate sensor nodes to send data packets. Wireless sensor network for data collection, through the comparison and analysis of wireless sensor network correlation coefficient(RSSI, CCI, PRR and SNR), select the PRR as the link quality prediction parameters. Based on the PRR value of the ETX calculation of the routing metric, in order to assess the link quality of wireless sensor.There are some shortcomings in the link evaluation method of wireless sensor based on D-FNN. This is because the whole dynamic fuzzy neural network adaptive ability decreases with the increase of the number of target samples. Because the system error can not be estimated online, the operation speed of the algorithm will be affected. Signal in the noise, but also affect the quality of the wireless sensor network link quality assessment of the acquisition. In view of the shortcomings of this algorithm in wireless sensor networks, the EKF(extended Calman filter method) is proposed to update the center and width of the premise parameters. Based on the dynamic fuzzy neural network, the neural network is modified by using the self organizing map technique to get better input space partition. Finally, in order to improve the algorithm speed, using the recursive least squares(RLS) to replace the linear least squares method(LLS) for noise elimination. The error reduction ratio(ERR) method is used as the construction strategy. Experimental results show that the optimized D-FNN algorithm can increase the link prediction accuracy and prediction quality, redundancy reduction algorithm, which in a certain extent decreases node energy consumption, prolong WSN and self organization network usage and their service life.