Research Of ZigBee Channel Selection Mechanism Based On LMBP Neural Network Of Time And Frequency Domain

The popularity of wireless communication promoting the emergence and development of short-range wireless communication technology, as an emerging short-range wireless communication technology, ZigBee is getting unprecedented application and development with its short-range, micro-power and low-cost unique advantages. With the rapid development of such short-range wireless communication applications as ZigBee, more devices operate in the 2.4GHz band, which makes the ZigBee devices faced with more and more serious spectrum interference problem. And how to solve the problem of ZigBee devices coexistence in the 2.4GHz band, improve the electromagnetic compatibility and anti-interference capability of ZigBee in the interference environment, has become an imminent problem. On the other hand, the current ZigBee channel selection policy using the method of single energy detection, this shortcoming makes it difficult to adapt to the increasingly complex environment of 2.4GHz band, and the limitations it bring will be more and more serious, so the improvement of the channel selection mechanism of ZigBee is also imperative.In this study, we firstly introduced the technology characteristics of ZigBee and its standard protocol stack, analyzed the principle and implementation mechanism of the channel selection mechanism in the physical layer, proposed and analyzed the shortcoming of the current channel selection policy. Secondly, we collected the spectrum occupancy data of the 2.4GHz band in the field, and through quantified, normalized processing and correlation calculation to the spectrum data, we obtained the time-frequency correlation conclusion of 2.4GHz band. Thirdly, on the basis of the time-domain and frequency-domain correlation of 2.4GHz band in the field, an LMBP neural network of time and frequency domain was proposed and applied to achieve the spectrum prediction of 2.4GHz band, and the simulation result verified that the improved model had a better performance of prediction accuracy and convergence time of training. Finally, we set up the ZigBee network in the simulation software of OPNET, completed the LMBP neural network model of time and frequency domain in the background noise model of wireless communication pipeline stage, and applied the proposed spectrum prediction model in the channel selection mechanism of ZigBee. Besides in different scenarios of network's structures, we analyzed and validated that the application of the improved prediction model could optimize the result of channel selection, effectively improve the communication quality of ZigBee network, so that improve the electromagnetic compatibility and anti-interference capability of ZigBee devices in the 2.4GHz band.