Measurement: Sensors (Feb 2023)
Hybrid blockchain-based spectrum sharing algorithm for dynamic channel selection in cognitive radio
Abstract
Currently, the available CR spectrum is not being used to their full potential. Certainly, the CR system is capable of meeting this challenge. Whenever a licensed user has to start another transmission on that channel, spectrum handoff allows an unlicensed user to leave its current channel. The SU then switches to a different channel to finish the incomplete transmission. This technique uses the handshaking protocol to transmit and receive an acknowledgment before data transmission to obtain the user's wait time.This paper describes the Convolution Neural Network automatically extracts the features without any human supervision. For classification, a two-layer hidden neural-network is utilized once features are extracted. The CNN performs convolution operation to extract the complex features of the data that is significant for doing regression. First, the preprocessed data is delivered to the CNN's input layer. The CNN conducts a convolution operation on the data to extract the complex features that are important for regression. . The developed model consists of a sequence layer, Long Short Term Memory (LSTM) layer, fully connected layer, and a regression layer. LSTM memory cells were used in the LSTM layer to provide extended memory support. LSTM units save the essential past state information to increase performance by accounting for dependencies, and they erase the unnecessary information to save memory an [1]d energy. The fully linked layer collects the processed output from all of the LSTM hidden units. The regression layer receives a single output as a result of this. The regression layer computes an output for which the loss value is computed and propagated backwards during the training phase. This is done for all of the training iterations until the loss is zero. The anticipated values of waiting time are next tested using a fully trained regression model. The observed waiting time as a function of the number of rounds played. The loss and RMSE (Root mean square error) keep decreasing for increasing number of iterations of training.
