Design and Implementation of Intelligent Dynamic Spectrum Management DSM Techniques to Mitigate Spectrum Scarcity Issues in 5G Network

Abstract

Massive expansions of mobile applications have increased the demand of bandwidth. The insufficient spectrum resource in 3G and 4G network affects the growth of wireless communications. This spectrum scarcity is a prominent drawback of 3G and 4G network, which encouraged the innovation of 5G technologies. To mitigate this spectrum scarcity issue, efficient spectrum sensing and maximum utilization of spectrum resources is essential. Primary user (PU) transmission is unpredictable and during some time slots, frequency band or geographic location PU remains inactive which is known as spectrum hole. Spectrum sensing is performed to indentify such unoccupied or vacant spectrum where PU is inactive. In wireless communication CR technique has a great impact in enhancing the optimum spectrum utilization. Probability of false alarm (PFA) and probability of detection (PD) are the two important metrics that are applied to estimate the performance of sensing. A desirable model should achieve high PFA along with low PD. Implementing deep learning-based approaches have increased accuracy in spectrum detection. newlineWe have explored several spectrum sensing strategies. It is observed that, in low SNR condition accurate sensing is little tricky. According to IEEE 802.22 standard, a good sensing strategy should achieve high PD while the PFA should be between 0 to 0.1. In our work we have proposed deep learning-based approaches and designed some hybrid models like ResNet-LSTM, DeepSenseNet and DeepMLLHSNet (Deep Modified LeNet-LSTM Hybrid Sensing Network) for efficient spectrum sensing. Hence, the present research work was undertaken to address the problem of spectrum sensing in low SNR and improves accuracy. This research penetrates into the use of deep neural network (DNN) for sensing the vacant spectrum accurately. For our simulation RadioML2016.10b dataset is used. The results are also studied. The proposed approach shows betterment in sensing than other traditional as well as previously proposed techniques. We have mainly focused