Non-orthogonal multiple access (NOMA) is a promising solution to the problem of spectrum deficiency. NOMA receivers require information about the modulation type of the co-scheduled user's signal to perform successive interference cancellation (SIC). Automatic modulation classification (AMC) is used to reduce the signal overhead created due to sharing of this information. In this paper, a convolutional neural network (CNN) with a modified residual block (MR-CNN) is proposed for AMC in NOMA systems. The classification performance of MR-CNN is evaluated on the input signal with four different modulation formats at varying signal to noise ratios (SNRs). Experimental results demonstrate the proposed model can achieve more than 90% classification accuracy for SNRs higher than 10dB. Further, comparative analysis depicts that the proposed model outperforms existing methods in terms of classification accuracy at low SNRs and has similar performance at high SNRs.

Cooperative spectrum sensing (CSS) has been shown as an effective method to improve the detection performance by exploiting the spatial diversity. In this paper, we propose a detection algorithm for cooperative wideband spectrum sensing considering partial band Nyquist sampling (PBNS) in which all the cooperating secondary users (CSUs) posses antenna diversity with two receive antennas (diversity branches). We make use of hard combining at the fusion center (FC) since it requires reduced control channel bandwidth when compared to the soft combining. Novelty of our approach lies in use of hard combining to sense the wide band and making use of PBNS instead of sampling the entire wide band. Complete theoretical analysis of the proposed algorithm is carried out considering Nakagami fading channels. The proposed approach outperforms the ranked channel detection used under no cooperation and the performance is better than both ranked square law combining (R-SLC) and ranked square law selection (R-SLS) algorithms.

In this paper we study the use of diversity for performance improvement in wideband spectrum sensing under fading. A new detection algorithm, namely, ranked square law selection (R-SLS) is proposed. The performance of the proposed algorithm is carried out under Nakagami fading. The analysis shows that the proposed algorithm outperforms ranked channel detection (RCD) algorithm without diversity. Analytical results are verified by Monte Carlo simulations. Effect of different parameters on the performance of proposed algorithm is also studied.