Search

Deep learning (DL) technology is an effective tool for automatic modulation recognition (AMR) in the field of cognitive radio (CR). Most of the existing DL-based approaches usually design a deep network with many layers, in which only refined features from the final layer are used for AMR. However, rough features from other layers (i.e., features captured in a shallow network with fewer layers) can also provide useful information for modulation recognition. These rough features are not carefully exploited in previous approaches. In this paper, we propose a novel multi-path features fusion network for AMR, in which both refined and rough features are learned. The proposed approach identifies 11 signals including digital modulation and analog modulation generated by the GNU radio and compared to the classic network in all SNR. The experiment results show that the effectiveness of our approach. Especially, our approach is able to achieve 99.04% accuracy in +18dB SNR which outperforms all comparison approaches. [ABSTRACT FROM AUTHOR]

An energy efficient protocol for routing and scheduling in wireless body area networks is proposed by considering node energy, path loss, traffic type, and other relevant factors. First, a channel competition procedure is proposed to reduce network flooding, making use of the related information that sensor nodes collected in the initialization phase. Then, energy abundant routes are selected according to the node type and energy consumption in routing setup phase. Finally, an adaptive slot assignment method is designed to meet requirements of data rate and priority in slot assignment phase. The simulation results show that the proposed protocol not only save energy, but also improve the channel utilization. [ABSTRACT FROM AUTHOR]