Automatic modulation classification (AMC) is a key technology of cognitive radio used in non-cooperative communication. Recently, deep learning has been applied to AMC tasks. In this paper, an AMC scheme based on deep learning is proposed, which combines random erasing and attention mechanism to achieve high classification accuracy. Firstly, we propose two data augmentation methods, random erasing at sample level and random erasing at amplitude/phase (AP) channel level. The former replaces training samples with noise information, while the latter replaces AP channel information of training samples with noise information. Erased data segments are randomly stitched to enable training data expansion. Training data of different qualities enables deep learning model to have stronger generalization capability and higher robustness. Then, we propose a single-layer Long Short-Term Memory (LSTM) model based on attention mechanism. In the first part of this model, we propose the signal embedding, which enables the input to contain modulation information more comprehensively and accurately. Then hidden state output by LSTM is input into the attention module, and weighting is applied to the hidden state to help the LSTM model capture the temporal features of modulated signals. Compared to a model without attention mechanism, this model has faster convergence speed and better classification performance. Lastly, we propose a random erasing-based test time augmentation (RE-TTA) method. Test data is randomly erased for multiple times and classification results are comprehensively evaluated, in order to further improve classification accuracy. Experimental results on dataset RML2016.10a show that classification accuracy of the proposed scheme is competitive compared with the state-of-the-art methods.

Spectrum handoff plays an important role in cognitive radio networks (CRNs). Secondary users (SUs) use spectrum handoff to hold on the idle channel or to free the channel for primary users (PUs). Spectrum handoff scheme greatly affects the transmission quality and the success rate of SUs connection. In this Letter, a reinforcement learning‐based spectrum handoff scheme with the measured packet drop rate (PDR) for multimedia transmissions over CRNs is proposed. In a system model with multiple PUs and SUs, a new state space description method is designed and an observed state includes not only the status whether PUs arrive on each channel but also several other important factors. Also, the measured PDR, instead of the calculated one, is presented to update the mean opinion score, the Q‐table and the handoff policy. Compared with the existing schemes with the calculated PDR from the Quality‐of‐Experience model, the authors' proposed scheme can converge more rapidly in the dynamic radio environment, and reduce the PDR of SUs more significantly.

To address the issue of neural network training and prediction failures due to the missing channel impulse response (CIR) data from the anchors under the non-line-of-sight (NLOS) conditions, this letter proposes an Ultra-Wideband (UWB) positioning scheme based on a new position fingerprint and a parallel convolutional neural network (CNN). Firstly, a new position fingerprint composed of multi-channel images is designed in order to avoid the subsequent CNN’s collapse caused by the missing CIR data. Secondly, a parallel CNN is proposed to predict each dimension of position coordinates respectively for an improvement in prediction accuracy. Finally, the Savitzky-Golay filtering is employed to further reduce the mean absolute error (MAE) obtained by the whole proposed scheme. Evaluation is conducted using measured public datasets. Results indicate that, this scheme ensures the model runs stably and reliably even when the CIR data from some anchors are missing. Compared to the traditional single CNN, the MAE of the proposed parallel CNN is reduced by 20.9%. Moreover, the final MAE resulted from the proposed scheme in this letter is 0.173 meters.

A tunable photonic microwave frequency up-conversion scheme using opto-electronic oscillator (OEO) based on a monolithic integrated three-section laser without any electro-optic modulator (EOM) or filter is proposed. By adjusting the monolithic integrated three-section laser to operate in the period-one (P1) dynamical state, the signal carrier frequency up-converted from the intermediate frequency (IF) to the frequency at the difference and the summation between the oscillation frequency of OEO and IF can be realized. The tunability of the up-converted frequency can also be greatly guaranteed just by altering the control currents of the monolithic integrated three-section laser induced P1 frequency. In this paper, an IF of 2 GHz with 10 Msymbol/s 16-QAM signals is successfully up-converted to 16 GHz and 20 GHz simultaneously. The up-converted frequencies can be tuned from 16 GHz/20 GHz to 19GHz/23 GHz, and the error vector magnitudes(EVMs) of all those up-converted signals satisfy the limitation in satellite communication.

A novel frequency down-converter using optoelectronic oscillator (OEO) based on monolithic integrated three-section laser is proposed. The experiment is carried out to realize the down-conversion from 17 GHz to 2 GHz with 10Msymbol/s 16QAM signal. In addition, the flexible tunability of the scheme is also experimentally demonstrated.