With the rapid development of global communication technology, the problem of scarce spectrum resources has become increasingly prominent. In order to alleviate the problem of frequency use, rationally use limited spectrum resources and improve frequency utilization, spectrum prediction technology has emerged. Through the effective prediction of spectrum usage, the number of subsequent spectrum sensing processes can be slowed down, and the accuracy of spectrum decisions can be increased to improve the response speed of the whole cognitive radio technology. The rise of deep learning has brought changes to traditional spectrum predicting algorithms. This paper proposes a spectrum predicting method called Back Propagation-Long short-term memory Time Forecasting (B-LTF) by using Back Propagation-Long Short-term Memory (BP-LSTM) network model. According to the historical spectrum data, the future spectrum trend and the channel state of the future time node are predicted. The purpose of our research is to achieve dynamic spectrum access by improving the accuracy of spectrum prediction and better assisting cognitive radio technology. By comparing with BP, LSTM and Gate Recurrent Unit (GRU) network models, we clarify that the improved model of recurrent time network can deal with time series more effectively. The simulation results show that the proposed model has better prediction performance, and the change in time series length has a significant impact on the prediction accuracy of the deep learning model.

As the bandwidth increases, the high-speed sampling rate becomes the bottleneck for the development of wideband spectrum sensing. Wideband spectrum sensing with sub-Nyquist sampling attracts more attention and modulated wideband converter (MWC) is an attractive sub-Nyquist sampling system. For the purpose of breaking the system structure limit, an advanced sub-Nyquist sampling framework is proposed to simplify the MWC system structure, adopting the single sampling channel structure with a frequency shifting module to acquire the sub-Nyquist sampling values. In order to recover the signal support information, the sensing matrix must be built according to the only one mixing function. Most existing support recovery methods rely on some prior knowledge about the spectrum sparsity, which is difficult to acquire in practical electromagnetic environment. To address this problem, we propose an adaptive residual energy detection algorithm (ARED), which bypasses the need for the above-mentioned prior knowledge. Simulation results show that, without requiring the aforementioned prior knowledge, the ARED algorithm based on the advanced sub-Nyquist sampling framework has the similar performance as MWC and even higher than MWC in some cases using only one sampling channel.