Limited data spectrum sensing method based on semi-supervised deep neural network

Abstract

The invention provides a limited data spectrum sensing method based on a semi-supervised deep neural network. The method comprises the following steps: S1, constructing a deep learning network; s2, pre-training the deep learning network through the labeled samples to obtain a pre-trained network; s3, performing category prediction on the unlabeled samples by using the pre-training network, and taking prediction results as labels of the corresponding unlabeled samples to obtain pseudo-label samples; s4, carrying out confidence calculation on the pseudo label sample, and correcting a cross entropy loss function of the network; s5, retraining the network by using the high-confidence pseudo label sample to obtain a spectrum sensing prediction model. According to the method, a deep learning network is pre-trained through a small number of labeled samples, a large number of unlabeled samples are marked through the pre-trained network, and then the confidence coefficient of pseudo labels is calculated by using a confidence coefficient function to improve the proportion of correct labels in expanded samples; meanwhile, the cross entropy loss function is corrected, the interference of error tags on the training model is reduced, and the performance of the final model is improved.

Description

technical field [0001] The invention belongs to the field of cognitive radio in wireless communication, and in particular relates to a limited data spectrum sensing method based on a semi-supervised deep neural network. Background technique [0002] With the rapid development of communication technology, wireless spectrum is widely used in communication systems such as broadcasting, satellite, and military. Studies have shown that the utilization rate of authorized frequency bands is 15% to 80%, while unlicensed frequency bands are increasingly in short supply. As an intelligent wireless communication technology, cognitive radio (CR) can intelligently discover available idle spectrum for users to use, thereby improving spectrum utilization. Spectrum sensing (SS) is the key technology of CR. The secondary user (SU) node detects whether there is a primary user (PU) on the frequency band of interest to determine whether there is an available idle spectrum. Therefore, improvin...

AI Technical Summary

Problems solved by technology

[0006] However, the existing spectrum sensing method based on deep learning can obtain excellent detection performance when the training data is sufficient, but the training of the network depends on a large number of labeled training samples and expanded data sets.

In order to collect these data, the SU needs to communicate frequently with the PU to determine their true state, which will increase the huge communication overhead. At the same time, when the position of the PU or SU changes, it may be necessary to obtain a new tens of thousands of training set

Images