Multi-modal feature fusion modulation recognition method and system based on neural network

Abstract

The invention provides a multi-modal feature fusion modulation recognition method and a multi-modal feature fusion modulation recognition system based on a neural network. The method comprises the following steps: converting a to-be-identified signal into a predetermined modal domain; extracting a feature vector from the corresponding modal domain signal by using a pre-trained heterogeneous neuralnetwork; fusing the feature vectors extracted by the neural network from the different modal domains of the to-be-recognized signal, and completing recognition and classification by using a pre-trained classifier. The features are learned from the to-be-recognized signal by using the strong representation learning ability of the neural network, so that a large amount of manual operation is saved.According to the invention, various modal domain information of the signal is comprehensively utilized. According to the method, the abstract feature vectors are extracted by using the heterogeneousneural network, the fused feature vectors have more comprehensive representation on the to-be-identified signals and have better robustness and robustness on influences such as noise, the obtained recognition classification result has higher reliability, and a higher recognition rate is still kept when the signal-to-noise ratio is low and the communication environment is poorer.

Description

technical field [0001] The invention relates to the technical field of signal modulation recognition, in particular to a neural network-based multimodal feature fusion modulation recognition method and system. Background technique [0002] The automatic identification technology of communication signal modulation mode is the key technology between signal detection and demodulation. In the civilian field, it is mainly used for the management of space spectrum resources, to confirm the identity of spectrum resource users, to prevent illegal use of wireless spectrum, and to ensure communication safe conduct of the activity. As the communication environment becomes more and more complex, in order to improve the utilization rate of the spectrum, different communication signals usually adopt a variety of different modulation methods. Therefore, the automatic identification of communication signal modulation methods has important application value. [0003] At present, the communi...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problem that the existing modulation recognition method based on deep neural network can only preprocess the signal in a single mode, and the recognition effect is not good when the communication environment is poor or there are many types of recognition signals, the present invention provides a method based on Multimodal feature fusion modulation recognition method and system based on neural network

Images