Decentralized Automatic Modulation Classification Method Based on Lightweight Neural Network

Dong, Biao; Xu, Guozhen; Fu, Xue; Dong, Heng; Gui, Guan; Gacanin, Haris; Adachi, Fumiyuki

doi:10.1109/pimrc54779.2022.9978060

Cited by 2 publications

(1 citation statement)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Luan et al [22] proposed a lightweight structure based on unordered units and GRU, which reduces the potential risk of overfitting and solves the timeconsuming problem. Fu et al [23] proposed a decentralized learning recognition method, using model merging and lightweight design to reduce communication overhead. Based on the tired positivity feature, Teng et al [24] proposed a classification method with a channel compensation mechanism.…”

Section: A Related Workmentioning

confidence: 99%

Lightweight signal recognition based on hybrid model in wireless networks

Gao,

Guo,

Tang

2024

Preprint

View full text Add to dashboard Cite

As a key technology in wireless networks, signal recognition is widely used in various military and civilian fields. By correctly recognizing the modulation scheme of the received unknown signal, the performance of the communication system can be improved. With the comprehensive digitization and intelligence of the world, the rapid development of wireless communication puts forward higher requirements for signal recognition: 1) Accurate and efficient recognition of various modulation modes and 2) Lightweight recognition of intelligent hardware. Therefore, through in-depth study of hybrid model and lightweight modulation recognition, this paper first designs a hybrid signal recognition model based on convolutional neural network and gating recursive unit (CnGr). By combining the spatial module with the temporal module, the multi-dimensional extraction of the original signal is promoted and the recognition accuracy is effectively improved. Further, a lightweight signal recognition method is given by combining pruning and depthwise separable convolution. The network can be reduced effectively on the premise of ensuring the recognition accuracy, which facilitates the deployment and implementation on edge devices. Extensive experiments demonstrate that the proposed method can effectively improve the recognition accuracy, and reduce the model significantly without reducing the accuracy.

show abstract

Section: A Related Workmentioning

confidence: 99%