Abandon Locality: Frame-Wise Embedding Aided Transformer for Automatic Modulation Recognition

Yantao Chen

doi:10.1109/LCOMM.2022.3213523

Abandon Locality: Frame-Wise Embedding Aided Transformer for Automatic Modulation Recognition

2023-01-01

25

PRE

AI

DOI：10.1109/LCOMM.2022.3213523

Original

Original request for help

Save

Abstract

En 中文

Automatic modulation recognition (AMR) has been considered as an efficient technique for non-cooperative communication and intelligent communication. In this work, we propose a modified transformer-based method for AMR, called frame-wise embedding aided transformer (FEA-T), aiming to extract the global correlation feature of the signal to obtain higher classification accuracy as well as lower time cost. To enhance the global modeling capability of the transformer, we design a frame-wise embedding module (FEM) to aggregate more samples into a token in the embedding stage to generate a more efficient token sequence. We also present the optimal frame length by analyzing the representation ability of each transformer layer for a better trade-off between the speed and the performance. Moreover, we design a novel dual-branch gate linear unit (DB-GLU) scheme for the feed-forward network of the transformer to reduce the model size and enhance the performance. Experimental results on RadioML2018.01A datasets demonstrate that the proposed method outperforms state-of-the-art works in terms of recognition accuracy and running speed.

Keywords:

Transformers

Feature extraction

Task analysis

Modulation

Finite element analysis

Symbols

Costs

Automatic modulation recognition

deep learning

transformer

Abandon Locality: Frame-Wise Embedding Aided Transformer for Automatic Modulation Recognition

Abstract

Journal

Organization