TAP-SNet: Lightweight Few-Shot Modulation Recognition via Selective Mamba Mechanism

Abstract

Automatic Modulation Recognition (AMR) is a cornerstone of cognitive radio; however its practical deployment is often hindered by the scarcity of labeled data. While Few-Shot Learning (FSL) offers a promising solution, existing models are typically plagued by excessive computational overhead and large parameter volumes. To tackle these limitations, this letter develops a lightweight FSL framework based on Selective Structured State-Space Models (SSMs) and the Mamba architecture. By replacing traditional recurrent units with efficient Mamba blocks, our design leverages an input-dependent selection mechanism to dynamically model continuous sequence characteristics, while significantly boosting parallel efficiency during training and inference. Experimental results on the RML2018.01A and HisarMod2019.1 datasets demonstrate that TAP-SNet achieves strong recognition performance across both datasets, with its clearest advantage on RML2018.01A, while substantially reducing model complexity. These findings validate the effectiveness of our approach in resource-constrained environments.