Skip to content
Open access

DRASST-MTL: Attention Guided Multi-Task Network for Joint Modulation Recognition and Symbol Rate Estimation

2026 · IEEE Open Journal of Vehicular Technology · Vol 7, pp. 3041-3054 · 0 citations · 46 references

Abstract

In modern wireless communication systems, especially cognitive radio and non-cooperative communication, modulation recognition and symbol rate estimation are critical for accurate demodulation and decoding since modulation recognition is a prerequisite for subsequent signal demodulation, while symbol rate provides essential time synchronization information. Both tasks are critical for achieving reliable blind demodulation of the entire link. However, existing approaches treat them as separate tasks, leading to duplicated computation, fragmented feature learning, and degraded robustness in noisy, dynamic time-varying channels. Motivated by the inherent correlation between the two tasks, this paper proposes a novel parameter-hard sharing multi-task network, named DRASST-MTL, for joint feature learning and task optimization. The proposed network integrates depthwise separable convolution with residual connections and a dual attention (DRA) module to extract discriminative local signal features, and adopts a simplified Swin Transformer (SST) module to capture long-range global and temporal dependencies. Moreover, by embedding the convolutional block attention module (CBAM) into depthwise and pointwise convolutions and introducing the efficient channel attention (ECA) mechanism into skip connections, DRASST-MTL effectively strengthens feature representation capability without introducing extra computational overhead. Experimental results verify that the proposed method reduces the overall computational complexity by 50% and achieves superior performance compared with single-task networks. In particular, compared with the MobileNetV3L, DRASST-MTL improves the modulation recognition accuracy by 2.23% and reduces the symbol rate estimation error by 4.31%, while only consuming 51.86% FLOPs and 9.88% network parameters of the former.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.