DKDNet: Dual Knowledge and Data-Driven Network for Cross-Domain Automatic Modulation Classification

2026-07-08

AI Summary

This work addresses the challenge of inter-domain distribution shifts caused by dynamically varying communication environments, which severely degrade the generalization capability of automatic modulation classification models. For the first time, it systematically identifies a set of signal priors—namely in-phase/quadrature (IQ), amplitude/phase (AP), and autocorrelation function (ACF)—that jointly exhibit modulation discriminability, domain stability, and complementarity. Building upon these priors, the paper proposes a dual knowledge-driven cross-domain modulation classification architecture that employs a multi-representation feature encoder to extract complementary signal views, integrates a dynamic lightweight fusion unit for adaptive feature combination, and optimizes classification and adversarial domain alignment objectives end-to-end. Extensive experiments on both simulated and publicly available datasets validate the effectiveness of the selected priors and demonstrate significant performance gains over existing unsupervised domain adaptation methods.

Abstract

The dynamics of communication environments induce significant distribution shifts across domains, challenging the generalization of deep learning-based automatic modulation classification (AMC) models. While existing UDA methods alleviate this problem by aligning source and target features, they give limited consideration to modulation-specific structures that remain informative across domain conditions. In this paper, we consider signal prior knowledge, grounded in communication protocols and physical principles, as a potential way to enhance cross-domain representation learning. Given that different priors may vary in modulation discriminability, domain stability, and complementarity, this paper first analyzes five commonly adopted signal representations that instantiate different signal priors. From them, in-phase/quadrature (IQ), amplitude--phase (AP), and autocorrelation function (ACF) are selected as compact prior-guided inputs. Based on that, a dual knowledge and data-driven network (DKDNet) is proposed for cross-domain AMC. The multi-representation feature encoder (MRFE) and dynamic lightweight fusion unit (DLFU) are designed to achieve unified representation learning and adaptive feature fusion, and the resulting fused features are optimized with modulation classification and adversarial domain alignment objectives. Experiments on both simulated and public datasets validate the rationality of the prior selection and demonstrate the superiority of the proposed method.

Problem

• cross-domain
• automatic modulation classification
• distribution shift
• generalization
• domain adaptation

Innovation

• cross-domain AMC
• signal prior knowledge
• multi-representation fusion
• adversarial domain alignment
• DKDNet