Heterogeneous 2D/1D Signal Representation Fusion for Underwater Acoustic Modulation Recognition Under Distribution Shift
arXiv:2606.23702v1 Announce Type: cross Abstract: Modulation recognition systems rely on heterogeneous signal representations. 2D signal-image modalities such as time-frequency and cyclostationary maps capture structural patterns, while 1D statistical descriptors such as higher-order power spectra encode complementary cues. Under distribution shift, these modalities degrade unevenly, making robust fusion a central challenge for practical deployment. Progress is further limited by the lack of a unified evaluation protocol that systematically separates different shift types. This paper addresses
This research addresses a fundamental challenge in deploying AI/ML systems in real-world, dynamic environments where data distribution shifts are common and degrade performance.
Improved robust signal recognition under distribution shift is crucial for reliable AI operation in critical applications, particularly in domains like defence or autonomous systems.
The proposed research aims to provide a more unified evaluation protocol and improved fusion techniques, which could lead to more resilient AI systems for signal processing.
- · Defence contractors
- · Autonomous vehicle developers
- · Underwater sonar manufacturers
- · AI/ML research institutions
- · Developers of brittle AI/ML systems
- · Organizations relying on static models
Enhances the reliability of AI systems operating in dynamic or adversarial underwater environments.
Accelerates the deployment of AI-driven capabilities in marine surveillance, exploration, and military applications.
Potentially reduces the human reliance on manual signal analysis in complex aquatic scenarios, shifting labor towards oversight of autonomous systems.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI