Physics-aware AI enables robust full-color acoustic vortex communication
Acoustic orbital-angular-momentum (OAM) modes provide an orthogonal basis for multiplexed communication, yet existing acoustic OAM links mainly use discrete topological-charge (TC) selection, limiting the information space to resolvable modes and preventing direct full-color image encoding, which requires both color-channel identity and continuous pixel intensity. Here, we demonstrate a physics-aware AI framework for robust full-color acoustic OAM communication. The encoder maps RGB information onto a hybrid TC–amplitude representation: TC labels the color channel, while independently tunable modal amplitude carries pixel intensity. This scheme extends acoustic OAM communication from discrete symbol assignment to full-color image encoding without additional OAM modes. To recover this denser representation under source imperfections, inter-modal crosstalk, and dynamic noise, we develop a physics-embedded dual-branch neural decoder that reconstructs modal amplitudes with mean-squared-error loss and phases with a 2 π -periodic cosine loss respecting phase topology. The framework maintains Pearson correlation coefficients above 0.99 under source nonidealities, 0.91 under hybrid static noise down to S N R = − 5 ~dB, and 0.94 under dynamic non-stationary interference, reconstructing complex full-color images without task-specific retraining. These results establish TC–amplitude encoding and physics-constrained AI decoding as a practical route to higher-capacity, robust acoustic OAM communication.
Authors
- Nengyin Wang
- Yong Li (ORCID: https://orcid.org/0000-0001-8049-9128)
- Quansen Wang
- Shanshan Liu
- Xu Wang
- Hua Ding
Publication Details
- Journal
- APS Open Science
- Published
- 2026-11-16
- DOI
- https://doi.org/10.1103/mzz4-szf3
- Primary Topic
- Acoustic Wave Phenomena Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00