AI company DeepSig, which focuses on the use of AI in wireless communications, has started up a new repository for models that apply AI to radio systems.
The OmniSIG Model Hub is meant for models that fuel the DeepSig’s OmniSIG Engine (formerly known as the OmniSIG Sensor) and aims to be a location that operators, developers and researchers can use to store, manage and retrieve pre-trained models as well as customized or proprietary models. DeepSig’s OmniSIG Engine detects and identifies RF signals and enables near-real-time, “spectrum-aware” reporting of anomalies, changes and threats in RF systems, according to the company.
The hub already includes “numerous models” that DeepSig has developed, tested and validated, and it allows access to OmniSIG capabilities beyond the default model, the company said—including models for signals such as push-to-talk radios or other communications devices, and commercial drone signals. The hub allows users to search for models using criteria such as the signal type or frequency range and offers a secure location for management and sharing or both pre-trained or custom models.
The hub also has a community collaboration aspect, which DeepSig said can accelerate development, reduce training costs and let users focus on fine-tuning models for specific needs.
“By offering a central model repository, we accelerate the ability of users to leverage powerful deep learning-driven AI/ML spectrum sensing with new signals and bands. This allows people to develop spectrum-aware systems and applications more quickly, without needing to build, curate, and train their own models to get up and running for a wide variety of use cases. In the future, we envision DeepSig’s Model Hub extending to vetted OmniPHY AI models and collaboration for 5G, 6G and RAN Digital Twin applications,” said DeepSig’s CTO Tim O’Shea.
DeepSig was among the first grant recipients of the National Telecommunications and Information Administration’s Public Wireless Supply Chain Innovation Fund, also known as the Wireless Innovation Fund; the grant bolstered research and development related to its generative AI and tools for modeling and measuring the wireless environment under real-world conditions, with a focus on applicability to Open RAN.
