The OmniSig sensor provides a new class of RF sensing and awareness using DeepSig’s pioneering application of Artificial Intelligence (AI) to radio systems. Going beyond the capabilities of existing spectrum monitoring solutions, OmniSIG is able to not only detect and classify signals but understand the spectrum environment to inform contextual analysis and decision making. Compared to traditional approaches, OmniSIG provides higher sensitivity and accuracy, is more robust to harsh impairments and dynamic spectrum environments, and requires less computational resources and dynamic range to process. The OmniSIG software can be deployed on a wide range of target devices ranging from low-SWaP mobile and embedded systems to cloud-based high-performance computational clusters, and it provides a set of streamlined web-based user interfaces, as well as open, standards based, low-latency streaming sensor metadata interfaces and control APIs for seamless integration and state of the art radio awareness integration into customer systems and applications.
The OmniSIG sensor is able to perform detection and classification of RF emissions across very large bandwidths of spectrum on the order of milliseconds, giving it the ability to report anomalies, changes or threats in real-time. It works with both wide- and narrow-band signals, and delivers accurate results for highly dynamic signals and within harsh or contested environments. Detection and recognition has been validated across a wide range of signal types, including numerous QAM, PAM, FSK, and analog single carrier modulation, multi-carrier modulation schemes, cellular and infrastructure signals (e.g., GSM, LTE, WiMAX), ISM-band signals (e.g., WiFi, BlueTooth), and mobile radio services (e.g., P25, GMR), as well as other numerous other types of emissions, and can be readily extended to include additional emissions and protocols based on customer requirements and applications. It's performance is robust to interference, both intentional and unintentional, and to a wide range of impairments caused by the receiver hardware and other sources.
The OmniSIG sensor’s machine learning architecture enables it to provide reliable, state of the art performance for a very wide range of applications. Using DeepSig’s dataset management, model training tools and software infrastructure, OmniSIG can be tuned and optimized for many different sensing needs, and deployed flexibly and efficiently anywhere, across wide range of operating platforms.
Deploying with OmniSIG
The OmniSIG sensor is a software product that customers can integrate into their own systems or third-party platforms. The OmniSig software is highly flexible, and can be targeted to a wide variety of processing platforms and elements, can use standard or custom radio interfaces, and is easily deployed and scaled using software including Docker containers.
The OmniSIG software typically requires the presence of at least one general purpose processor, such as an x86 or ARM core, and can accelerate signal processing throughput and efficiency using Graphics Processing Units (GPUs) including integrated NVIDIA Tegra cores or discrete GPU cards, Tensor and Vector processing accelerators, and FPGA resources. The OmniSIG sensor typically operates in a real-time streaming fashion, ingests radio samples from many common radio interfaces, and can make use of packet formats like VITA49 or SDDS. The web-based user interface can be used from any device with a browser, including mobile handsets, and the OmniSIG software also provides its metadata output stream in JSON form for use by other applications.
Theory of Operation
DeepSig’s revolutionary approach to signal processing leverages machine learning directly on time-series radio samples and channel measurements. By creating algorithms that learn from raw signal data and effects, DeepSig’s systems achieve significantly better performance than traditional heuristic and expert-feature based techniques.
Our approach allows us to optimize the system as a whole for a particular set of performance requirements rather than optimizing individual components, as is the practice in existing approaches.
The sensing techniques used within the OmniSig sensor have demonstrated a sensitivity of 4 to 10 dB better than prior state of the art methods, and can in some instances provide 10x or higher reductions in computational complexity through reduced sample precision and dynamic range requirements, enabling reductions in antenna size and receiver expense. It can also produce accurate sensing results with substantially reduced data requirements and dwell times providing further power reductions and throughput increases on equivalent processors. Sensing can be easily parallelized, scaled to application needs, and deployed on everything from a cloud compute cluster to a mobile ARM processor.
To learn more about using in your applications OmniSIG, please contact us!