Radio frequency (RF) and wireless systems are intertwined with all fields of technology and are essential for cellular, communications, and surveillance equipment. Since inception, RF technology has relied upon engineering experts to innovate and build the next generation of products with ever increasing complexity.
In 2012, Alex Krizhevsky, Geoff Hinton, and Ilya Sutskever redefined the image classification world by demonstrating that deep learning, a new type of artificial intelligence (AI), can outperform the best human-engineered solutions. With this development the field of computer vision has grown in reliability and is currently utilized in self-driving vehicles, robotics, and many other aspects of AI.
A significant number of tasks in the telecommunication, defense, and internet of things (IoT) sectors are still reliant on human-driven analytics. For example, detecting, locating, and shutting down spurious interference signals within the RF spectrum is still, primarily, executed by people.
Imagine a smart city with an ecosystem of fully inter-connected sensors measuring weather, traffic, pollution, energy consumption, and terrorist threats. The city is likely reliant upon a wireless RF network for synchronization and receiving the sensor data. Any interfering signal could significantly impact the ability to monitor these sensors. Using humans-driven analytics will become unsustainable as the number of sensors and networks increase in complexity. Using artificial intelligence and deep learning, the task of monitoring the spectrum and locating nefarious signals may be automated.
The first step is to recognize and identify the tasks that are either human-engineered or human-analyzed within your domain. If any of these tasks are limiting development, deployment, or analysis time, a deep learning approach may be suitable to increase productivity while decreasing cost.
RF and wireless systems have the task of transmitting and receiving data (voice, content, power) and are, inherently, a hardware technology. No matter the application, any RF or wireless solution will a require hardware component. This is why Deepwave has invented the Artificial Intelligence Radio Transceiver (AIR-T). The AIR-T system incorporates the compute engine of AI, the GPU, with a software-defined RF transceiver.
Deepwave’s AIR-T system provides a low-cost platform to acquire signal data, train deep learning algorithms, and perform inference (deploy) your newly created deep learning RF technology. We abstract away the difficult task of communicating with the receiver and data acquisition systems so you can focus on the interesting part: your deep learning application.
