The high-resolution output from the Acconeer radar sensor is perfect for use with deep learning algorithms. Using the new Deep Learning Interface (DLI) in our Exploration Tool, you can record data and train your own Artificial Intelligence (AI) model without any need for programming. This is useful when evaluating and implementing use cases such as gesture detection for control of electronic devices, surface classification for use in robotics and many more.
The DLI is built on the open source AI tools TensorFlow and Keras, and lets you train and use AI networks with the A111 radar sensor in a simple and convenient way. To make it easier to evaluate and implement radar-based use cases built on AI, we have also released detailed DLI documentation explaining how to use it and the steps needed to get started.
One example of how the deep learning in Exploration Tool can be used has been implemented as a master’s thesis collaboration between Lund University and Acconeer. In her work, Mathematical Statistics student Eda Dagasan trained an AI network to recognize a set of gestures using the radar data from Acconeer’s A111 radar sensor. The resulting model managed to identify gestures with more than 95% accuracy.
The Deep Learning Interface and master’s thesis are examples of several investments we’ve made in the AI area recently. Another example is the Gesture-controlled headphone demo created together with Edge AI company Imagimob, a project that is partly funded by Swedish innovation agency Vinnova through their program Smarter Electronic Systems.
To get started using radar with AI, find Exploration Tool on GitHub, learn how to use the deep learning features in the DLI documentation and get one of our modules or evaluation kits (EVK) from Digi-Key. Once you have the hardware, you can have it up and running in less than an hour and start playing with your own AI-based radar use cases.
