Intelligent object classication for radar based presence detection 

Traditional sensing technologies like PIR, UWB, and ToF typically only provide basic presence information. Our radar sensing solutions leverage edge AI to enable advanced sensing capabilities like object classification. Allowing presence detection sensors to identify detected objects like humans, and animals to respond accordingly in real time.

Powered by our low power, cost optimized IWRL6432 radar, engineers can easily enable advanced object classification features to the next generation of presence detection. 

Capture accurate radar data with the IWRL6432 devices, enabling full AI processing done completely on the edge through the integrated M4F MCU and HWA.

Our radar sensors support a wide range of output formats, including range profiles, range-doppler heatmaps, point clouds, and more. For pose estimation, the model leverages object point cloud data to achieve high accuracy performance.

Customers can get started immediately using the sample human vs fan dataset available in the Radar Toolbox. Alternatively, users can train their own point cloud data and experiment with new objects and optimize performance, as detailed in the Human vs Non-Human classification user guide.

Retraining the human vs. non-human radar model with additional object point cloud data can be performed using the MATLAB Visualizer found in the Radar Toolbox. 

Users can select which features they want to use in the model, the number of radar data frames to consider as input into the model, and even augment data to create larger, more robust datasets.

Once trained and validated, Radar Toolbox provides an end to end workflow for deploying trained models directly to TI's Radar devices without any other systems needed.