Millimeter wave (mmWave) Doppler radar can be used to detect different types of human activity, and that includes the ability to identify if someone is doing sports, if they are having dinner, if they have come home … activities that a smart assistant could use to help on a day-to-day basis.
Currently, many cameras are used for this type of task, capturing images and videos that, using intelligent identification systems, send information to “digital brains” to warn of what is happening, but this technique is not very friendly to privacy precisely.
What they are doing Researchers from Carnegie Mellon University’s Future Interfaces Group is to demonstrate that activity can be tracked without relying on cameras. mmWave offers signal richness approaching that of microphones and cameras, and over time they may have enough data to train AI models to recognize different human activities.
In the video you can see what they have achieved so far. The system detects if someone is cycling, clapping, waving, or squatting. It manages to interpret the mmWave signal generated by the movements, and as it has been properly trained, it shows the correct result.
This system is not used to detect different facial expressions, but it does differentiate between reading and eating.
You need a line of sight between the subject and the detection hardware, as well as special detection hardware, but it is possible that in the future our intelligent assistant will know what we are doing by spying on us with its invisible waves, not just listening to what we say.