Synthetic Sensors makes today’s smart devices feel dumb


In my quest to add smarts to my home, I’m quickly learning that it’s harder than it looks. When you finally do find the right device which meets your needs, having them all talk with each other is an even bigger challenge. And that’s even without addressing all your existing dumb devices. There’s more to think about than I had first thought, hence my personal home research project to see how far I can get.

But what if you could relieve the device of the burden of the most basic smart functions? A Google-funded Carnegie Mellon University research project thinks they might have the answer.

Synthetic sensor board

Synthetic Sensors, a single tiny device packed with sensors can make an entire room smart. By capturing a broad range of environmental data, these sensors are able to create a digital fingerprint of each device in the room. Even if more than one device is operating at a time.

Energy monitoring manufacturers have already caught onto device fingerprinting. By detecting fluctuations in the mains current, major running appliances are identifiable. This provides a less intrusive way to monitor power usage without having to go to the source device. Synthetic Sensors are performing a similar task but in much more detail, and about other data. I can’t help but think Synthetic Sensors and energy monitoring fingerprinting could make a nice match.

Understanding device events, not just state

This sensor data when mixed with machine learning is capable of identifying precise details about complex events in the room. One example shown in the video understands the state of a ‘dumb’ microwave as the door is opened and closed, or whether it’s in use or completed cooking. This sequence of device states creates an event. Or how about knowing when a paper towel dispenser is not only in use, but also how many sheets have been dispensed. Data such as this could help further automate the facilities management of a building.

Synthetic sensors signals

And since most of the processing is performed on the device, the data is not transmitted off premise. Also in keeping with privacy concerns it doesn’t include a camera.

The goal is not for the sensor to act as a device itself with a user interface. Instead it complements other devices through constant detailed environment analysis in the background. It’s this data which can then be used by smart home hubs and virtual assistants.