Sensors of tomorrow know everything that happens in homes

Smart home devices are great at automating single devices or performing specific tasks, but the all-knowing home depicted in science fiction is still just fiction. Synthetic Sensors could change that. This project from Carnegie Mellon University uses sensors and machine learning to deduce precisely what devices are being used in a home. A demonstration on the project's website shows researchers have been able to make Synthetic Sensors eerily accurate, at least in a controlled setting.

The technology works by using data from inputs like power, audio and light. Each input requires their own sensors for by themselves they might not be very intelligent. Synthetic Sensors is able to combine the data and determine precisely what is happening in a room. An open refrigerator door, for example, produces a specific sound and turns on a light, both of which can provide a unique data set to nearby sensors. Synthetic Sensors can then pass on to the user that the refrigerator door is open.

The system can also tell whether a blender or a coffee maker is in use by their unique sounds and power consumption. This works even when a device is moved to a different part of the room. Given these, researchers have been able to make the system incredibly accurate in isolating the right data to determine what's happening.

Synthetic Sensors can also be paired with other sensors and trigger specific actions based on what it detects. When a faucet is turned on, the system could also trigger a sensor that helps users keep track of water consumption. Just as smart home devices already promise to help people save money by conserving electricity, this logic can be easily extended to other utilities with Synthetic Sensors.

Equipping every room with various types of sensors might not be feasible or practical, but the fact that something like Synthetic Sensors can be made to work so well shows the promise of the smart home future. It also adds tremendous value to sensors that are already on the market.