One in 10 households will have a home robot by 2020, up from one in 25 last year, according to Juniper, although that includes less-smart robots like the Roomba intelligent vacuum.
But the wider acceptance of robots in homes depends on people getting more comfortable with being watched, and read, by a robot.
Generally, facial recognition technology starts with identifying where a face is, then what its features are, explains Jonathon Phillips, face recognition program manager at the National Institute of Standards and Technology.
"[Robots] learn mathematically what distinguishes different faces," said Phillips, whose agency works with the tech companies to develop new technologies used both by businesses and the government for national security purposes.
Once a robot identifies a face, it makes additional calculations to measure expressions, Phillips said. Changes in features, for instance whether the corners of the mouth are moving up or down, are sensed and interpreted, Phillips explained. There are a variety of programs that are used to recognize faces and interpret expressions, many of them based on proprietary, often closely guarded algorithms, Phillips said.
The more often a robot sees its user's face, the more accurate its emotion reading should become, Phillips said.