Innovation Artificial Intelligence

New MIT tech could make it easier for robots to understand human emotion

A device called EQ-Radio can measure changes in your breathing and heart rate to determine with 87% accuracy if you are happy, excited, angry, or sad.

Image: Jason Dorfman/MIT CSAIL

MIT researchers are another step closer to building robots that can detect human emotions, which might eventually help your business improve customer service.

On Tuesday, researchers from the school's Computer Science and Artificial Intelligence Laboratory (CSAIL) announced the development of EQ-Radio, a device that measures changes in breathing and heart rate to determine how an individual is feeling at that moment.

The device reflects wireless signals off a person's body, and uses algorithms to determine individual heartbeats and breathing patterns, and their corresponding level of arousal, according to a report. The radio can zoom in on human reflections, and ignore those from other objects in the area. After analyzing the results, EQ-Radio can tell if a person is excited, happy, angry, or sad--with an 87% accuracy rate.

"This work shows that wireless signals can actually capture meaningful information about human behavior that is otherwise invisible to the naked eye or hard to discern," said Mingmin Zhao, lead author of the EQ-Radio study. "By measuring small changes in breathing and heart rhythms, we can detect a person's emotions without on-body sensors or facial-recognition software."

SEE: Why the number of jobs that will be replaced by robots is lower than you think

Zhao said she believes the device could be used across a variety of industries, such as entertainment and health care.

"Smart homes could use information about your emotions to adjust the music or heating. The system could even make recommendations based on your mood, like suggesting that you go outside for a walk if you've been sad for a few days," Zhao said. "We believe a system like this could someday even be used to help monitor and diagnose conditions like depression and anxiety."

Since the machine analyzes the time between heartbeats, it eventually might be useful for health monitoring, the researchers said.

Image: MIT


In the future, EQ-Radio could also be combined with new specialized virtual assistants for better customer service experiences. "At a higher level, we view this work as the next step in trying to develop computers that can better understand humans at an emotional level and potentially interact with us similarly to how we interact with other people," Zhao said.

Existing emotion-detection methods--such as Microsoft's Emotion API--rely on audiovisual cues or on-body sensors, Zhao said. However, a person could potentially trick software based on facial expressions, such as by smiling when they are upset. And on-body sensors such as chest bands can be inconvenient to wear and need to stay in the same position at all times to remain accurate, Zhao said.

While EQ-Radio's detection capabilities vary from person to person, the device can detect emotions with 70% accuracy on the first try with a new person, the researchers said.

The 3 big takeaways for TechRepublic readers

  1. On Tuesday, MIT researchers announced EQ-Radio, a device that can detect whether a person is happy, sad, excited, or angry.
  2. EQ-Radio works by reflecting wireless signals off of a person's body, and using an algorithm to analyze the emotion associated with the pattern of arousal.
  3. The device might someday be integrated into specialized virtual assistants for customer service, and could have applications in the health care and entertainment industries.

Also see

Visit TechRepublic