TOKYO — We all know the phrase: “If you’re going to talk the talk, then walk the walk”—usually said by surly cops to snotty punks in B-grade action movies. Robotic scientists in Japan, however, discovered that you don’t have to speak a word to get your meaning across—your gait can pretty much convey your emotions. Scientists from Japan and France, working together, announced last month that they had developed an algorithm that can recognize emotions from a human gait.
The research will be used to make advances in robotic technology and facilitate smoother interactions between humans and robots.
“When it comes to designing robots to interact with humans, we have to understand humans, so we’re also doing research in understanding human non-verbal communication,” said Gentiane Venture, an associate professor at Tokyo University of Agriculture and Technology and the head researcher for the project.
“We don’t want to have a robot that is asking questions all the time,” said Venture, “but rather understanding what you mean by your body language and your nonverbal communication.”
In the study, which was published recently in the International Journal of Social Robotics, four actors performed different types of walks that conveyed five main emotions (neutral, happiness, anger, sadness, and fear) and the research team had 20 test subjects attempt to recognize which emotions were being conveyed. The researchers found that certain parameters, such as walking speed and posture, can affect the recognition of emotion.
Based on these results, the research group produced an emotion recognition algorithm. While changing the weight of the parameters (speed and the position of the torso and head) and verifying the success rate of emotion recognition, they concluded that specific parameters greatly affect emotional recognition. In other words, you don’t need to look at the entire human body to read what the person is thinking—almost everything can be read in the torso.
“Some studies say that nonverbal communication is 90 percent of communication,” said Venture. “So the idea is that having a robot that can understand human nonverbal communication will really help in having a system that people would like to interact with and would like to play with or to work with.”
In an aging society that faces a rapidly shrinking workforce, this data could be used to program robots as better caregivers and companions. Venture notes that Japan has been thinking for several years about using robots to work with the elderly: “If you have a robot who doesn’t understand your feelings, then you don’t want to interact with the robot. It would be like just going to your fridge. We don’t want fridge robots. We want robots that can support you when you need to be supported and want to be supported.”
The research on gait may also be used to make robots move in a more natural way. As part of this research they have programmed one particularly cute robot named NAO to walk like a human and express emotions such as “happy.” The research teams also uses NAO in the lab for testing some human robot interaction and works in conjunction with psychologists to understand human behavior toward humanoid robots.
It should be said, that while the focus of the research is on creating helpful and friendly automatons, like all technology it could also be subverted for other uses, such as creating killer robots. One of the five gaits that can be recognized with the algorithm is “anger”—straight back, forward leaning, fast pace--which could be a precursor to armed attack, and that is something a military robot would seek to pre-empt. However, it’s hard to visualize cute little NAO blowing up hostile combatants. The goal of the present research is to help create the programming for a robot that is “a sociable partner.”
“Without having a system that can understand you, it’s impossible in the long term to think of working with robots,” said Venture. “Without having a system that understands humans and knows what’s going on, it’s like working with a colleague that never asks you, ‘How are you today?’”
It’s an interesting point in an age where many tasks once conducted by humans are now farmed out to drones. Even the Associated Press has started using “robots” to write some routine business stories, rather than humans.* We may all have to learn to work with robots in the near future—and it would be great if we could get along.
By the way, dear reader, how are you today?
*This article was written by two humans.