The new year has only just started and in 3 days the “Consumer Electronics Show…
AI supported Smartwatch analyses mood in conversations
Whether the conversation partner is happy or sad is often not easy for people with anxiety disorders or Asperger’s syndrome, which in turn can lead to stress among those affected. A team of researchers from the Massachusetts Institute of Technology (MIT) would like to provide people with the assessment of social situations with artificial intelligence, which analyzes the mood during a conversation .
Algorithm detects mood
Tuka Alhanai and Mohammad Ghassemi have developed the technology for the Samsung Simband and recorded 31-minute conversations. The system analyzes audio recordings, text transcriptions and physiological measurements – with a probability of 83% the algorithm correctly assigns whether a conversation is predominantly cheerful or sad. A second algorithm analyzes the change in perceptivity from positive to neutral to negative in a five-second cycle, performing 7.5 percent better than existing systems.
Subjects had to tell either a happy or a sad story to test the AI. Long pauses and a monotonous voice point to more sad stories, while energetic, varied linguistic patterns are combined with more cheerful narratives. In analyzing the body language, it was noticeable that more sad stories led to more fidgeting and increased cardiovascular activity.
Detail accuracy is improved
“Our next step will be to improve the detail accuracy of the algorithm so that we can name boring, tense and exciting moments better, rather than describe interactions as” positive “or” negative, “Alhanai tells moobilux.com. They also hope to be able to gather more data in the future, possibly on commercial devices like the Apple Watch.