ONCE MORE, WITH FEELING: EDUCATING ROBOTS ABOUT EMOTION

摘要

In the 2004 Will Smith summer blockbuster I Rabot, the most advanced robots had the ability to interpret the emotions of the human beings around them. They did it by analyzing the stress patterns in the voices they heard, an intentional echo of an earlier computer who did just the same thing—the HAL-9000 computer from 2001: A Space Odyssey. Today, the Scotland-based Affective Media Limited is working to help computers better understand people in various stages of emotional stress. Affective Media even has an online demo with an animated character named Tetchy the Turtle, who accepts voice samples and analyzes them. [See this online at http://www.affectivemedia.com/demo.htm] As soon as you have given Tetchy a four-second voice sample, it starts to process your sample. Eventually, the turtle begins to imitate your emotions, feeding them back to you, and providing hours of fun for the self-absorbed. In an interview with the British newspaper The Observer, Christian Jones, the chief executive of Affective Media, explained that humans "betray our emotions as we talk in dozens of subtle ways." When people are feeling blue or depressed, the pitch of the human voice drops and speech is slower; when angered, the vocal pitch rises and the volume goes up. Affective Media's recognition system identifies 40 of these subtle differences, Jones says. "It ignores the words you use, and concentrates exclusively on the sound quality of speech. It can tell your emotional state the very first time it hears your voice."