The Huffington Post Bianca Bosker First Posted: 12/04/12 EST Updated: 12/05/12 EST
Your future smartphone could know from the sound of your voice that you're stressed, and automatically book you a massage. Or it might sense you're feeling sad, and cue up an ad for antidepressants.
Artificial intelligence, meet artificial intuition: Researchers at the University of Rochester have developed an algorithm that can detect human emotions through speech with greater accuracy than was previously possible and without analyzing the meaning of the speakers' words.
To create the computer program, the University of Rochester team identified 12 different acoustic features of a speaker's voice, such as pitch and loudness, and used those to classify short snippets of speech as expressing one of six different emotions, from "anger," sadness" and "disgust" to "happiness," "fear" and "neutral." The software was then instructed to associate certain speech characteristics with specific positive and negative emotions, and produced a new algorithm that can correctly identify human emotions with 80.5 percent accuracy, up from the 55 percent accuracy offered by prior programs.
University of Rochester graduate student Na Yang, who co-authored the research paper, has already developed a Windows Phone smartphone app, Listen-n-Feel, that functions as a "mobile emotion sensor" and allows the phone to detect a speaker's feeling, showing a smiley face or frown-y face depending on the speaker's tone.
And the algorithm doesn't even need to follow what the speaker is saying -- only how.
"We actually used recordings of actors reading out the date of the month -- it really doesn't matter what they say, it's how they're saying it that we're interested in," said Wendi Heinzelman, a professor of electrical and computer engineering at the University of Rochester, according to a press release issued by the university.
The report's authors hypothesize that the technology could be used by psychologists or clinicians to monitor patients in a more natural environment, or integrated in smartphones so a gadget could "choose songs based on the user's current emotion."
It's easy to imagine a whole host of seemingly sci-fi applications. What if Siri could whisper sweet nothings to you when you're stressed? Or serve up Prozac ads when you're sad, promos for beach getaways when it senses frustation, or Starbucks ads when you sound sleepy? If a phone can sense what you're feeling, so, more likely than not, could advertisers.
Affectiva, a company that tracks emotion using facial cues, is already being used by brands for market research and to gauge the effectiveness of ads. In an interview with the Technology Review published in July, then-Facebook engineer Andrew Bosworth suggested the company would consider targeting ads using sounds picked up via phone's microphone.