For years, Affectiva has been able to detect emotions with its facial recognition software. Now it’s adding the ability to do so by listening to your voice.
Affectiva, one of the top 10 funded artificial intelligence startups in Boston, announced that it is launching cloud-based API software that can distinguish emotion in speech, which the company says brings it closer to developing the first ever multi-modal AI. The announcement was made at the company’s Emotion AI Summit on Wednesday.
In an interview with BostInno, Abdelrahman Mahmoud, a product manager at Affectiva, said the company envisions a number of use cases for its new software. That includes customer support environments where honing in on a customer’s verbal expressions could indicate whether they need to be elevated from an automated voice service to a human support professional. The company also sees the potential for use in market research, an area Affectiva has already been working in with its facial recognition tech.
But perhaps the most compelling use case Affectiva envisions is for the automotive industry, where the ability of a vehicle to detect emotions like anger could determine whether you need more help with directions, among other things.
“Artifical emotional intelligence is going to be just as critical to our AI systems.”
In an interview earlier this year, Affectiva CEO and co-founder Rana el Kaliouby told BostInno that emotion AI could help with things like detecting driver distraction or drowsiness. She added that it could also become a critical component in autonomous vehicles, where detecting a passenger’s emotions could help a car determine whether to slow down or pull over.
“Artifical emotional intelligence is going to be just as critical to our AI systems,” she said.
Taniya Mishra, a lead speech scientist at Affectiva, told BostInno that the new software focuses on how people speak and not what they actually say. For example, this can help to determine if someone is confused because of their speech pattern, Mishra said.
Affectiva’s new voice software can currently detect anger, arousal, laughter and gender, Mishra said, and the company plans to expand its emotional recognition abilities over time. She said it takes a few seconds for the software to recognize emotion, which is longer than the 700 milliseconds it takes for humans but fast enough to still be effective.
The company’s goal is to combine its facial and speech recognition capabilities to gain a more nuanced understanding of how people express emotions. Mishra said this can be important, for example, when the tone of your voice is communicating one thing while your facial expressions are indicating something else.
“It’s a challenging problem, but the payoff is huge,” she said.
The company said it is already beginning to integrate the new tech with some clients, though it’s unable to disclose who they are. For its facial recognition tech, Affectiva works indirectly with 1,400 brands that rely on the software for things like market research. The company’s facial recognition has also been used in a psychological thriller video game that adapts its gameplay when it senses the player is emotionally distressed.
Founded in 2009 as an MIT Media Lab spinoff, Affectiva has raised $26 million in venture capital. Investors include Kleiner Perkins Caufield & Byers and Fenox Venture Capital.