Siri: Speech Recognition Software Being Developed to Detect Human Emotion

Tweet

“I’m getting a little bit flustered, Siri,” you might say to your iPhone 4S in frustration. But while Siri understands human speech instructions without the need for pre-set commands, it doesn’t really understand human emotion, does it? That might change soon, as a team of university researchers are developing software that could pinpoint negative emotion in human speech, leading to all sorts of applications.

Researchers from two universities in Spain have come up with software that detects human emotion in speech. Scientists from the Universidad Carlos III in Madrid and Universidad de Granada say they have developed their software to particularly point out negative emotions, such as anger, boredom and doubt. These, they say, have applications in speech-powered telephone prompts, which usually earn the ire of callers for being impersonal and degrading.

Once the system has identified the mood of a person, it can change its programming and dialog accordingly. So, for instance, if a caller is agitated, the system might find ways to calm the caller down, which can be particularly helpful in customer service calls. Or, the system can refer a person to a human correspondent.

The software uses a set of 60 acoustic parameters, which include pauses, tone, speed, stress and energy of the voice. This technology can be paired with one that likewise detects human emotion based on facial recognition, or even motion-sensing technologies. While the system is not currently being marketed, perhaps companies that incorporates speech recognition, such as customer service call centers, and even Apple’s Siri speech recognition engine, will find this useful in the future. Maybe someday, Siri will not only make witty comments based on what you ask or say, but can even know when you’re happy or not feeling well.