People who spend lots of time around Shrikanth “Shri” Narayanan may not hear him ask “What are you thinking?” too often. That's not because he doesn't care but because he might just try to analyze their gestures and vocal sounds instead.
“We are studying human behavior of various sorts,” says the professor of electrical engineering, linguistics and computer science at USC's Viterbi School of Engineering. “These are complex things that are mostly very qualitative. How can engineering — technology and the underlying mathematics, as well as psychology — be used to study human behavior in a faster, more economical way? We're trying to create those tools.”
Narayanan's research utilizes “behavioral signal processing,” or BSP, to measure a person's word choice, vocal synchronization, gesturing and even subtle physiological changes.
Among other groundbreaking efforts, while studying sophisticated motion capture of couples in therapy, Narayanan and his team identified signs that a relationship was under greater strain than acknowledged: if a couple excessively used the word “you,” if the pair physically leaned away from one another, or if one spoke softly and the other loudly — a failure to mirror vocal affect. The data are meant to help therapists dig deep into problems that might have proved elusive.
His team also created a futuristic software program that measures the electrical waveforms created by a person's voice when he or she calls a company's phone-answering system. The system, still under study, transfers the caller to a human operator if his or her wave patterns reach a zone his team identifies as “angry” or “frustrated.” Naturally, the program also recognizes expletives.
Narayanan and his team analyzed 1,400 calls to a real company in order to formulate their baseline thresholds, and their research shows that 80 percent of the time, their program is right about the caller's feelings. It's also self-improving, its algorithm adding to its knowledge base with every new example.
“My formal training as an engineer gave me understanding of systems, and I think humans are one of the most complicated systems,” says the lean, wavy-haired Narayanan, his expressive smile appropriate to the nature of his research. “I look at things across disciplines — with linguists, psychologists, artists — and it's very collaborative.”
Originally from India, Narayanan earned a Ph.D. in electrical engineering from UCLA and worked at Bell Labs on the East Coast before returning to California to work at USC. Now, the soulful elements of his work dovetail perfectly with his personal interests. “I'm a musician, interested in Indian classical music, and I play every day to have fun,” he says.
His instrument is the vina, a sort of forebear to the sitar, which was played, according to the Hindu religion, by the goddess of learning. “It's almost meditative for me.”
One of Narayanan's groundbreaking studies involved using magnetic resonance imaging to understand how an American beatboxing expert — a musician who uses his mouth, lips, tongue and voice to create percussion — achieves those sounds. The MRI revealed some unusual and difficult sound formations that Narayanan says also are present in languages indigenous to Chechnya, Peru, British Columbia and certain African tribes.
Narayanan yearns to address human emotional and mental suffering, especially when its victims can't. “Besides better understanding human communication,” he says, he hopes his research can someday be applied to “autism, addiction, depression and other ubiquitous human conditions.”