Rita Singh Talks Cutting-Edge Vocal Recognition Technology
The voice is a significant component of human interaction that allows us to characterize each other, and voice-activated and identification technologies are at the forefront of research to personalize day-to-day interactions with technology. Carnegie Mellon University's Rita Singh, a senior systems scientist in the Language Technologies Institute, recently spoke to CSO Online about her research in the field, which focuses on the core algorithmic aspects of computer speech recognition and learning from speech signals. She has found that voices can help identify personal characteristics beyond gender, age and race — like a person's education level, sociological trends in the population and socioeconomic status. Such profiling can be especially useful for law enforcement, by determining fake callers or finding perpetrators of voice-based crimes like bank fraud.
Singh further explained to Extreme Tech that even when individuals attempt to conceal their voices by using accents or changing modulation, her algorithms would still be able to identify them. This is due to micro expressions contained within a voice — for instance, the speed at which a person rises from a "t" to an "a" when pronouncing the word "tap." Such micro expressions are beyond an individual's control, as they cannot be easily faked or altered, and thus are picked up by the vocal-recognition algorithms.
"The goal of my work is to enable computing machines to recognize speech better in general, especially in high noise and complex environments," said Singh. "We are making judgments all the time through only hearing the signal a person produces."