Language is based on a set of phonetic sounds that make up the words that we speak. Jonathan Pearl, president and chief scientist at Racine-based Perceptral LLC, has developed a way to identify the key components of phonetic speech in order to analyze, morph, modify and blend the sound in a way that previous technology would not allow.
His firm is currently working with the U.S. Navy and the U.S. Army on projects related to sound modification and blending that could revolutionize sound technology capabilities.
“What we have done is taken a new approach to sound analysis,” Pearl said. “Right now, linguistics is based on the phenomenon of the phonetic alphabet or letter sounds, but we know that those letter sounds don’t make up everything that we hear.”
Historically, voice recognition and sound technology has always forced the user to accommodate the device, Pearl said. “We’ve always had to speak more clearly in order for the device to recognize phonetic similarities,” Pearl said. “What we can do now is not only identify the similarities but also the subtle differences between similar sounds so that we can more aptly modify the sounds how we want.”
Perceptral was founded in May of 2008. In August of 2008 Pearl and his team were awarded an approximately $80,000 Phase I Small Business Innovation Research grant from the U.S. Navy. That contract included a $70,000 option, which has since been extended to a Phase II project. Since 2008, Perceptral has earned approximately $970,000 in Small Business Innovation Research grants for its technology, Pearl said.
“The project for the Navy was based on speech synthesis,” Pearl said. “The Navy posed a problem they had with traditional synthesized voices for their training routines and we came up with a solution.”
Military officers have a distinct tone and feel to their voice, Pearl said. “They speak sternly, quickly and clearly,” he said. “With traditional synthesized voices you didn’t get all of those qualities. When you tried to speed it up, the pitch sometimes changed or the intelligibility of the voice was diminished.”
Perceptral’s technology is capable of identifying the two definitive components of any sound clip, Pearl said. Once those components are identified, all other components can be compressed or modified to reflect the desired changes in the sound.
For the Navy, those changes were making a traditional synthesized message compressed enough to speed it up without losing the stern pitch or the intelligibility of the voice.
In the future, those changes could be used for morphing an individual’s accent, inflection and emotion into another language, or morphing a human’s voice with a cat’s meow for animation purposes.
“With our technology we can potentially reconstruct a degraded signal,” Pearl said.
“We can focus on the components that make up certain accents for accent identifications, that technology can then be implemented in call centers or in the military,” he said.
Perceptual recently won a Phase I U.S. Army contract for a project that uses sound technology to identify individual voices from within a crowd of people or background noise, and to pull out individual voices from multiple voice streams if there was cross channel interference on a radio signal or telephone conversation, Pearl said.
“The possibilities are endless,” Pearl said. “Not only for entertainment purposes like morphing a popular actor’s voice into another language for a foreign version of the same film, but also on the military front when it comes to identifying, learning and speaking in different accents of a foreign language.”
Pearl used the example of dubbing one of Sean Connery’s films in a different language like French. “Rather than trying to find a Scottish person who could speak French with a Scottish accent, you could potentially use our technology to modify Sean Connery’s accent into a French speaking person’s sound,” he said. “That’s where this technology is going.”
