South African researchers are working on a new type of artificial larynx that offers a more natural kind of speech, doing away with the “robot voice” that characterizes current implants.
Using a palatometer, which looks like an orthodontic plate and is normally used for speech therapy, the new system tracks the contact between the tongue and palate using 118 embedded touch sensors to determine which word is being mouthed.
It uses a speech synthesizer to actually generate sounds.
About 10,000 Americans are diagnosed with laryngeal cancer each year, according to the National Cancer Institute. Most patients with advanced cancer must have their larynx removed.
Want to hear what current implants sound like? Just watch any of Geovision’s anti-smoking ad spots, which ran in the northeastern U.S. and featured Ronaldo Martinez, a cancer survivor who has an artificial voicebox.
To use the device, a patient puts the palatometer in his or her mouth and mouths words normally. The system translates those mouth movements into words and reproduces them on the synthesizer.
The software for the device was written by researchers at the University of the Witwatersrand. For now, the system can recognize 50 common English words, just by saying each word multiple times with the palatometer in the mouth.
It translates that information by representing it on a binary space-time graph and entering it in a database. Each time the user speaks, the contact patterns are compared against the database to identify the correct word.
The team has developed several techniques for word identification. One approach aligns and averages the data produced while training the device for a few instances of a word to create a template for comparison. Other approaches compare words on the graph, or with voting, or even using predictive-analysis based on the previous word spoken.
The accuracy? 94.14 percent for words it doesn’t consider “unknown,” which occurs 18 percent of the time. (It skips those to avoid awkward social situations.)
More testing is expected to come, and the team is exploring wireless data transfer, better predictive analysis, faster processing and user-controlled modulation of pitch and volume for a more natural sound. But the development of a device that doesn’t require surgery is promising.
The team’s work is being presented at the International Conference on Biomedical and Pharmaceutical Engineering this week in Singapore.