With almost one hundred percent accuracy Virtual assistants Sber Salyut, Yandex.Station – all this can be controlled by voice. But what if a person cannot speak, for one reason or another? The answer was found at the Cornell College of Computing and Informatics. Discuss ” alt=”Engineers have created a necklace that recognizes the wearer's words without pronouncing them” Engineers have created a necklace that recognizes the wearer's words without pronouncing them />
Cheng Zhang, assistant professor of computer science and doctoral student Ruidong Zhang, developed SpeeChin, a silent speech recognition device that can identify silent commands using images of skin deformation on the neck and face taken with an infrared camera mounted on the neck.
“This device has the potential to study human speech patterns, even when speech is silent,” said Cheng Zhang.
The SpeeChin features an infrared camera mounted on a 3D printed necklace body that is hung from a chain with the camera pointed at the wearer's chin. To increase stability, the developers created a wing on each side and placed a coin on the bottom.
During the first experiment, which was attended by 20 people (10 in English and 10 in Mandarin Chinese dialect), measurements were taken to determine the initial position of the chin, then the differentiated images were used to train the device to recognize simple commands. , consisting of numbers, interactive commands, voice assistant commands, punctuation commands, and navigation commands. It then did the same with 44 simple Mandarin words or phrases.
SpeeChin recognized English and Mandarin commands with an average accuracy of 90.5% and 91.6% respectively.
Source: Cornell