AI-equipped glasses read silent speech Neuroscience News


Summary: Researchers have developed a wearable interface called EchoSpeech, which recognizes silent speech by tracking lip and mouth movements through acoustic sensing and AI. The device requires minimal user training and recognizes 31 unannounced commands. The system can be used to give voice to people who are unable to speak sounds or communicate silently with others.

Source: Cornell University

Researchers at Cornell University have developed a silent-speech recognition interface that uses acoustic-sensing and artificial intelligence to continuously recognize 31 nonvocalized commands based on lip and mouth movements.

The low-power, wearable interface — called EchoSpeech — requires only a few minutes of user training data before it can recognize commands and run them on a smartphone.

Ruidong Zhang, doctoral student in informatics, is lead author of “EchoSpeech: Continuous Silent Speech Recognition on Minimally-Obstructive Eyewear Powered by Acoustic Sensing,” which will be presented at the Association for Computing Machinery’s Conference on Human Factors in Computing Systems. CHI) this month in Hamburg, Germany.


“For people who cannot raise their voice, this silent speech technology can be an excellent input for voice synthesizers. It can give patients back their voice,” said Zhang, with further development of the technology’s potential uses. Said about

In its current form, EcoSpeech can be used to communicate with others via smartphone where speech is inconvenient or inappropriate, such as noisy restaurants or quiet libraries. Silent Speech Interface can be combined with a stylus and used with design software such as CAD, but eliminates the need for a keyboard and mouse.

The low-power, wearable interface — called EchoSpeech — requires only a few minutes of user training data before it can recognize commands and run them on a smartphone. image is in the public domain

Crafted with a microphone and a pair of speakers smaller than a pencil eraser, the EcoSpeech glasses become a wearable AI-powered sonar system, sending and receiving sound waves across the face and sensing mouth movements . A deep learning algorithm then analyzes these echo profiles in real time with approximately 95% accuracy.

“We’re taking sonar to the body,” said Cheng Zhang, assistant professor of informatics and director of Cornell’s Smart Computer Interface for Future Interaction (SciFi) Lab.

“We’re very excited about this system,” he said, “because it really advances the field on performance and privacy. It’s small, low-power, and privacy-sensitive, which enables new, wearable devices to be used in the real world.” All are important characteristics for applying qualified techniques.”

credit: Ruidong Zhang

Cheng Zhang said that most technology in silent-speech recognition is limited to a select set of predetermined commands and requires the user to face or wear a camera, which is neither practical nor feasible. There are also major privacy concerns associated with wearable cameras — both for the user and those with whom the user interacts, he said.

Acoustic-sensing technology like EchoSpeech removes the need for wearable video cameras. And because audio data is much smaller than image or video data, it requires less bandwidth to process and can be relayed via Bluetooth to a smartphone in real time, said informatics professor Franois Guimbretier. Said.

“And because data is processed locally on your smartphone rather than uploaded to the cloud,” he added, “privacy-sensitive information never leaves your control.”

Summary written with the help of ChatGPT AI technology.

About this AI research news

Author: Becca Bowyer
Source: Cornell University
contact: Becca Bowyer – Cornell University
image: image is in the public domain

Basic Research: The findings will be presented at the Association for Computing Machinery’s conference on Human Factors in Computing Systems (CHI).

Source


Related News

How to Delete Your Reddit Account and History

When you've gotten used to posting on Reddit, it's easy to go down the rabbit hole and start posting and commenting everywhere. If and when you want to delete

Android Auto offers support for one-tap quick replies to messages

Months after the feature was announced, Google is now rolling out support for quick replies to messages via Android Auto.

This bitcoin wallet is as small as a credit card, but with an unbreakable metal construction

It comes complete with a QR code for easy transfer, an unbreakable design that's water and fireproof, and due to its analog nature, it's virtually hack-proof

How to make Windows 11 look like Windows 10

Despite Windows 11's modern and stylish appearance, some people prefer the familiar look of Windows 10. If you're one of them, there are plenty of ways to

Vodafone presents its new WiFi 6 router with which it promises to quadruple the speed: this is how you can request it

Now Vodafone is betting on the same system with the Vodafone WiFi 6 station. A new model that the red operator begins to distribute among fiber customers,

The opening of Shikimori’s Not Just a Cutie reached 1 million views on YouTube

On 9 April 2022 theanime Shikimori's Not Just a Cutie (Kawaii Dake Ja Nai Shikimori-san) adaptation of the manga of the same name Keigo Maki, was released on