While whispering, without moving the lips, the vocal cords move minimally. The machine will decode the movements and translate them into clear words. A more sophisticated system will help the astronauts
By: Yuval Dror, Haaretz, Walla News!
For decades, scientists from all over the world worked on a machine that could read minds. But the brain is too complex, and the information scientists have about the way the brain works is too little.
Yesterday, the American space agency NASA announced that it is working on a development that comes close to its ability to read minds. The device, which combines special sensors and computer software, is able to decode words spoken by the subject even though he did not utter them out loud.
The development is based on the fact that even when people read a text to themselves in a quiet voice, even without moving their lips, the vocal cords, tongue and throat muscles move in a tiny, almost imperceptible way. "A person using his sub-vocal systems thinks of a sentence and says it to himself so that the words cannot be heard, but the tongue and vocal cords receive messages from the brain and move accordingly," explained Chuck Jorgensen, who heads the team developing the technology for sub-speech recognition vocal. "It's like a person whispering, and gradually weakens the intensity of the whisper until no sound is heard," he explained.
Surf the Internet without touching the keyboard
According to him, his team came to the conclusion that the tiny movements of the tongue and vocal cords can be tracked, analyzed and translated into real words. The scientists attached special sensors under the subject's chin as well as on both sides of the glottis. The sensors monitor the activity of the throat muscles and the activity in the vocal cords and send the information to a computer program. The software sifts through the silence and amplifies the weak signals, which are not normally heard. Using the signals, she tries to decipher the vocal pattern and convert it into a clear word.
The received signals are not clear enough, so the scientists are forced to "teach" the system to recognize the words, according to the patterns picked up by the sensors. The first words that the system learned to recognize are "stop", "go", left", "right", "alpha" and "omega", as well as all the numbers. The success rate of the system is estimated at 92%.
To prove that the software can understand subsonic speech, the scientists developed a simple system: the subject whispers to himself a combination of two digits. Each combination (for example "1" and "4") refers to a certain letter in the alphabet, and thus the subject can spell complete words. After spelling a word, it is sent to a search engine such as "Google", which links the person whispering to a website corresponding to the word. "The experiment proved that we are able to surf the Internet without touching the keyboard," said Jorgensen. "We don't read thoughts but the vocal results of the thoughts".
The technology will be applied in the space program to conquer Mars
Subsonic speech recognition technology may have many uses. Jorgensen believes that the technology can be applied in NASA's space program to conquer Mars. "An expanded and more sophisticated system could help injured astronauts control computers and machines," Jorgensen explained. "If an astronaut sent to Mars suffers from muscle weakness as a result of the lack of prolonged exposure to gravity, he will be able to use the system to land the spacecraft on Mars or during the journey back to Earth."
Security forces around the world may also find the system effective, for example when assaulting forces have to coordinate their activities without causing noise that would attract the enemy's attention.
NASA believes that the technology also has many civilian applications. For example, a person could speak in a whisper to his cell phone device even when he is in a noisy environment and the device knew how to translate his words to the other party. Even flight controllers who work alongside other controllers will be able to give precise instructions to the pilots without having to raise their voices to overcome the noise in the room. Obvious candidates for utilizing technology are people who have difficulty speaking. "In order to improve the system, it is necessary to work on its two main components: the sensors that pick up the signals and the software that processes them," Jorgensen concluded.
