[ad_1]
For years, Pancho communicated by spelling out words on a computer using a pointer attached to a baseball cap; this was a demanding method that allowed him to type about five correct words per minute.
“I had to bow/lean my head forward and down and poke an important letter one by one to type,” he emailed.
Last year, researchers gave him another device containing a head-controlled mouse, but still not as fast as brain electrodes in research sessions.
Pancho communicated between 15 and 18 words per minute through the electrodes. This was the maximum rate allowed to run as the computer waited between requests. Dr. Faster decoding is possible, Chang says, but it’s unclear whether this will approach the typical speech-to-speech rate: around 150 words per minute. Speed is an important reason why the project focuses on speaking directly by tapping into the brain’s word generation system rather than the hand gestures involved in typing or typing.
“It’s the most natural way for people to communicate,” he said.
Pancho’s lively personality has helped researchers overcome challenges, but it sometimes makes speech recognition erratic.
“Sometimes I can’t control my emotions and laugh a lot and don’t do very well in the experiment,” she emailed.
Dr. Chang recalled times when the algorithm successfully identified a sentence, “you could see it visibly trembling and it looked like it was giggling.” When that happened, or when he yawned or got distracted during repetitive tasks, “it didn’t work out well because he wasn’t really focused on getting those words. So there are some things we need to work on because frankly we always want it to work.”
The algorithm sometimes confused words with similar phonetic sounds, describing “go” as “bring”, “do” as “you”, and words beginning with “F” – “faith”, “family”, “feel” as V. -word, “a lot”
[ad_2]
Source link