Facebook is actively funding a bleeding-edge technology that aims to let people type by merely thinking. The study conducted by scientists of the University of California San Francisco extensively focused on developing machine-learning algorithms capable of translating speech from brain activity.
The social media giant believes that this innovative study will help them develop a “non-invasive, wearable device that lets people type by simply imagining themselves talking.” The study was successful in epilepsy patients with electrodes placed on their head for assessing their seizures before surgery.
The scientists had given patients a list of multiple-choice questions in random order and asked them to read the answers out loud. According to this experiment, it was found that the algorithm was able to identify the questions asked 75% of the time and their answers 61% of the time.
While several past approaches focused solely on decoding the speech, this study focused on developing a solution that catered to both the questions and answers making it a fairly advanced and sophisticated solution.
“Most previous approaches have focused on decoding speech alone, but here we show the value of decoding both sides of a conversation – both the questions someone hears and what they say in response. This reinforces our intuition that speech is not something that occurs in a vacuum and that any attempt to decode what patients with speech impairments are trying to say will be improved by taking into account the full context in which they are trying to communicate,” said Prof. Eddie Chang.
The scientists said that people suffering from disabilities such as speech loss or paralysis are usually limited to spelling words through residual actions such as eye movements or muscle twitches to a computer interface. However, they’re well equipped and possess enough information in their brains to communicate fluently.
This study thereby aims to create a technology that can help such people express fluently by using electrodes and decoding their thinking to produce fluent speech. As of now, this technology was tested with a very limited vocabulary; the scientists expressed that they “hope to increase flexibility as well as the accuracy of what we can translate from brain activity.”
Currently published in Nature Communications, the scientists have revealed that the algorithm was able to generate a rate of 100 words per minute based on their initial experimentation.