A Team of Engineers Reached a New Intelligible Speech Landmark
A team of specialized neuroengineers from the University of Columbia has reached a new landmark. They managed to create a new system that is able to convert thoughts into intelligible speech.
The task has been achieved by monitoring the brain activity of a person in order to transform the patterns into words that can be understood by anyone. A mix of speech synthesizers and AI is able to output the speech and it is already thought that the technology could open new ways of communicating directly with a brain.
It would also be a great relief for those that are unable to speak due to debilitating conditions like amyotrophic lateral sclerosis (ALS) or damage caused by stroke.
Speech has remained the primary means of communication since the humans began to use articulated language. To lose the capacity to speak is something that will dramatically alter your life in several ways, from professional opportunities to social relationships. The breakthrough discovery shows that it is possible to restore speech by harnessing the power of the brain.
Research which spans decades into the past notes that active speech or the wish to speak will prompt the brain to emit distinguishable patterns that can be identified. Similar patterns will also appear when we are actively listening to someone or even pretend to do it. Many scientists hoped to discover a method that would allow them to interpret these signals and use them in order to convert the thoughts into speech that can be heard and understood by others.
The challenge was far more complicated than it was thought at first. Several attempts were met with failure until now. The team of researchers led by Doctor Mesgarani decided to use a vocoder, a specialized algorithm that is able to synthesize speech after it is trained in order to understand how humans converse.
The technology is still in the early stages but it has the potential to improve the life of millions of people from all over the world. The preliminary results have been published in a peer-reviewed journal.