Spending time in nature and listening to birds sing can be a relaxing way to enjoy free time. But for researchers at the University of California, San Diego, the hobby means business.
In a proof-of-concept study, a team of engineers and neuroscientists implanted silicon electrodes into zebra finches’ brains that recorded neural activity as they chirped away. With the help of artificial intelligence, the researchers were able to reproduce the pitch, volume and quality of the birds’ songs by translating their brain activity.
The computer-generated copies of the melodies may help inspire new types of vocal prosthetics that could translate the brain activity of people who have lost the ability to speak into any sound or word they think of, “freeing them to communicate whatever they wish,” according to a statement from Timothy Gentner, senior author of the study and a professor of psychology and neurobiology at UC San Diego.
Now, the team has to work on showing its translation system can do the job in real-time to prove it can accommodate the complexity of human speech. The study was published June 16 in the journal Current Biology.
“In many people’s minds, going from a songbird model to a system that will eventually go into humans is a pretty big evolutionary jump,” study co-author Vikash Gilja, a professor of electrical and computer engineering at UC San Diego, said in the statement. “But it’s a model that gives us a complex behavior that we don’t have access to in typical primate models that are commonly used for neural prosthesis research.”
UC San Diego Jacobs School of Engineering · Researchers re-create a bird’s song by reading its brain activity
The system was able to reproduce zebra finches’ songs with the help of mathematical equations that modeled the physical changes, such as pressure and tension, that occur in the birds’ vocal organ when they sing. The researchers then trained machine learning algorithms to connect the neural activity recorded with electrodes to these “representations” of the songs, instead of the actual songs themselves.
“If you need to model every little nuance, every little detail of the underlying sound, then the mapping problem becomes a lot more challenging,” Gilja said. “By having this simple representation of the songbirds’ complex vocal behavior, our system can learn mappings that are more robust and more generalizable to a wider range of conditions and behaviors.”
Some of the most advanced communication prosthetics for those who have lost their ability to speak include implantable devices that can use people’s brain activity to generate “textual output, writing up to 20 words per minute,” Gentner said.
Other more common prosthetics are valves placed in small openings between the trachea and esophagus that allow people to make sounds by pushing air through it and up into their mouth.
But a more successful device would need to be fast and intricate enough to keep up with people’s constant changes in speech and thinking that occurs while communicating.
“Imagine a vocal prosthesis that enables you to communicate naturally with speech, saying out loud what you’re thinking nearly as you’re thinking it,” Gentner said. “That is our ultimate goal, and it is the next frontier in functional recovery.”
The device could help the roughly 1 million Americans who have aphasia — the loss of ability to speak or understand speech because of brain damage — and the up to 10% with communication disorders.