Brain Computer Interface (BCI)
BCI basically is direct communication between the brain and an external device. BCI technology provides possibilities for those who are unable to communicate verbally due to paralysis. In this study Berezutskaya used a speech reconstruction method using brain recordings while test subjects had to speak Dutch words out loud. This study was conducted in collaboration between the AI lab at Donders Institute for Brain, Cognition, and Behaviour and the Medical Centre in Utrecht. Developing clinical applications or enhancements of cognitive functions in humans is highly interdisciplinary. The study reveals several findings that could help the development of BCI technology. For instance, Individual word decoding in reconstructed speech achieves 92-100% accuracy. The findings in this study pave the way for the development of advanced communication tools.
Future studies and applications
Between five or ten years from now Berezutskaya hopes to demonstrate this technology with the help of an implant in people who cannot speak but there are a lot of steps to be taken first. Berezutskaya: “This technology is highly interdisciplinary. For instance, there is a lot of hardware work to be done on the implants and in the computational processing and interaction with the users. At least now we are getting closer to providing the basic algorithms that make it possible. We have some clinical trials planned and think it is realistic to demonstrate our technology in actual people who cannot speak in the next 5 – 10 years. That would be a major step for us.”
Marcel van Gerven, chair of the AI department of the Donders Institute and co-author of the paper: ‘This study is a step on the way to the ultimate goal of decoding inner speech. Next to that, the developed approach may provide new insights into brain function. However, we still need to take a lot of steps. Especially from saying words out loud in people who are not paralyzed, as in this study, to decoding internal speech in those who cannot speak. But we are confident we can do so by making use of the latest neurotechnology and AI. The scientific community in the Netherlands plays a key role in this area with grass-roots initiatives such as NeurotechNL and research programs such as The Dutch Brain Interfaces Initiative funded by the Gravitation program of NWO.
The success of this study emphasizes the importance of optimizing decoding strategies to achieve the best possible results. Furthermore, the direct reconstruction of speech from neural signals offers a glimpse into the future of BCI technology, where seamless communication could become a reality for those who have long been silenced by paralysis.
For more information, read the full paper ‘Direct speech reconstruction from sensorimotor brain activity with optimized deep learning models’ published in the Journal of Neural Engineering.