Tiny Brain-Machine Interface Chip Achieves High Text Conversion Accuracy

Swiss researchers at EPFL have developed a tiny brain-machine interface (BMI) chip, smaller and more efficient than Elon Musk's Neuralink. This micro BMI (MiBMI) measures just 8 square millimeters, contrasting with Neuralink's 23 x 8 millimeters. It's designed to read brain activity and convert it into text on a screen.

The MiBMI operates with minimal power, is minimally invasive, and processes data in real-time. Unlike Neuralink, which requires 64 electrodes and external processing, the MiBMI uses a compact, integrated system.

Key to its success is a unique neural code (DNC) system that simplifies data. Instead of processing thousands of bytes per letter, it handles just 100 bytes per DNC, a shorthand for each letter. This innovation boosts efficiency and reduces training time for users.

Currently, the MiBMI decodes 31 characters, a record for integrated systems, with potential to expand to 100. It aims to aid communication for those with severe motor disabilities, like ALS patients.

Researchers are also exploring its use in speech decoding and movement control, aiming for a versatile BMI tailored to various neurological conditions.

The breakthrough is detailed in the IEEE Journal of Solid-State Circuits.