The Chinese startup NeuroXess has achieved two groundbreaking advancements in brain-computer interface (BCI) technology, providing life-changing assistance to a patient with brain damage. The first milestone involves real-time decoding of thoughts into speech, while the second enables remote control of a robotic arm through mere thought.
In an experiment conducted in August 2024 at Huashan Hospital, a 21-year-old patient suffering from epilepsy and motor cortex damage received an implant of a flexible, high-bandwidth BCI device with 256 channels. This interface successfully captured and analyzed high-frequency brain signals in the 70–150 Hz range, offering insights into motor and sensory activities.
After training with a neural network model, the patient achieved speech decoding from brain signals with a latency of less than 60 milliseconds. Within 48 hours, she began playing table tennis and video games, and after two weeks, she mastered controlling smartphone applications such as WeChat and Taobao.
The interface further empowered the patient to manage smart home systems and a wheelchair, significantly enhancing her daily life. A notable triumph was decoding speech from brain signals—an especially challenging feat for the tonal and logographic structure of the Chinese language.
NeuroXess achieved an unprecedented 71% accuracy in decoding speech from 142 common Chinese syllables, with a latency of less than 100 milliseconds, setting a national record. The startup showcased its technology by enabling the patient to operate a robotic arm, perform gestures, and communicate phrases like “Happy New Year, 2025!” She also interacted with a digital avatar linked to an AI model.
Tao Hu, the founder of NeuroXess, emphasized that the company’s method minimizes the risk of brain tissue damage by operating on the cortical surface. Looking ahead, the startup envisions deploying this technology to assist individuals with speech and movement limitations, including patients with amyotrophic lateral sclerosis (ALS), paralysis, and stroke-related impairments.
