Cell：“读心术”来了！脑接机口实时解读“内心独白”，自带密码保护，防止隐私泄漏

Core Viewpoint - The article discusses a breakthrough in brain-computer interface (BCI) technology that allows for the decoding of internal speech, providing a promising tool for individuals with speech impairments [2][3]. Group 1: Technology Overview - The newly developed BCI system can decode internal speech by relying on signals from the supramarginal gyrus, a brain area closely related to language [2]. - This BCI can achieve a decoding accuracy of 74% for imagined speech, comparable to previous systems that required vocalization [4][8]. - The system incorporates a password protection feature, ensuring that decoding only occurs when the user thinks of a pre-set password, thus safeguarding privacy [3][4]. Group 2: Research Methodology - The research involved four participants with speech difficulties, including one stroke survivor and three individuals with motor neuron disease, who were instructed to either attempt to speak or imagine speaking specific words [6]. - The study found that both attempted speech and internal speech originate from the same brain region, producing similar neural signals, albeit with weaker signals for internal speech [7]. - An AI model was trained using the collected neural data to identify phonemes, enabling real-time assembly of words and sentences from a vocabulary of 125,000 words [9]. Group 3: Key Findings - The study concluded that there are shared representations of attempted speech, internal speech, and perceived speech in the motor cortex [15]. - The BCI can decode general sentences and improve user experience, while also interpreting aspects of personal internal speech during cognitive tasks like counting [15]. - The high-fidelity solution effectively prevents the unintended decoding of personal internal speech, addressing privacy concerns [15].