Digital News

　　Brain-Computer Interface (BCI) technology establishes a direct communication pathway between the brain and external devices by detecting and regulating brain activity. This has created a revolutionary form of human-computer interaction, turning “thought-based communication” from science fiction into reality. In recent years, with rapid advancements in technology, many countries have conducted experimental research in related fields, particularly achieving significant breakthroughs in speech BCIs. These include real-time communication through “brainwave-to-speech” technology for stroke patients with paralysis, brain-controlled robotic arms for writing Chinese characters, and improvements in the quality of life for Amyotrophic Lateral Sclerosis (ALS) patients... This emerging technology is not only enabling individuals with speech impairments to communicate with the world but also providing new possibilities and solutions for treating neurological diseases and other related conditions.

　　Real-time Conversion of Brain Activity into Speech

　　The brain is a powerful yet isolated organ, securely protected by the skull, and responsible for processing information related to sensations, emotions, memory, decision-making, and movement. Information flows between the external world and the brain through our body’s biological information interfaces, namely, our sensory organs and nervous system. Thanks to advancements in modern technology, humans can detect brain activity signals, decode the information they carry, and use this data to bypass the muscular system, thus directly controlling external devices. Essentially, this establishes an artificial information interface between the brain and the external world, which is known as BCI technology.

　　Speech BCIs, as a specific application of BCI technology, work by directly detecting brain activity, particularly extracting speech-related signals from the motor control areas of the brain. These signals are then decoded to extract the speech content, which is used to control speech synthesis devices to “speak” the words that patients intend to say. Ideally, it would function like a real-time simultaneous interpretation system that not only accurately interprets people’s intentions and thoughts but also outputs natural language as quickly and faithfully as possible. To make this possible, scientists need to solve a series of technical challenges, including signal decoding, speech synthesis, and output delays.

　　With advancements in neuroscience and engineering technology, multiple research projects worldwide are advancing the rapid iteration of speech BCI technology from different dimensions, promising to usher in a new stage of medical applications featuring “millisecond-level decoding + natural dialogue.” In March this year, China’s independently developed “Beinao No.1” semi-invasive system completed its third human implantation. With its flexible, high-density electrodes, the system enables synchronous signal collection across 128 channels, allowing ALS patients with aphasia to regain their ability to communicate while reducing surgical trauma risks. Recently, a research team at the University of California, Davis, unveiled a new speech BCI system. The team implanted a 256-channel microelectrode array into the brain of a 45-year-old male patient with aphasia caused by ALS. Using deep learning algorithms, they captured relevant signals from his brain, thereby decoding the words he intended to say. The system can capture brainwave signal features every 10 milliseconds, enabling near real-time decoding of sounds that aphasia patients attempt to produce. It can also display tonal variations and hum a sequence of notes across 3 different pitches, making overall expression more natural and fluent.

　　Artificial Intelligence (AI) Algorithms Are the Key to Technological Breakthroughs

　　Integrating and utilizing advanced AI models is crucial for BCIs to decode brain neural signals and generate natural language output.

　　In recent years, research institutions worldwide have successively released the latest developments in this field. The team from the University Medical Center Utrecht in the Netherlands and Radboud University optimized deep learning models to convert the neural activity in the sensory-motor cortex into recognizable speech in real time. This model achieves an impressive classification accuracy of 92% to 100% for individual words, while also preserving the tone and timbre of the synthesized voice. Similarly, the research team at the University of California, Davis, developed a deep learning model that uses audio recordings from the patient before the onset of aphasia to train the AI algorithm. This enables the model to synthesize and produce speech that closely mirrors the patient’s original voice.

　　Chinese has 418 syllables and four tones, which pose greater challenges for developing neural encoding and decoding mechanisms and information processing methods tailored to Chinese language characteristics, compared to languages such as English. A collaborative team from the Huashan Hospital Affiliated to Fudan University, ShanghaiTech University, and Tianjin University has developed a Chinese-oriented speech BCI. This multi-stream neural network model simultaneously decodes both tones and syllables, achieving a classification accuracy of up to 76% for tone and syllable decoding, and 91% for single-character decoding in individual subjects.

　　These research breakthroughs have provided a strong foundation for the practical application of speech BCIs. Looking ahead, the biggest challenge may lie in decoding intentions and meaning. Current research primarily focuses on decoding speech motor commands from the cerebral cortex that controls speech production. However, a significant number of aphasia patients have difficulty forming coherent sentences because the encephalic region responsible for language organization is damaged, rather than the area controlling vocalization. To address this, it will be necessary to directly record signals from a higher-level cerebral cortex involved in language processing to decode the patient’s intent and combine AI technologies, such as large language models, to generate the appropriate language expressions. As of now, decoding complex intentions within the brain is still in the early stages of research. However, there is hope that future speech BCIs will achieve further breakthroughs to realize true “thought-to-speech”

　　Expected to Revolutionize the Treatment of Neurological Diseases

　　In the medical field, BCI technology can not only help aphasia patients regain their language abilities, but also holds promise for triggering more revolutionary changes in the treatment of neurological injuries or diseases.

　　For instance, researchers from the Swiss Federal Institute of Technology in Lausanne and Lausanne University Hospital have developed a brain-spinal cord interface. This system decodes the brain’s motor control signals and stimulates the spinal cord areas responsible for walking, enabling paralyzed patients to regain the ability to walk. The system has been operating stably for over a year since its implantation, allowing patients to use it independently at home without the need for frequent recalibration.

　　A collaborative team from Fudan University and the Chinese Academy of Sciences has recently developed the world’s first visual prosthesis capable of capturing both visible light and infrared radiation. When implanted into the retina, the device replaces the photoreceptor cells, converting incoming light signals into electrical signals. These signals then activate the ganglion cells in the retina, transmitting visual information to the brain. This technology has enabled blind experimental animals to regain the ability to perceive both visible light and infrared radiation. It holds significant potential for future breakthroughs in the treatment of retinal diseases.

　　Additionally, BCIs can precisely regulate the activity of specific targets in the brain through implanted electrodes or non-invasive techniques, such as transcranial electrical or magnetic stimulation. Successful examples of the former include the use of deep brain stimulation to treat Parkinson’s disease, while the latter is currently being extensively researched for treating a wide range of brain diseases, from major depression to Alzheimer’s disease.

　　However, BCI technology still faces numerous challenges that urgently need to be addressed. Implanted BCIs require further validation to ensure their long-term stability and safety within the body, further reduce the trauma caused by electrode implantation, and improve the accuracy and operation stability of neural signal decoding. Additionally, neuroscience research needs to uncover more about the brain’s information processing mechanisms and patterns, enabling BCIs to interact with the brain more effectively.

　　Additionally, since BCI technology directly involves the detection and intervention of brain activity, its future development must carefully address potential risks related to ethics, privacy, and data security. These issues have already attracted significant attention from international organizations such as the United Nations, as well as from relevant regulatory authorities in China. Ethical guidelines, standards, and regulations related to BCI research and applications are gradually being refined to ensure that the technology and its applications develop in a healthy and sustainable manner, ultimately benefiting all of mankind.