EEG生物传感器有帮助AI读取你的心 引导机器人,汽车和更多的

全文内容 Full Text

Sep 12, 2026 10:55am Sensors Applications Magnetic / Electrical Medical / Healthcare Chemical / Biosensors --> Electroencephalography (EEG) systems rely on biosensors, individual metal discs that are placed on the scalp to detect tiny voltage fluctuations caused by the firing of brain neurons. They have been in medical use for more than a century and offer profound insights for medical diagnoses and, more recently, AI research. A steadily emerging field of this research is brain-to-computer interface (BCI) technology. EEGs fit into this work for non-invasive work, where companies such as g.tec, Emotive, Neuroelectrics and others are active. Elon Musk’s Neuralink company has won attention for implants of electrode threads into the cortex of the brain to help people with disabilities, but others in the invasive space include Synchron, Precision Neuroscience, Paradromics and Blackrock Neurotech. Some or all of these companies are working to productize their work, or already have produced dozens of portable wireless EEG products that people can buy for use in gaming, meditation, attention tracking and other purposes on the consumer side. Research-grade products are plentiful and far more costly. (More on that below.) Many researchers and companies now see BCI as just another sensor modality that will be very useful in developing robots. EEG signals can be fused with cameras, eye tracking, IMUs, muscle sensors, speech, tactile sensors and even sensors for temperature and other environmental inputs. What’s most interesting within these capabilities, which are directly reliant on biosensors in EEG systems, are how researchers are finding ways to detect what the human brain thinks, which can be applied to AI training. If AI in a computer or physical setting (robot) can be expected to collaborate naturally, it must accurately understand what a human wants, more than just inferring human intent from speech, actions or gestures. New research from the Korea Advanced Institute of Science and Technology (KAIST) is looking into using neural feedback through EEG applied to humans to help AI decide things like “that’s not what I meant” from human brain signals. With that insight, AI can then adjust its behavior in real time according to human goals. TechXplore described their work based on research published in IEEE Transactions on Cybernetics. AI can misunderstand human intent and the ambiguity can cause problems. For example, if a person picks up a cup, it might not be clear if they intend to drink from it or hand it off to someone. When an AI misunderstands human intent, users often have to correct it with additional commands or actions. The researchers focused on such “prediction error” signals that arise unconsciously in the brain. The team focused on finding the brain’s “that’s not what I meant” response when AI performed the wrong action or pursued the wrong goal. The researchers measured real-time EEG signals from people as they observed AI performing tasks. What they found is that the brain produced different signals, depending on whether the AI misunderstood the goal or chose the wrong method while pursuing the correct goal. They found distinctive brainwave patterns that appeared when both types of errors happened at the same time. They developed technology from EEG insights alone of how a person is interpreting the AI’s behavior even without a person saying “that’s wrong." The AI can recognize whether the human brain is seeing the goal is wrong or the method is wrong. They proposed an algorithm to feed the decoded brain signals back to the AI in real time, so it can correct its behavior. The researchers believe the technology could advance to robots in industrial settings or autonomous vehicles to quickly reflect human judgment. Or it could be extended to medical or rehab robots for patients who have difficulty speaking or moving. It isn’t clear from the research paper exactly how these EEG signals would be gathered from a