Recently, researchers from several U.S. universities are working on a project. It is said that the mature performance of the project is that, according to the scanning of human brain activity by the equipment, it can accurately depict the human face image that people thought at the time. Perhaps this sounds incredible, but the technology will soon be applied to various fields. “It's like reading minds,†said Marvin Pei, a professor of psychology, cognitive science, and neurobiology at Yale University.
According to the physicist's organization network, Alan Corvin, a student of Professor Marvin Biao, wanted to know whether he could use the pattern of brain activity to construct a person's face. This task is extremely arduous because people's facial differences are much smaller than the differences between buildings, and the process of forming human faces in the brain involves many brain areas. “We need to come up with a face and the details involved are much more subtle than thinking of other things,†said Kevin.
With funding from the Yale Provost Office, Corwin and Bryce Cole, Associate Professor at New York University, conducted an fMRI scan of six volunteers to construct 300 different facial images. These images were repeatedly revised. Researchers call it "training."
They first built a statistical library with data. The statistical library can describe how the brain responds to a single human face. Six volunteers then watched several new faces while scanning with fMRI. The researchers entered the newly acquired fMRI data into the original statistical library model and reconstructed the faces that the volunteers saw. Kevin said that with the development of technology, the accuracy of reconstructing facial images will be higher and higher. He predicts that this method can also be used as a research tool, such as studying the response of children with autism to the face.
The core of fMRI is big data analysis. The principle is actually very simple. By stimulating the brain to repeat some kind of thinking, action or recall, using magnetic resonance imaging to obtain scan data to determine which brain regions have signal changes in the process, and then use statistical methods. Find out the relevance of the area, intensity, and these thoughts, actions, or memories of brain activity. Therefore, the more accurate the location of brain activity areas, the more samples in the database, the higher the accuracy. This project is an extension application of fMRI, which is theoretically feasible. However, to reach the realm of “reading the heart†requires breakthrough progress in scanning accuracy and sample size, which is far from a day's work.