HTML5 Webook

16/64

14Data utilization and analytics platformCenter for Information and Neural Networks The Center for Information and Neural Networks (CiNet) conducts fundamental research integrating neuroscience and information and communications technology (ICT). It operates around a core of NICT, Osaka University, and the Advanced Telecommunications Research Institute (ATR), and engages in widespread collaboration with other universi-ties, research institutes, and enterprises.The human brain is the most complex information and communication appara-tus known to man. In our fourth Medium- to Long-Term Plan, we have set a goal of establishing technology able to measure brain activity related to cognition, percep-tion, and movement, and to encode and decode it efficiently. Our ultimate goal is to create a new generation of ICT that will help improve health and welfare. To achieve this, we are analyzing high-order brain information processing and applying it to tasks such as designing information processing architectures and discovering biomarkers. We are also promoting R&D on technologies to improve the physical, sensory, and social capabilities of individu-als. We are also conducting basic research to evaluate appropriateness and safety based on brain information technologies, and promoting basic technologies to infer human emotion and cognition, based on human responses to multi-sensory fluctu-ations and changes in brain data. Brain information decoding tech-nologiesBrain information decoding technolo-gies read details of perception and cogni-tion from brain activity. We expect that ba-sic technologies like the brain-machine interface will play an important role, but to make such technologies practical, it is nec-essary to read the complex and varied con-tent of perception produced in the real world. In earlier research, technologies to identify what is being viewed, or to infer the content of dreams, from brain activity have been developed, but they have not been able to visualize more than an extremely small subset of the complex and varied perceptions arising from the real world. This research has focused on language it-self as a means of expressing varied per-ceptions, and we have conceived a tech-nology that decodes perception details from brain activity by incorporating feature spaces of language in a brain-information decoder. We used this approach to analyze brain activity while watching advertise-ment videos, and were successful in infer-ring details of perception and cognition in-duced by the video, in the form of objects (nouns), actions (verbs), and impressions (adjectives), using a vocabulary of several tens of thousands of words (Fig.1). Technology to predict depres-sion using fMRIWe have established a technology able to predict tendencies toward depression, current and in the following year, from pat-terns of activity in the amygdala, which responds to comparison of one’s situation relative to another (social value orienta-tion). This has been reported in the journal Nature Human Behaviour. Specifically, 94 subjects performed the Beck Depression Inventory II, which is a test for depressive tendencies, and were then given a task called the Ultimatum Game while under-going an MRI scan. In the Ultimatum Game, one person has the role of proposer, and proposes a way to distribute a sum of Fig.1 : Decoding brain activity while watching video and estimating perception contentResearch and Development