Study of brain cognitive computational interface for user's mind reading

为了实现利用大脑认知过程模型来预测行为的目的，我们构建了一个真实的世界问题中的行为决策模型，并通过行为实验对模型进行了评价。在实验中，我们进行了两个平行的研究活动。一种是在实验室中使用游戏任务建立工作模型的精确行为分析。另一个是对实际汽车驾驶场景的分析，对模型进行评价。我们将受试者的行为建模为行动决策策略的动态切换，并将该模型在计算机模拟中的行为与典型受试者和ASD受试者的实际行为进行了比较。结果表明，我们的大脑过程的存在，设置一个参数的行动决策过程取决于对手的游戏。在后一项研究中，我们测量了眼动的司机，并给出了日期的标签描述，使计算机模拟的眼动在一个伪驾驶环境。我们尝试了在计算机模拟中重建眼动行为，并对模拟模型进行了迭代改进。因此，我们开发了一个模型的驱动程序内部过程，其中包括一个风险依赖的行动因素，除了传统的因素的显着性，对象检测和驾驶计划的因素。

For the realization of behavior prediction using a model of brain cognitive process that can estimate internal process of other's mind, we constructed a model of action decision in a real world problem and evaluated the model through behavioral experiment. In the experiment, we conducted two research activities in parallel. One is a precise behavioral analysis using a game playing task in the laboratory to establish the working model. Another is an analysis of actual motor car driving scene to evaluate the model.In the former study, we used a competitive game "Matching pennies". We model the subject's behavior as a dynamic switching of action decision strategies, and compared the behavior of the model in a computer simulation with the actual behavior of the typical subjects and the ASD subjects. The results suggested us an existence of brain process that sets a parameter of the action decision process depending the opponents of the game.In the latter study, we measured eye movements of the drivers and gave tag description of the date to enable a computer simulation of the eye movement in a pseudo driving environments. We tried a reconstruction of the eye movement behavior in the computer simulation and iterated improvement of the simulation model. As the result, we developed a model of driver internal process that includes a risk dependent action factor in addition to the conventional factors of saliency, objects detection and driving plan factors.