人脑视觉系统具有非凡的信息处理能力，很多研究尝试利用神经信息编解码的方法，来揭示大脑视觉信息加工的机制。随着功能磁共振等脑成像技术的进步和人工智能技术的发展，我们已经可以建立被试在观看简单刺激时的脑活动与刺激内容间的信息编解码模型，并实现了视觉刺激重建。然而，由于脑信号的信息编码复杂性高、时空维度高、被试个体差异大、以及脑活动数据难以大量获取等问题，现有方法难以对自然视觉刺激引起的脑活动进行准确地语意解码和视觉重建。本研究将结合神经科学和人工智能技术研发具备高鲁棒性、适应性、准确性的神经信息编解码方法，从数据、算法和任务三个层面，研究基于多被试数据融合、对偶学习和多任务学习等方法的编解码技术，提高模型解码准确率。该研究不仅对深入了解人脑视觉信息编码机制具有重要科学意义，也可以帮助我们实现利用机器智能对人类视觉感知功能的模拟，提升计算机处理信息的能力，为类脑计算模型等人工智能研究提供帮助。

The visual system of human brain has extraordinary information processing capability. Its visual processing mechanism has been widely concerned by researchers. Many studies have attempted to use brain encoding and decoding methods to reveal the mechanism of brain visual information processing. With the advancement of neuroimaging techniques such as functional magnetic resonance imaging and the development of AI technology, we have been able to establish the quantitative mapping between brain neural activity and stimulus content when participants are watching simple stimuli, and achieve visual stimulation reconstruction. However, due to the high complexity of the brain function signal, high temporal and spatial dimension, the large variance of the individual subjects, and the small sample size, the existing methods are difficult to accurately decode the semantic information and reconstruct the natural visual stimulus. This study will combine neuroscience and artificial intelligence technology to develop a robust, adaptive, and accurate neural information encoding and decoding method. We will study multi-subject data fusion, dual learning based on data, algorithms, and tasks. And multi-task learning and other methods to improve the accuracy of model decoding. This research not only has important scientific significance for understanding the processing and coding mechanism of human brain visual information, but also be helpful for using machine intelligence to realize the simulation of human visual perception function and improve the ability of computer to process visual information, to promote brain inspired intelligence research.