盲人通过听觉与外界环境进行有效地交互，大脑的补偿可塑性和跨通道的信息处理让盲人的视觉皮层参与听觉信息的加工。然而盲人是如何通过听觉双通路模型加工听觉内容信息和空间信息的机制仍不清楚。本项目利用功能磁共振成像采集先天盲、早盲和正常被试在听觉识别和定位范式下的脑影像数据，基于脑区激活特征和脑区间功能连接特征，运用多变量模式分析方法解码听觉刺激内容和空间信息的皮层表征，研究听觉刺激内容信息和空间信息的加工机制。同时，利用动态因果模型分析方法，构建人脑处理内容信息和空间信息的最优网络模型，揭示人脑对听觉信息的时空处理机制。最后结合深度自编码模型和隐藏马尔科夫模型设计新的深度学习模型，发掘听觉信息加工过程脑区间的动态变化关系，从网络建模的角度探索听觉信息的加工机制。本项目从脑信息解码的角度探索盲人的听觉信息加工机制，有助于促进神经康复医学的发展，为开发基于听觉通路的盲人群体视觉补偿提供科学依据。

Blind individuals interact efficiently with environment through perception of auditory information. The occipital visual cortex is involved in the auditory information processing because of compensatory plasticity and cross-pathway processing in blind. However, it remains unclear how the auditory content and spatial information is processed by dual-pathway model in congenitally blind. In this project, we use functional magnetic resonance imaging to study the congenitally and early blind and matched sighted humans with an auditory identification and location paradigm, and make the multivariate pattern analysis (MVPA) to decode the cortical representation of auditory content and spatial information processing based on the brain activation intensity and functional connectivity between specific regions. Meanwhile, dynamic causal modelling analysis is performed to select the best optimized model during the processing of the auditory content and spatial information in blind subjects, which can elucidate the spatial-temporal mechanism of auditory information processing and verify the role of primary visual cortex in the model. We finally try to propose a novel methodological architecture that combines the Deep Auto-Encode and Hidden Markov Model, in order to discover the hierarchical non-linear relationships between regions. This project tries to explore the auditory processing mechanism in blind with decoding methods, which will promote the development of neural rehabilitation and make a great contribution to developing the prosthetic vision system via auditory pathway for blind people.