1 Sora技术带来的冲击和思考
2 脑信号视觉重构技术
2.1 对人脑视觉表征的认识
2.2 神经解码与视觉重构技术的演进
2.2.1 基于fMRI数据的视觉重构
图4 不同视频/音频对应的fMRI激活区域图案Fig.4 FMRI activation area patterns corresponding to different video/audio |
图5 Miyawaki等人用fMRI数据对简单图形的重构结果Fig.5 Reconstruction results of simple graphs using fMRI data by Miyawaki et al |
2.2.2 基于脑电信号的视觉重构
2.3 立体定向颅内脑电图(SEEG)对视觉认知研究的帮助
图9 应用颅内电极可覆盖枕叶及周围相关视觉认知皮质,在视觉认知过程中,准确反映不同脑区的高频信号随时间的变化,具有高时空分辨率特性Fig.9 The intracranial electrodes can cover the occipital lobe and the surrounding visual cognitive cortex, accurately reflect the high-frequency signals in different brain regions during the visual cognitive process with time, and have the characteristics of high spatial and temporal resolution |
3 视觉认知引导的孪生AI深度网络
3.1 光电耦合性脑信息数据智能计算模式分布转化
3.1.1 视觉认知脑区功能定位
3.1.2 仿生网络结构设计
3.2 视觉认知引导的AI深度网络构建与验证
3.2.1 视觉深度网络构建
3.2.2 视觉重构有效性验证
图12 实验场景重构。a)为真实的实验场景,图中的女士并没有看向患者,而图c)中患者大脑信号的视觉重构结果显示,患者在大脑中认为这位女士是看向他的,故而为视觉重构技术重构了患者的幻觉内容Fig.12 The reconstruction of experimental scenarios. Figure a) is the real experimental scenario, in which the lady is not looking at the patient. However, the visual reconstruction result of the patient's brain signals in Figure c) shows that the patient believes in his brain that the lady is looking at him, so the visual reconstruction technology reconstructs the hallucination content of the patient |
中国指挥与控制学会会刊 