情景知觉中的信息界定:从主观预期到“惊奇”算法

情景信息加工是情景知觉研究领域内的核心问题。然而,不同的研究者对"情景信息"的界定并不相同。在过去80多年的时间里对情景信息的界定先后经历了研究者主观确定、观察者评估、基于情景物理特征的计算模拟等方法。Itti和Baldi等人通过比较先验概率和后验概率之间的差异,提出了"惊奇"理论。文章介绍了真实情景知觉研究中"惊奇"理论的应用及进展,并指出"惊奇"算法的拓展及界定、情景知觉过程中的注意转换与"惊奇"的关系,以及社会情景知觉中的"惊奇"应用可能是未来相关领域需要进一步探索的问题。

The Definition of Information in the Scene Perception:From Subjective Expectancy to Surprise Arithmetic

It is a key issue on how to process scene information in the high-level scene perception fields. However, the information definition varies from dif ferent researchers. During the past eight decades, the development of scene information definition experienced several approaches as researches′ subjective affirmation, viewers′ evalua-tion and computing simulations based on scene physics characteristic. Itti and Baldi (2005, 2009) compared the different distribution between the prior probability and posterior probability , and developed a model of surprise. This paper introduced the application of ″surprise″ to the scene perception field, and analyzed the development of the relevant research progress. Finally, it noted that the extension of surprise computation, the definition of sur-prise, attention switching during scene perception and application of surprise in complex social scenes that need to be further explored.