模板
视觉搜索
计算机科学
对偶(语法数字)
模板匹配
匹配(统计)
模式识别(心理学)
人工智能
数学
统计
图像(数学)
文学类
艺术
程序设计语言
作者
Haomin Lian,Sen Liu,Hua Chen,Zuomin Wang,Xiaowei Che
摘要
Abstract Neural activation of the target representation (template) facilitates attentional guidance, allowing humans to effectively perform visual search. However, visual search is not always very effective, especially when searching for multiple templates. The reduced search efficiency under dual-target compared with single-target searches is known as the dual-target cost and might be caused by decreased precision, increased resource consumption, or the switch cost between activated templates. The activation of templates and the underlying mechanism of multitarget visual search were explored in this study. In Experiments 1 and 2, participants searched for one or two targets under different precision requirements or memory loads, respectively. The results showed that the precision requirement, rather than resource consumption, influenced the dual-target cost. The impact mechanism of precision requirement was explored in Experiment 3 by measuring ERPs reflecting attentional selection and memory matching. The sustained posterior contralateral negativity, which reflects memory matching, was smaller in the dual-target search compared with the single-target search, especially under low-precision requirement. The activation patterns of templates during the dual-target search were investigated in Experiment 4 using EEG decoding. Under low-precision requirement, the matched template was activated subsequent to the unmatched template, whereas under high-precision requirement, there was an overlap in the activation periods of the two templates during the template matching stage. These findings demonstrate that increasing the precision requirement of working memory keeps the activation of the template and promotes template matching. The dual-target cost might be attributed to the inappropriate template activation, which consequently hinders accurate matching with potential objects.
科研通智能强力驱动
Strongly Powered by AbleSci AI