mEBAL

我们介绍了mEBAL，这是一个用于眨眼检测和注意力水平估计的多模态数据库。mEBAL是一个数据库，最初在arXiv技术报告中介绍，然后被第22届ACM多模式相互作用国际会议接受。
眨眼频率与认知活动有关，已经提出了眨眼的自动检测器，用于许多任务，包括注意力水平估计，神经退行性疾病分析，欺骗识别，驾驶乏力检测或面部反欺骗。但是，该领域中的大多数现有数据库和算法仅限于仅涉及数百个样本和单个传感器 (例如面部摄像头) 的实验。
拟议的mEBAL在眨眼样本和可用的认知信息方面改善了以前的数据库。特别地，同时考虑三种不同的传感器: 近红外 (NIR) 和RGB相机来捕获面部手势，以及脑电图 (EEG) 带来捕获用户的认知活动和闪烁事件。

We introduce mEBAL, a multimodal database for blink detection and attention level estimation. Originally presented in an arXiv technical report, mEBAL has been accepted by the 22nd ACM International Conference on Multimodal Interaction.

Blink frequency is correlated with cognitive activity, and automated blink detectors have been proposed for a wide range of tasks including attention level estimation, neurodegenerative disease analysis, deception detection, driver fatigue monitoring, and facial anti-spoofing. However, most existing databases and algorithms in this field are limited to experiments involving only hundreds of samples and a single sensor, such as a facial camera.

The proposed mEBAL improves upon prior databases in terms of blink sample size and accessible cognitive information. Specifically, it incorporates three distinct sensors: near-infrared (NIR) and RGB cameras for capturing facial gestures, and an electroencephalogram (EEG) device for recording users' cognitive activity and blink events.

搜集汇总

背景与挑战

背景概述

mEBAL是一个多模态数据库，用于眨眼检测和注意力水平估计，通过近红外、RGB相机和脑电图三种传感器收集数据，样本数量和认知信息较现有数据库更丰富。该数据集由马德里自治大学于2020年发布。