GLips (German Lips)

GLips（德语唇读数据集）是由汉堡大学信息学院知识技术系创建，包含250,000个来自黑森州议会演讲者的公开视频，用于词级唇读研究。每个视频时长1.16秒，格式与英语LRW数据集相似，便于跨语言转移学习研究。数据集创建过程涉及视频采集、多模态处理和面部检测等步骤。GLips适用于增强听力障碍支持、嘈杂环境中的通信、提高自动语音识别系统准确性等领域，旨在解决语言交流中的信息缺失问题。

GLips (German Lip-Reading Dataset) was developed by the Department of Knowledge Technologies, School of Informatics, University of Hamburg. It contains 250,000 publicly available videos of speakers from the Hessian State Parliament, and is intended for word-level lip reading research. Each video has a duration of 1.16 seconds, and its format is similar to that of the English LRW dataset, which facilitates cross-lingual transfer learning research. The creation of the GLips dataset involves procedures such as video collection, multimodal processing, and face detection. GLips is applicable to fields including enhanced support for hearing-impaired individuals, communication in noisy environments, and improving the accuracy of automatic speech recognition systems, with the aim of addressing the issue of information loss in language communication.