LRW-1000
收藏资源简介:
LRW-1000是由中国科学院智能信息处理重点实验室创建的大规模野外唇读数据集,包含1000个普通话词汇类别,总计718,018个样本,覆盖超过2000名说话者。数据集旨在模拟实际应用中的自然变异性,包括不同的语音模式和成像条件。创建过程涉及从电视节目中自动收集视频,并结合手动标注和额外过滤以确保数据一致性。该数据集适用于唇读技术的研究与开发,特别是在提高多说话者和嘈杂环境下的语音识别性能方面。
LRW-1000 is a large-scale wild lip-reading dataset developed by the Key Laboratory of Intelligent Information Processing, Chinese Academy of Sciences. It encompasses 1000 Mandarin vocabulary categories, with a total of 718,018 samples covering over 2000 speakers. The dataset is designed to simulate natural variability in real-world applications, including diverse speech patterns and imaging conditions. Its creation involves automatically collecting video materials from television programs, combined with manual annotation and additional filtering to ensure data consistency. This dataset is suitable for research and development of lip-reading technologies, particularly for improving speech recognition performance in multi-speaker and noisy environments.




