中文手语识别数据集

本数据集整合了三个高质量中文手语公开资源：1）CE-CSL（https://www.selectdataset.com/dataset/afe9b23529e029d45767448f72b5dfc2）包含5,988段真实场景连续手语视频，覆盖70+复杂背景，适用于复杂环境下的连续手语识别；2）CSL500（https://pujunfu.github.io/openresources/cslr-dataset-2015/）由50名 signer 录制500个词汇，提供RGB、深度图与3D骨骼关节点，专用于孤立词识别，仅限科研使用并需签署协议；3）SpreadTheSign中文手语词汇集（https://spreadthesign.com/zh.hans.cn/search/）通过网络爬取，涵盖基础与日常词汇，适用于词汇级识别与教学研究。衷心感谢哈尔滨工程大学、中国科学技术大学VSLRG团队及SpreadTheSign平台的开放共享，为手语识别研究与聋健无障碍沟通提供宝贵支持。

This dataset integrates three high-quality open-source resources for Chinese sign language:
1) CE-CSL (https://www.selectdataset.com/dataset/afe9b23529e029d45767448f72b5dfc2) contains 5,988 real-world continuous sign language videos covering over 70 complex backgrounds, which is applicable to continuous sign language recognition in complex environments;
2) CSL500 (https://pujunfu.github.io/openresources/cslr-dataset-2015/) includes 500 vocabulary terms recorded by 50 signers, and provides RGB images, depth maps and 3D skeletal joint points. It is specifically designed for isolated word recognition, and is for research use only and requires a signed agreement;
3) SpreadTheSign Chinese Sign Language Vocabulary Dataset (https://spreadthesign.com/zh.hans.cn/search/), which is crawled from the web, covers basic and daily vocabulary, and is suitable for vocabulary-level recognition and teaching research.

We sincerely thank Harbin Engineering University, the VSLRG team of the University of Science and Technology of China, and the SpreadTheSign platform for their open sharing, which provides valuable support for sign language recognition research and barrier-free communication between deaf and hearing people.

搜集汇总

背景与挑战

背景概述

该数据集整合了三个公开的中文手语资源，包括CE-CSL连续手语视频、CSL500孤立词多模态数据和SpreadTheSign词汇视频，覆盖孤立与连续手语识别任务。它适用于手语识别、翻译、教育和无障碍人机交互等研究领域。