Speak in Colombian Sign Language: A Dynamic LSC70 Database

The dynamic LSC70 database consists of a collection of images of Colombian Sign Language (LSC), created to support research in artificial intelligence for developing sign language recognition applications that promote the social inclusion of individuals with hearing disabilities. This database was developed with the participation of 70 volunteers and contains 47 signs in transitions of six frames. It comprises a total of 35,208 images, including signs of the alphabet, numbers, and common words. Additionally, the dataset is divided into three sections: LSC70AN, which includes full-body signs of the alphabet and numbers with a resolution of 640x480 pixels; LSC70ANH, which similarly covers alphanumeric signs but focuses on the dominant hand of the volunteers with a resolution of 120x120 pixels; and LSC70W, which encompasses common words used in basic conversations, such as greetings and identifications, also at 640x480 pixels. Each sign was recorded under uncontrolled lighting and clothing conditions using a cellphone and was later filtered to remove duplicates and enhance the quality of the database. The images are in JPG format and organized in folders by volunteer and sign, facilitating their use in research and the development of sign language recognition models.动态LSC70数据库是一套哥伦比亚手语（Colombian Sign Language, LSC）图像集，旨在为支撑手语识别应用开发的人工智能研究提供数据支持，助力听障群体的社会融入。该数据库由70名志愿者参与构建，涵盖47个六帧过渡状态的手语词汇，总计包含35208张图像，内容覆盖字母、数字及常用词汇的手语动作。此外，该数据集分为三个子部分：LSC70AN，包含分辨率为640×480像素的字母与数字全身手语数据；LSC70ANH，同样涵盖字母数字类手语，但聚焦志愿者的优势手，分辨率为120×120像素；LSC70W，包含基础对话中常用的常见词汇手语，如问候、身份识别相关表达，分辨率同样为640×480像素。所有手语动作均通过手机在光照、着装不受控的环境下录制，后续经去重处理以提升数据库整体质量。图像均采用JPG格式存储，并按志愿者与手语类别分文件夹组织，便于在手语识别模型的研究与开发中使用。