基于手形的骨骼手语库构建研究

Study of Skeleton Sign Language Based on the Hand Shape

摘要: 手语词库的构建是手语识别与合成的基础。本文通过分析中国手语的特点，提出了一种基于扩展Kinect人体骨骼模型，以手形组合构建骨骼手语库的方法。通过采集手语表达过程当中的关键状态，记录手形、位置和方向等数据，形成手语词关键帧数据，实现了一个中文骨骼手语编辑系统，有效的提高手语词汇的合成效率。生成的数据通过关键帧插值算法可实现手语词的动态组合，为基于骨骼手语库的手语识别及合成奠定了基础。

Abstract: Building of sign language thesaurus is the basis of sign language recognition and synthesis. Through analyzing the characteristics of Chinese sign language, this paper proposes a human skeleton model which is based on expansion Kinect, and built skeleton sign language thesaurus with a combination of hand shape. It achieves an editing system of Chinese skeleton sign language by acquisition of the key state in sign language expression, then recording data of hand shapes, positions and orientations, and forming key frame data of sign language at last. It effectively improves synthesis efficiency for sign language. Data generates from which can make dynamic combination through the key frame interpolation algorithm, and lays the foundation for the recognition and synthesis of sign language thesaurus.