机器学习技术被广泛应用于人体运动合成，取得了很多研究成果，这得益于运动捕获技术的成熟和使用。运动捕获技术使数据驱动的方法成为人体运动合成的主流，而机器学习技术能够有效地处理人体运动数据这类维数高且在时间和空间两个维度都具有较大相关性的数据。文中以机器学习方法为主线，结合各类方法的特点，对人体运动合成中的应用进行总结和归纳；同时，还展望了人体运动合成和运动数据处理技术未来的发展趋势。-Machine learning techniques are widely used in human motion synthesis, made a lot of research results, thanks to motion capture technology maturity and use. Motion capture technology enables data-driven approach into the mainstream of human motion synthesis, and machine learning techniques to deal effectively with human motion data of such high dimension and two dimensions in space and time are associated with large data. In this paper, the main line of machine-learning method, combining the characteristics of various methods of synthesis of human motion for summary and induction　the same time, prospects of human motion synthesis and motion data processing technology development trend of the future.