人体行为识别数据集研究进展

Research Advances on Human Activity Recognition Datasets

人体行为识别是计算机视觉领域的一个研究热点,具有重要理论价值和现实意义.近年来,为了评价人体行为识别方法的性能,大量的公开数据集被创建.本文系统综述了人体行为识别公开数据集的发展与前瞻:首先,对公开数据集的层次与内容进行归纳.根据数据集的数据特点和获取方式的不同,将人体行为识别的公开数据集分成4类.其次,对4类数据集分别描述,并对相应数据集的最新识别率及其研究方法进行对比与分析.然后,通过比较各数据集的信息和特征,引导研究者选取合适的基准数据集来验证其算法的性能,促进人体行为识别技术的发展.最后,给出公开数据集未来发展的趋势与人体行为识别技术的展望.

Human activity recognition is an important research field of computer vision with important theoretical value and practical significance. In recent years, a large number of public datasets have been created for evaluation of human activity recognition methodologies. This paper reviews the progress and forecast the future of public datasets for human activity recognition. First, the hierarchy and contents of the public datasets are summarized, and the public datasets are divided into four categories according to the characteristics and acquiring methods. Then, the four categories are described and analyzed separately. Meantime, the state-of-the-art research results and corresponding methods of the public datasets are introduced to researchers. By comparing the information and characteristics of each dataset, researchers can be guided in the selection of the most suitable dataset for benchmarking their algorithms, so as to promote the technology progress of human activity recognition. Finally, the future trends of the public datasets and the prospects of human activity recognition are discussed.