基于骨骼三维信息结合隐马尔科夫模型人体动作识别方法

Human Motion Recognition Method Based on Bone 3D Information Combined with Hidden Markov Model

摘要: 本文在人体动作描述采用关键骨骼点的深度三维坐标，通过各个关节点的位置坐标关系来描述人体动作。进一步为每一类人体动作训练一个深度信息隐马尔科夫模型，从一定程度上解决了人体动作部分遮挡问题，对训练样本，参照得到的数据库完成编码，通过隐马尔科夫模型训练得到相应参数模型，通过训练得到的模型进行验证和测试，实验数据结果表明该方法具有精确率高的优点。

Abstract: In this paper, the depth three-dimensional coordinates of key bone points are used in the description of human action, and the human action is described through the position coordinate relationship of each joint point. Further, a depth information hidden Markov model is trained for each kind of human action, which solves the problem of partial occlusion of human action to a certain extent. The training samples are encoded with reference to the obtained database, the corresponding parameter model is obtained through hidden Markov model training, and the model obtained through training is verified and tested. The experimental results show that this method has the advantage of high accuracy.