基于规则的身体动作快速识别方法

Fast Recognition Method of Body Movements Based on Rules

为了提升身体动作的识别效率,减少身体动作中整体位移对识别效果的影响,提出了一种基于规则的动作快速识别方法。首先,建立简化的关节点模型,并以髋关节中心点建立投影坐标系;其次,定义身体关节夹角和中心参数,设计动作判别函数,基于实验数据建立判别规则库与动作分类库;最后,将实时采集的关键帧数据与规则进行匹配,实现动作的快速识别。通过开展多数据集、多方法的比较实验,得到的平均识别准确率为90.72%,识别时间为0.996 s,识别效率优于其他方法。

In order to improve the recognition efficiency of body action and reduce the influence of the whole displacement on the recognition effect,a fast recognition method based on rules was proposed.Firstly,the simplified joint point model was established,and the projection coordinate system was established based on the center point of the hip joint.Secondly,the joint angle and the center parameters of the body were defined,and the action discrimination function was designed,and the discrimination rule base and action classification base were established based on the experimental data.Finally,the real-time key frame data were matched with the rules to realize the fast action recognition.The average recognition accuracy of this method is 90.72%,the recognition time is 0.996 s,and the recognition efficiency is better than other methods.