手机位置和朝向无关的活动识别技术研究

Research on Activity Recognition Technique of Smart Phone Position and Orientation Independent

文中针对基于智能手机的活动识别中如何消除手机位置和朝向的影响问题展开研究。首先,针对手机放置位置问题,提出了先识别手机位置再进行活动识别的方法,其特点是能够基于特定的手机位置选取最佳的活动识别模型,进而消除手机位置对活动识别的影响;其次,针对手机朝向影响问题,对传感器数据使用绝对值和简单移动平均线处理的方法,消除手机朝向对手机位置识别和活动识别的影响;最后,基于Android平台开发了一个传感器数据收集工具对传感器数据进行收集,设计了两组实验对上面提出的方法进行实验验证。结果表明,文中提出的方法能够很好地消除手机放置位置和朝向对活动识别的影响,活动识别的准确率能达到87.89%。

The research aims at how to eliminate the influence of smart phone＇ s position and orientation variation at human activity recog- nition. Firstly, a method is put forward which recognizes the phone position and human activity followed based on the specific position. The feature is to select the best activity recognition model based on specific position for elimination of influence of mobile position on ac- tivity recognition. Secondly, in order to decrease the impact of mobile phones on the phone toward the sensor data, the absolute value and a simple moving average method is applied to process the sensor data. Finally ,to prove the theory ,a sensor data collecting tool has been developed in the Android platform, which is used to collect sensor data in different position and orientation of mobile phone, and two ex- periments have been conducted based on the theory and data collected by the tool. The results show that the presented method can effec- tively eliminate the influence of the smart phone＇ s position and orientation on the activity recognition,and the activity recognition accura- cy can reach 87.89%.