基于多传感器数据融合的人体运动模式识别研究  

Research of Human Motion Recognition Based on Multi-Sensor Data Fusion

作　　者：张毅[1] 任同辉[1] 罗元[2] 

机构地区：[1]重庆邮电大学信息无障碍工程研发中心,重庆400065 [2]重庆邮电大学光纤通信技术重点实验室,重庆400065

出　　处：《计算机测量与控制》2016年第1期303-307,共5页Computer Measurement &Control

基　　金：科技部国际合作项目(2010DFA12160);国家自然科学基金(51075420)

摘　　要：针对单一传感器在人体运动姿态监测中误差较大的问题,提出了一种基于加速度传感器和陀螺仪数据融合的人体运动模式识别方法;该方法使用陀螺仪输出的人体运动信息对加速度传感器采集到的姿态角信息进行修正,采用卡尔曼滤波算法实现多传感器信息的融合,有效提高了姿态角度测量的准确度;根据人体日常的活动状态构建了基于人体姿态角度特征的隐马尔可夫模型实现人体运动模式的识别;实验表明,该方法比采用单一传感器方法识别的准确率高,可以有效区分不同的日常活动行为。To deal with the problem of large deviation of single sensor in the human motion attitude recognition monitoring, an human motion recognition method based on data fusion between acceleration transducer and gyroscope is proposed. In the method, the attitude angle output by acceleration transducer is modified by the human motion information output by gyroscope, multi--sensor data fusion is achieved by using Kalman filtering algorithm to effectively improve the accuracy of attitude angle measurement. According to the daily movement of hu man body, Hidden Markov Model based on attitude angle feature is built for human motion recognition. Experiments show that, the proposed method performs better with higher accuracy rate than method using single sensor, which can effectively distinguish between different daily activities.

关 键 词：运动模式识别 卡尔曼滤波 数据融合 姿态检测 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]