计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (17): 153-159.DOI: 10.3778/j.issn.1002-8331.1603-0279

• 模式识别与人工智能 • 上一篇    下一篇

基于多传感器融合的动态手势识别研究分析

马正华1,李  雷2,乔玉涛2,戎海龙3,曹海婷2   

  1. 1.常州大学 研究生部,江苏 常州 213164
    2.常州大学 信息科学与工程学院 数理学院,江苏 常州 213164
    3.常州大学 城市轨道交通学院,江苏 常州 213164
  • 出版日期:2017-09-01 发布日期:2017-09-12

Dynamic gesture recognition research and analysis based on multi-sensor fusion

MA Zhenghua1, LI Lei2, QIAO Yutao2, RONG Hailong3, CAO Haiting2   

  1. 1.Graduate Division, Changzhou University, Changzhou, Jiangsu 213164, China
    2.School of Information Science & Engineering, School of Mathematics & Physics, Changzhou University, Changzhou, Jiangsu 213164, China
    3.School of Urban Rail Transit, Changzhou University, Changzhou, Jiangsu 213164, China
  • Online:2017-09-01 Published:2017-09-12

摘要: 研究利用三类传感器(表面肌电仪、陀螺仪和加速度计)信号的特点进行信息融合,提高可识别动态手势动作的种类和准确率。将动态手势动作分解为手形、手势朝向和运动轨迹三个要素,分别使用表面肌电信号(sEMG)、陀螺仪信号(GYRO)和加速度信号(ACC)进行表征,利用多流HMMs进行动态手势动作的模式识别。对包含有5个运动轨迹和6个静态手形的识别实验结果表明,该方法可以有效地从连续信号中识别动态手势,三类传感器组合使用获得的全局平均识别率达到92%以上,明显高于任意两个传感器组合和仅采用单个传感器获得的平均识别率。实验表明该方法是一种有效的动态手势识别方法,并且相较于传统的动态手势识别的方法更具有优势。

关键词: 手势识别, 表面肌电信号(sEMG), 加速度信号(ACC), 陀螺仪信号(GYRO), 多流隐马尔可夫模型(MHMMs)

Abstract: This paper studies using three kinds of sensors(surface electromyography, gyroscope and accelerometer) singnal characteristics for information fusion. The purpose is to improve the types and accuracy of recognizable dynamic hand gestures. Dynamic hand gestures are divided into three elements: hand shape, gestures towards and trajectory. The three elements are respectively represented by surface Electromyogram signal(sEMG), Gyroscope signal(GYRO) and Acceleration signal(ACC). Multi-streaming HMMs are used for pattern recognition of dynamic hand gestures. In the recognition experiments that contain five motion trajectories and six static hand shapes, the results show that this method can effectively identify dynamic hand gestures from the continuous signals. The global average recognition rate obtained from the composite use of three kinds of sensors is higher than 92%. It is significantly higher than the average recognition rate that obtained from the combination of any two sensors or single sensor only. Experimental results show that this method is an effective dynamic gesture recognition method. And compared with the traditional dynamic gesture recognition method, this method has more advantages.

Key words: gesture recognition, surface Electromyogram signal(sEMG), Acceleration signal(ACC), Gyroscope signal(GYRO), Multi-stream Hidden Markov Models(MHMMs)