摘要
针对低成本和便捷的手势识别需求,提出了轴交叉特征码手势识别方法。该方法利用MEMS惯性传感器获取手势的姿态数据,通过互补滤波器融合传感器的加速度和陀螺仪数据,计算手势的加速度矢量。根据手势加速度矢量通过东北天(ENU)坐标系中坐标轴的顺序生成手势轴交叉特征码,最终与手势特征模板库对比,完成手势实时识别。所提方法在用户独立实验中的手势识别准确率可达97.4%,相比动态时间规整(DTW)手势识别方法提升19.8%,并且程序运算的闪存占用小于5 kB,内存占用小于1 kB。实验结果表明:该方法能达到较高的手势识别准确率并仅需占用较小的计算资源,能满足低成本和便捷的人机交互场景。
Aiming at the requirements of low-cost and convenient gesture recognition gesture recognition method with axis-crossing feature codes is proposed.The method MEMS inertial sensors to acquire gesture data,acceleration data and gyroscope data are fused with complementary filter to calculate the acceleration vector of the gesture.Then,the gesture acceleration vector is used to generate gesture axis-crossing feature codes based on the order of the axes in the ENU(East,North,Up)coordinate system,which are eventually compared with the gesture feature template library to achieve real-time gesture recognition.The proposed method has a gesture recognition accuracy of 97.4%in user independent experiments,which is 19.8%higher than the dynamic time warping gesture recognition method,and the flash memory consumption of the program operation is less than 5 kB and the memory usage is less than 1 kB.The experimental results show that the method can achieve a high accuracy rate of gesture recognition and only requires a small amount of computing resources,which can meet the low-cost and convenient human-computer interaction scenario.
作者
廖承鑷
蔡浩原
柳雅倩
王嘉铭
刘春秀
LIAO Chengnie;CAI Haoyuan;LIU Yaqian;WANG Jiaming;LIU Chunxiu(State Key Laboratory of Transducer Technology,Aerospace Information Research Institute,Chinese Academy of Sciences,Beijing 100190,China;University of Chinese Academy of Sciences,Beijing 100049,China)
出处
《传感器与微系统》
北大核心
2025年第3期134-137,142,共5页
Transducer and Microsystem Technologies
基金
国家重点研发计划项目(2020YFC2004501,2020YFC2004503)
国家自然科学基金资助项目(61774157,81771388)
北京自然基金资助项目(4182075)。
关键词
手势识别
惯性测量单元
互补滤波
Hand gesture recognition
inertial measurement unit
complementary filtering
