摘要
为了实现在空中手写的人机交互方式,给用户带来一种新型的人机交互体验,设计了一种基于惯性传感器的空中手写轨迹识别系统。系统主要包括数据采集滤波模块、四元数法坐标系转换模块、积分获取测量轨迹模块和神经网络识别模块四部分。本文在原始数据采集和积分获取轨迹两个环节采用卡尔曼滤波算法。为了验证系统的准确性,以在空中书写数字8为例,经卡尔曼滤波后在空间范围内的轨迹完整、清晰,对数字0~9轨迹的捕捉也验证了这一点。设计了AlexNet神经网络迁移学习模块进行轨迹识别,实验结果表明,识别准确率为87.3%,轨迹识别度较高,达到了预期效果。
In order to realize the human-computer interaction of handwriting in the air and bring a new human-computer interaction experience to users,an air handwriting track recognition system based on inertial sensor was designed.The system mainly included four parts of data acquisition and filtering module,quaternion coordinate system conversion module,integration acquisition and measurement track module and neural network identification module.Therefore,the Kalman filter algorithm was used in the two links of original data acquisition and integration acquisition.In order to verify the accuracy of the system,the writing of the number 8 in the air was taken as an example.The trajectory in the space after Kalman filtering was complete and clear,which was also verified by the capture of the number 0~9 trajectory.The AlexNet neural network migration learning module was designed for trajectory recognition.The experimental results show that the recognition accuracy is 87.3%,and the trajectory recognition degree is high,which achieves the expected effect.
作者
李忠虎
狄慧敏
王金明
LI Zhong-hu;DI Hui-min;WANG Jin-ming(School of Information Engineering,Inner Mongolia University of Science and Technology,Baotou 014010,China)
出处
《科学技术与工程》
北大核心
2020年第28期11659-11665,共7页
Science Technology and Engineering
基金
内蒙古自治区自然科学基金(2018LH06001)。
关键词
卡尔曼滤波
惯性轨迹捕捉
四元数
坐标系转换
迁移学习
Kalman filter
inertial trajectory capture
quaternion
coordinate system transformation
transfer learning
