基于多传感器融合的动态手势识别研究分析

发布时间：2018-03-23 22:04

本文选题：手势识别　切入点：表面肌电信号(s　出处：《计算机工程与应用》2017年17期

【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(s EMG)、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。
[Abstract]:The information fusion of three kinds of sensors (surface electromyograph, gyroscope and accelerometer) is studied in order to improve the classification and accuracy of recognizable dynamic gesture. The three elements of gesture orientation and motion trajectory were characterized by surface electromyography (EMG), gyroscope (gyroscope) and acceleration signal (ACCS), respectively. Multi-stream HMMs is used for dynamic gesture recognition. The experimental results show that the method can effectively recognize dynamic gestures from continuous signals, and the experimental results show that there are five motion trajectories and six static hand shapes. The global average recognition rate obtained by the combination of three kinds of sensors is over 92%. It is obviously higher than the average recognition rate obtained by any combination of two sensors and only using a single sensor. Experiments show that this method is an effective dynamic gesture recognition method and has more advantages than the traditional dynamic gesture recognition method.
【作者单位】：常州大学研究生部;常州大学信息科学与工程学院数理学院;常州大学城市轨道交通学院;
【基金】：国家自然科学基金(No.61201096) 机器人技术与系统国家重点实验室开放基金重点资助项目(No.SKLRS-2010-2D-09,No.SKLRS-2010-MS-10) 江苏省高校自然科学研究面上资助项目(No.10KJB510003) 江苏省自然科学基金(No.BK20140265) 常州市科技资助项目(No.CJ20110023,No.CM20123006)
【分类号】：TP212;TP391.41

【相似文献】