基于多源信息融合的肌电轮椅智能控制技术研究

Research on intelligent control technology of myoelectric wheelchair based on multi-source information fusion

传统手势识别设备很难同时满足识别精度与便携性的要求,同时,基于单一传感器的手势识别策略获取的信息较少,识别结果易受各种因素的影响。而相较于传统的摇杆控制方式,手势控制方式更为便捷,并且可实现远距离操作。基于此背景提出了基于多源信息融合的肌电轮椅智能控制技术研究,设计了一套便携式肌电信号与运动信息采集系统,其中运动信息包括加速度信号和角度信号。利用极限学习机构建识别算法模型,优化了融合分类结构,并以此识别结果控制轮椅运动。实验结果表明:相较于单一传感器手势识别方案,识别准确率提升了6.1%~12.3%,达到了94.7%;相较于传统模式识别方法,分类准确率提升了1.5%~6.0%,并且,轮椅控制系统的在线平均识别率达到了95.2%,满足实时性要求。

The traditional gesture recognition device is difficult to meet the requirements of recognition accuracy and portability at the same time.The gesture recognition strategy of a single sensor acquires less information,so the recognition result is susceptible to various factors.Moreover,compared with the traditional joystick control mode,the gesture control method is more convenient and can be operated at a long distance.Based on this,a research on the intelligent control technology of myoelectric wheelchair based on multi-source information fusion was proposed.A portable electromyogram(EMG)signal and motion information acquisition system was designed.The motion information includes acceleration signal and angle signal.The extreme learning machine was used to construct the recognition algorithm model,and the fusion classification structure was optimized and the wheelchair movement was controlled.The experimental results show that compared with the single sensor gesture recognition scheme,the recognition accuracy of the scheme is improved by 6.1%~12.3%,reaching 94.7%;and,compared with the traditional pattern recognition method,the classification accuracy is improved by 1.5%~6.0%.The controller’s online recognition rate of 9 kinds of actions reaches 95.2%,and meets the real-time requirements.