摘要
为提高肢体运动模式识别率,论文提出了一种经验模态分解与多重分形分析相结合的模式识别方法。先用经验模态分解得到代表肌电信号细节的多层内在模态函数,然后在内在模态函数上进行多重分形分析,提取其广义维数谱,作为肌电信号多模式识别的特征向量。最后以改进的K最近邻分类方法-KNN模型增量学习算法,实现对动作模式的识别。在对张开、合拢及腕内旋、腕外旋4个动作的识别实验中,正确识别率达到了93.0%。结果表明,方法具备一定的实用性,可用于遥操作机器人系统中操作者手腕运动模式识别。
In order to improve the pattern recognition rate of physical movement, a new pattern recognition method has been proposed through the combination of empirical mode decomposition ( EMD ) and multifractal analysis. Firstly, multilayer intrinsic mode functions (IMF), which represent the details of surface electromyography( sEMG), are obtained using EMD. Then multifractal spectrum, which can be used as eigenvector in pattern recognition of sEMG, is extracted from IMF by muhifractal analysis. Finally, the improved K nearest neighbor method-KNN model based incremental learning method is used to recognize various movements of hand. The experiment, designed to classify four hand gestures including hand open, hand grasp, wrist pronation and wrist supination, shows that by using this method, the recognition rate has reached 93.0%, which demonstrates the practicality of this approach and its possible application in the pattern recognition of manipulator's wrist movement.
出处
《传感技术学报》
CAS
CSCD
北大核心
2013年第2期282-288,共7页
Chinese Journal of Sensors and Actuators
基金
国家自然科学基金项目(60903084
61172134)
浙江省自然科学基金项目(Y1111189
LY12F03007
LY12F03006)
浙江省科技计划项目(2010C33075
2012C33075)
关键词
表面肌电信号
模式识别
多重分形分析
经验模态分解
K最近邻模型法
surface electromyography
pattern recognition
multifractal analysis
empirical mode decomposition
k nearest neighbor mode