手部抓取动作特征提取算法研究

针对人手抓取动作问题,如何有效地提取表面肌电信号特征是提高其模式识别率的关键。通过对前人不同手部抓取动作的分类方法及日常生活工作中使用的频度进行统计学分析,决定选取8种抓取动作进行研究。实验显示,随着手部动作姿态种类的增加,基于表面肌电信号的不同特征提取算法分类能力出现不同程度的下降甚至失效。为取得更为理想的抓取动作分类效果,提出将抓取动作分割为预抓取和抓取两个动作过程。选择采集预抓取动作前臂肌电信号,除对其时域、频域及时频域常用特征量进行分析对比外,还增添了对时频域中小波系数最大模值的分析,旨在找出最有效表征肌电信号动作分类的特征量。实验结果表明,小波系数最大模值量最有特征可分性,区分效果比较理想。

Feature extraction algorithm for hand motion modes

Extracting sEMG signal feature availably is the key to improve the pattern recognition rate . Eight kinds of grasping motions were studied according to the frequency statistical analysis on predeces -sors' various classification method for hand grassping motion used in daily life .The experimental results show that with the increase of hand gestures types ,the classification ability of different feature extraction algorithms based on the sEMG became decline in different levels and even failed .The grasping motions were divided into prefetch and grab action process to achieve better results .Prefetch forearm electromyo-graphic signals were chosen ,not only the commonly used features were analyzed and compared in time do-main and frequency domain as well as time and frequency domain , but also maximum modulus of wavelet coefficients was analyzed in time and frequency domain , which was to find out the most effective charac-teristic quantity of sEMG classification .The experimental results show that the maximum modulus of wave-let coefficients has the most characteristic separability , and the classification result is better .