高级检索
    胡昕卉, 李文彬, 阚江明. 面向整枝机控制的手势识别技术研究[J]. 北京林业大学学报, 2017, 39(2): 117-124. DOI: 10.13332/j.1000-1522.20160290
    引用本文: 胡昕卉, 李文彬, 阚江明. 面向整枝机控制的手势识别技术研究[J]. 北京林业大学学报, 2017, 39(2): 117-124. DOI: 10.13332/j.1000-1522.20160290
    HU Xin-hui, LI Wen-bin, KAN Jiang-ming. Gesture control technology based on surface[J]. Journal of Beijing Forestry University, 2017, 39(2): 117-124. DOI: 10.13332/j.1000-1522.20160290
    Citation: HU Xin-hui, LI Wen-bin, KAN Jiang-ming. Gesture control technology based on surface[J]. Journal of Beijing Forestry University, 2017, 39(2): 117-124. DOI: 10.13332/j.1000-1522.20160290

    面向整枝机控制的手势识别技术研究

    Gesture control technology based on surface

    • 摘要: 为了实现对自动立木整枝机的手势控制,需要完成手势识别和无线遥控两部分工作,因此开展了基于表面肌电信号的手势识别技术研究。首先根据整枝机6个工作状态的特点定义相对应的6个手势,即:握拳、上切、下切、外翻、内翻和展拳;然后对采集到的表面肌电信号进行预处理,包括消噪和活动段分割;再对表面肌电信号进行时域和时-频域分析,得到3类特征,即:平均绝对值、自回归参数模型系数和小波分解后的各子频段信号平均能量;最后,构建支持向量机分类器,通过V折交叉验证得到最佳参数,分别进行单用户和多用户的识别实验。手势识别实验结果表明:单用户识别的准确率最高达100.00%,平均准确率为98.07%,高于多用户识别(91.19%)。本研究为实现整枝机手势控制的后续工作奠定了基础,为推进林业机械的智能化与人机交互进程提供了一种有效的新思路。

       

      Abstract: To realize the gesture control for automatic pruning machine needs to complete two parts of work, gesture recognition and wireless remote control and therefore, the technology of hand gesture recognition based on the surface electromyography (sEMG) signal was studied. Firstly, according to six working states of pruning machine, six corresponding gestures were defined, which comprised fist, palm lateral supination, palm lateral pronation, palm supination, palm pronation and finger spread. Secondly, collected sEMG signal was preprocessed, including noise elimination and motion segmentation. The motion signals were then analyzed in time domain and time-frequency domain, and three kinds of features were computed, which were mean absolute value, coefficient of autoregressive parameter model, and average energy of each subband after wavelet decomposition. Lastly, support vector machine classifier was constructed to conduct single-user and multi-user recognition experiments and the best parameters of the classifier were obtained by V-fold cross validation. The results of gesture recognition experiments showed that, the highest accuracy of single-user experiment reached 100% and the average accuracy of single-user experiment (98.07%) was higher than that of multi-user experiment (91.19%). This study lays foundation for the implementation of the follow-up work of gesture control for pruning machine and provides an effective new way to promote the intelligent and human-computer interaction process of forestry machinery.

       

    /

    返回文章
    返回