Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Hong Kong, China.
Sensors (Basel). 2012;12(2):1130-47. doi: 10.3390/s120201130. Epub 2012 Jan 30.
The human hand has multiple degrees of freedom (DOF) for achieving high-dexterity motions. Identifying and replicating human hand motions are necessary to perform precise and delicate operations in many applications, such as haptic applications. Surface electromyography (sEMG) sensors are a low-cost method for identifying hand motions, in addition to the conventional methods that use data gloves and vision detection. The identification of multiple hand motions is challenging because the error rate typically increases significantly with the addition of more hand motions. Thus, the current study proposes two new methods for feature extraction to solve the problem above. The first method is the extraction of the energy ratio features in the time-domain, which are robust and invariant to motion forces and speeds for the same gesture. The second method is the extraction of the concordance correlation features that describe the relationship between every two channels of the multi-channel sEMG sensor system. The concordance correlation features of a multi-channel sEMG sensor system were shown to provide a vast amount of useful information for identification. Furthermore, a new cascaded-structure classifier is also proposed, in which 11 types of hand gestures can be identified accurately using the newly defined features. Experimental results show that the success rate for the identification of the 11 gestures is significantly high.
人手具有多个自由度 (DOF) 以实现高灵巧度的运动。为了在许多应用中执行精确和精细的操作，例如触觉应用，需要识别和复制人手运动。除了使用数据手套和视觉检测的传统方法外，表面肌电图 (sEMG) 传感器是识别手部运动的一种低成本方法。识别多个手部运动具有挑战性，因为随着添加更多的手部运动，错误率通常会显著增加。因此，本研究提出了两种新的特征提取方法来解决上述问题。第一种方法是提取时域中的能量比特征，这些特征对于相同的手势，对运动力和速度具有鲁棒性和不变性。第二种方法是提取描述多通道 sEMG 传感器系统中每两个通道之间关系的一致性相关特征。多通道 sEMG 传感器系统的一致性相关特征被证明可以提供大量有用的信息进行识别。此外，还提出了一种新的级联结构分类器，该分类器可以使用新定义的特征准确识别 11 种手势。实验结果表明，识别 11 种手势的成功率非常高。
